update github action

Import the committed KnowledgeFS snapshot dc4072ee302317145612087ce7440851dc329fd0 under knowledge-fs/ without its Git history, local IDE settings, or build artifacts.

.agents/skills/frontend-testing/SKILL.md

@@ -9,12 +9,11 @@ Use this skill for Vitest work under `web/` and `packages/dify-ui/`. Do not use
 
 ## Required Source
 
-Before writing, changing, or reviewing frontend tests, read `web/docs/test.md` completely. It is the single source of truth. This skill defines the execution workflow and must not add requirements that conflict with or duplicate that guide.
+Before writing, changing, or reviewing frontend tests, read `web/docs/test.md` completely. It is the single source of truth. This skill provides an execution checklist and must not redefine or extend that policy.
 
 ## Workflow
 
 1. Read the source, its behavior owner, nearby specs, and relevant public dependencies.
-1. Identify whether the contract belongs in `web/`, Dify UI Browser Mode, or a styled Storybook test.
 1. Apply the canonical guide to decide whether a test is needed and choose its boundary.
 1. For a behavior change or bug fix, write or identify the failing scenario first when practical.
 1. Implement one coherent scenario at a time and run the focused spec before expanding scope.
@@ -33,4 +32,4 @@ vp test run path/to/spec-or-directory
 vp test run --project unit src/path/to/spec
 ```
 
-For styled Dify UI behavior, run `vp test --project storybook --run`. Run broader checks only after the focused behavior passes.
+Run Dify UI Storybook tests with `vp test --project storybook --run`. Run broader checks only after the focused behavior passes.

.codex/hooks.json

@@ -0,0 +1,15 @@
+{
+  "hooks": {
+    "PreToolUse": [
+      {
+        "matcher": "Bash",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx -y block-no-verify@1.1.1"
+          }
+        ]
+      }
+    ]
+  }
+}

.dockerignore

@@ -8,6 +8,7 @@
 **/*.pyc
 **/.mypy_cache
 **/.ruff_cache
+knowledge-fs/
 .git
 .github
 *.md

.github/dependabot.yml

@@ -1,6 +1,15 @@
 version: 2
 
 updates:
+  - package-ecosystem: "npm"
+    directory: "/knowledge-fs"
+    open-pull-requests-limit: 10
+    schedule:
+      interval: "weekly"
+    groups:
+      knowledge-fs-dependencies:
+        patterns:
+          - "*"
   - package-ecosystem: "uv"
     directory: "/api"
     open-pull-requests-limit: 10

.github/workflows/api-tests.yml

@@ -16,7 +16,7 @@ concurrency:
 jobs:
   api-unit:
     name: API Unit Tests
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     env:
       COVERAGE_FILE: coverage-unit
     defaults:
@@ -75,7 +75,7 @@ jobs:
 
   api-integration:
     name: API Integration Tests
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     env:
       COVERAGE_FILE: coverage-integration
       STORAGE_TYPE: opendal
@@ -129,7 +129,7 @@ jobs:
 
   api-coverage:
     name: API Coverage
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     needs:
       - api-unit
       - api-integration

.github/workflows/build-push.yml

@@ -173,7 +173,7 @@ jobs:
 
   create-manifest:
     needs: build
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     if: github.repository == 'langgenius/dify'
     strategy:
       matrix:

.github/workflows/cli-release.yml

@@ -23,7 +23,7 @@ concurrency:
 jobs:
   validate:
     name: validate manifest + resolve target Dify release
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     if: github.repository == 'langgenius/dify'
     permissions:
       contents: read
@@ -87,7 +87,7 @@ jobs:
   release:
     name: build + attach standalone binaries (all targets)
     needs: validate
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     permissions:
       contents: write
     defaults:

.github/workflows/db-migration-test.yml

@@ -9,7 +9,7 @@ concurrency:
 
 jobs:
   db-migration-test-postgres:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
 
     steps:
       - name: Checkout code
@@ -59,7 +59,7 @@ jobs:
         run: uv run --directory api flask upgrade-db
 
   db-migration-test-mysql:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
 
     steps:
       - name: Checkout code

.github/workflows/deploy-agent.yml

@@ -13,7 +13,7 @@ on:
 
 jobs:
   deploy:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     if: |
       github.event.workflow_run.conclusion == 'success' &&
       github.event.workflow_run.head_branch == 'deploy/agent'

.github/workflows/deploy-dev.yml

@@ -10,7 +10,7 @@ on:
 
 jobs:
   deploy:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     if: |
       github.event.workflow_run.conclusion == 'success' &&
       github.event.workflow_run.head_branch == 'deploy/dev'

.github/workflows/deploy-enterprise.yml

@@ -13,7 +13,7 @@ on:
 
 jobs:
   deploy:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     if: |
       github.event.workflow_run.conclusion == 'success' &&
       github.event.workflow_run.head_branch == 'deploy/enterprise'

.github/workflows/deploy-saas.yml

@@ -13,7 +13,7 @@ on:
 
 jobs:
   deploy:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     if: |
       github.event.workflow_run.conclusion == 'success' &&
       github.event.workflow_run.head_branch == 'deploy/saas'

.github/workflows/hotfix-cherry-pick.yml

@@ -22,7 +22,7 @@ concurrency:
 jobs:
   check-cherry-pick-provenance:
     name: Require cherry-pick provenance
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
         with:

.github/workflows/knowledge-fs-ci.yml

@@ -0,0 +1,504 @@
+name: KnowledgeFS CI
+
+on:
+  pull_request:
+    branches: ["main"]
+  merge_group:
+    branches: ["main"]
+    types: [checks_requested]
+  push:
+    branches: ["main"]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pull-requests: read
+
+concurrency:
+  group: knowledge-fs-${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+
+env:
+  CI: true
+  DIFY_KNOWLEDGE_FS_API_IMAGE_NAME: >-
+    ${{ vars.DIFY_KNOWLEDGE_FS_API_IMAGE_NAME || 'langgenius/dify-knowledge-fs-api' }}
+
+jobs:
+  check-changes:
+    name: Check KnowledgeFS changes
+    runs-on: depot-ubuntu-24.04-4
+    outputs:
+      knowledge-fs: ${{ steps.changes.outputs.knowledge-fs }}
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+        with:
+          fetch-depth: 0
+          persist-credentials: false
+
+      - name: Detect KnowledgeFS changes
+        id: changes
+        uses: dorny/paths-filter@7b450fff21473bca461d4b92ce414b9d0420d706 # v4.0.2
+        with:
+          filters: |
+            knowledge-fs:
+              - 'knowledge-fs/**'
+              - 'knowledge-fs/packages/api/src/dify-capability-v2.ts'
+              - 'knowledge-fs/packages/api/src/knowledge-space-routes.ts'
+              - 'knowledge-fs/packages/api/src/upload-session-routes.ts'
+              - 'knowledge-fs/scripts/export-capability-v2-operations.mjs'
+              - 'knowledge-fs/scripts/export-openapi.mjs'
+              - 'api/dev/generate_knowledge_fs_contract.py'
+              - 'api/dev/knowledge_fs_product_contract.py'
+              - 'api/knowledge-fs-contract.lock.json'
+              - 'api/knowledge-fs-product-operation-gaps.json'
+              - 'api/knowledge-fs-product-operations.json'
+              - 'api/**/knowledge_fs/**'
+              - 'api/**/*knowledge_fs*'
+              - 'api/**/*knowledge-fs*'
+              - 'api/.env.example'
+              - 'api/app_factory.py'
+              - 'api/commands/__init__.py'
+              - 'api/controllers/console/__init__.py'
+              - 'api/controllers/console/workspace/rbac.py'
+              - 'api/controllers/service_api/__init__.py'
+              - 'api/core/agent/base_agent_runner.py'
+              - 'api/core/app/apps/agent_app/runtime_request_builder.py'
+              - 'api/core/rbac/entities.py'
+              - 'api/core/tools/__base/tool_runtime.py'
+              - 'api/core/tools/builtin_tool/_position.yaml'
+              - 'api/core/workflow/node_runtime.py'
+              - 'api/core/workflow/nodes/agent_v2/runtime_request_builder.py'
+              - 'api/extensions/ext_celery.py'
+              - 'api/extensions/ext_commands.py'
+              - 'api/models/__init__.py'
+              - 'api/services/account_service.py'
+              - 'api/services/agent_tool_inner_service.py'
+              - 'api/services/enterprise/rbac_service.py'
+              - 'api/services/entities/agent_tool_inner.py'
+              - 'api/services/knowledge_fs/**'
+              - 'api/services/knowledge_fs_capability.py'
+              - 'api/tests/unit_tests/dev/test_generate_knowledge_fs_contract.py'
+              - 'api/tests/unit_tests/controllers/console/workspace/test_rbac.py'
+              - 'api/tests/unit_tests/core/agent/test_base_agent_runner.py'
+              - 'api/tests/unit_tests/core/app/apps/agent_app/test_runtime_request_builder.py'
+              - 'api/tests/unit_tests/core/workflow/nodes/agent_v2/test_runtime_request_builder.py'
+              - 'api/tests/unit_tests/core/workflow/nodes/tool/test_tool_node_runtime.py'
+              - 'api/tests/unit_tests/core/workflow/test_node_runtime.py'
+              - 'api/tests/unit_tests/services/enterprise/test_rbac_service.py'
+              - 'api/tests/unit_tests/services/test_account_service.py'
+              - 'api/tests/unit_tests/services/test_agent_tool_inner_service.py'
+              - 'api/tests/unit_tests/services/test_knowledge_fs_capability.py'
+              - 'api/tests/unit_tests/services/test_knowledge_fs_product_operations.py'
+              - 'api/pyproject.toml'
+              - 'api/uv.lock'
+              - 'dify-agent/src/dify_agent/layers/dify_core_tools/client.py'
+              - 'dify-agent/tests/local/dify_agent/layers/dify_core_tools/test_client.py'
+              - 'packages/contracts/generated/api/console/**'
+              - 'packages/contracts/generated/api/service/**'
+              - 'docker/.env.example'
+              - 'docker/README.md'
+              - 'docker/dify-env-sync.py'
+              - 'docker/dify-env-sync.sh'
+              - 'docker/docker-compose-template.yaml'
+              - 'docker/docker-compose.yaml'
+              - 'docker/envs/core-services/api.env.example'
+              - 'docker/envs/core-services/knowledge-fs.env.example'
+              - 'docker/generate_docker_compose'
+              - 'docs/design/knowledge-fs*'
+              - '.github/dependabot.yml'
+              - '.github/workflows/knowledge-fs-ci.yml'
+
+  build:
+    name: Build KnowledgeFS API production image
+    needs: check-changes
+    if: needs.check-changes.outputs.knowledge-fs == 'true' || github.event_name == 'workflow_dispatch'
+    runs-on: depot-ubuntu-24.04-4
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+        with:
+          persist-credentials: false
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@bb05f3f5519dd87d3ba754cc423b652a5edd6d2c # v4.2.0
+
+      - name: Login to Docker Hub
+        if: github.event_name == 'workflow_dispatch' || (github.event_name == 'push' && github.ref == 'refs/heads/main')
+        uses: docker/login-action@af1e73f918a031802d376d3c8bbc3fe56130a9b0 # v4.4.0
+        with:
+          username: ${{ secrets.DOCKERHUB_USER }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: Extract KnowledgeFS image metadata
+        id: meta
+        uses: docker/metadata-action@dc802804100637a589fabce1cb79ff13a1411302 # v6.2.0
+        with:
+          images: ${{ env.DIFY_KNOWLEDGE_FS_API_IMAGE_NAME }}
+          tags: |
+            type=raw,value=latest,enable=${{ github.ref == 'refs/heads/main' }}
+            type=ref,event=branch
+            type=sha,format=long
+
+      - name: Build KnowledgeFS API image
+        uses: docker/build-push-action@53b7df96c91f9c12dcc8a07bcb9ccacbed38856a # v7.3.0
+        with:
+          context: ./knowledge-fs
+          file: ./knowledge-fs/apps/api/Dockerfile
+          labels: ${{ steps.meta.outputs.labels }}
+          platforms: linux/amd64
+          push: ${{ github.event_name == 'workflow_dispatch' || (github.event_name == 'push' && github.ref == 'refs/heads/main') }}
+          tags: ${{ steps.meta.outputs.tags }}
+
+  quality:
+    name: Run KnowledgeFS quality and contract gates
+    needs: check-changes
+    if: needs.check-changes.outputs.knowledge-fs == 'true' || github.event_name == 'workflow_dispatch'
+    runs-on: depot-ubuntu-24.04-4
+    defaults:
+      run:
+        shell: bash
+        working-directory: ./knowledge-fs
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+        with:
+          fetch-depth: 0
+          persist-credentials: false
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@0ebf47130e4866e96fce0953f49152a61190b271 # v6.0.9
+        with:
+          package_json_file: knowledge-fs/package.json
+          run_install: false
+
+      - name: Setup Node
+        uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7.0.0
+        with:
+          node-version: 22
+          cache: pnpm
+          cache-dependency-path: knowledge-fs/pnpm-lock.yaml
+
+      - name: Install KnowledgeFS dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Scan KnowledgeFS secrets
+        run: pnpm security:secrets
+
+      - name: Audit KnowledgeFS production dependencies
+        run: pnpm security:dependencies
+
+      - name: Run KnowledgeFS checks
+        run: pnpm check
+
+      - name: Build KnowledgeFS
+        run: pnpm build
+
+      - name: Lint KnowledgeFS
+        run: pnpm lint
+
+      - name: Setup UV and Python
+        uses: astral-sh/setup-uv@11f9893b081a58869d3b5fccaea48c9e9e46f990 # v8.3.2
+        with:
+          enable-cache: true
+          python-version: "3.12"
+          cache-dependency-glob: |
+            api/uv.lock
+            dify-agent/uv.lock
+
+      - name: Verify Dify dependency lock
+        working-directory: .
+        run: uv lock --project api --check
+
+      - name: Install Dify contract dependencies
+        working-directory: .
+        run: uv sync --project api --locked --dev
+
+      - name: Collect Dify KnowledgeFS gate targets
+        working-directory: .
+        run: |
+          set -euo pipefail
+
+          target_dir="${RUNNER_TEMP:?}/knowledge-fs-ci-targets"
+          mkdir -p "$target_dir"
+
+          production_targets=()
+          add_production_target() {
+            local path="$1"
+            if [[ ! -f "$path" ]]; then
+              echo "required Dify KnowledgeFS production target is missing: $path" >&2
+              exit 1
+            fi
+            production_targets+=("$path")
+          }
+
+          while IFS= read -r -d '' path; do
+            if [[ "$path" == *knowledge_fs* ]]; then
+              add_production_target "$path"
+            fi
+          done < <(
+            find api \
+              \( -path 'api/.venv' -o -path 'api/tests' -o -path 'api/storage' \) -prune \
+              -o -type f -name '*.py' -print0
+          )
+
+          production_touchpoints=(
+            api/app_factory.py
+            api/commands/__init__.py
+            api/controllers/console/__init__.py
+            api/controllers/console/workspace/rbac.py
+            api/controllers/service_api/__init__.py
+            api/core/agent/base_agent_runner.py
+            api/core/app/apps/agent_app/runtime_request_builder.py
+            api/core/rbac/entities.py
+            api/core/tools/__base/tool_runtime.py
+            api/core/workflow/node_runtime.py
+            api/core/workflow/nodes/agent_v2/runtime_request_builder.py
+            api/extensions/ext_celery.py
+            api/extensions/ext_commands.py
+            api/models/__init__.py
+            api/services/account_service.py
+            api/services/agent_tool_inner_service.py
+            api/services/enterprise/rbac_service.py
+            api/services/entities/agent_tool_inner.py
+          )
+          for path in "${production_touchpoints[@]}"; do
+            add_production_target "$path"
+          done
+          printf '%s\0' "${production_touchpoints[@]}" > "$target_dir/glue-files"
+
+          if ((${#production_targets[@]} == 0)); then
+            echo "Dify KnowledgeFS production target set is empty" >&2
+            exit 1
+          fi
+          printf '%s\0' "${production_targets[@]}" > "$target_dir/production-files"
+
+          test_targets=()
+          add_test_target() {
+            local path="$1"
+            if [[ ! -f "$path" ]]; then
+              echo "required Dify KnowledgeFS unit test is missing: $path" >&2
+              exit 1
+            fi
+            test_targets+=("$path")
+          }
+
+          while IFS= read -r -d '' path; do
+            if [[ "$path" == *knowledge_fs* ]]; then
+              add_test_target "$path"
+            fi
+          done < <(find api/tests/unit_tests -type f -name '*.py' -print0)
+
+          test_touchpoints=(
+            api/tests/unit_tests/controllers/console/workspace/test_rbac.py
+            api/tests/unit_tests/core/agent/test_base_agent_runner.py
+            api/tests/unit_tests/core/app/apps/agent_app/test_runtime_request_builder.py
+            api/tests/unit_tests/core/workflow/nodes/agent_v2/test_runtime_request_builder.py
+            api/tests/unit_tests/core/workflow/nodes/tool/test_tool_node_runtime.py
+            api/tests/unit_tests/core/workflow/test_node_runtime.py
+            api/tests/unit_tests/services/enterprise/test_rbac_service.py
+            api/tests/unit_tests/services/test_account_service.py
+            api/tests/unit_tests/services/test_agent_tool_inner_service.py
+          )
+          for path in "${test_touchpoints[@]}"; do
+            add_test_target "$path"
+          done
+
+          required_test_scopes=(
+            /commands/
+            /configs/
+            /controllers/
+            /core/agent/
+            /core/app/apps/agent_app/
+            /core/tools/builtin_tool/providers/knowledge_fs/
+            /core/workflow/
+            /dev/
+            /extensions/
+            /migrations/
+            /models/
+            /repositories/
+            /services/
+            /tasks/
+          )
+          for required_scope in "${required_test_scopes[@]}"; do
+            scope_found=false
+            for path in "${test_targets[@]}"; do
+              if [[ "$path" == *"$required_scope"* ]]; then
+                scope_found=true
+                break
+              fi
+            done
+            if [[ "$scope_found" != true ]]; then
+              echo "required Dify KnowledgeFS test scope is empty: $required_scope" >&2
+              exit 1
+            fi
+          done
+
+          if ((${#test_targets[@]} == 0)); then
+            echo "Dify KnowledgeFS unit test target set is empty" >&2
+            exit 1
+          fi
+          printf '%s\0' "${test_targets[@]}" > "$target_dir/unit-test-files"
+
+      - name: Lint Dify KnowledgeFS integration
+        working-directory: .
+        run: |
+          set -euo pipefail
+          targets=()
+          while IFS= read -r -d '' path; do
+            targets+=("$path")
+          done < "${RUNNER_TEMP:?}/knowledge-fs-ci-targets/production-files"
+          if ((${#targets[@]} == 0)); then
+            echo "Dify KnowledgeFS production target manifest is empty" >&2
+            exit 1
+          fi
+          uv run --project api --dev ruff format --check "${targets[@]}"
+          uv run --project api --dev ruff check "${targets[@]}"
+
+      - name: Type-check Dify KnowledgeFS integration
+        working-directory: .
+        run: |
+          set -euo pipefail
+          targets=()
+          while IFS= read -r -d '' path; do
+            targets+=("$path")
+          done < "${RUNNER_TEMP:?}/knowledge-fs-ci-targets/production-files"
+          if ((${#targets[@]} == 0)); then
+            echo "Dify KnowledgeFS production target manifest is empty" >&2
+            exit 1
+          fi
+          PYREFLY_OUTPUT_FORMAT=github ./dev/pyrefly-check-local "${targets[@]}"
+
+          mypy_targets=()
+          for path in "${targets[@]}"; do
+            if [[ "$path" != api/migrations/* ]]; then
+              mypy_targets+=("${path#api/}")
+            fi
+          done
+          if ((${#mypy_targets[@]} == 0)); then
+            echo "Dify KnowledgeFS Mypy target set is empty" >&2
+            exit 1
+          fi
+          uv run --directory api --dev mypy \
+            --explicit-package-bases \
+            --exclude-gitignore \
+            --exclude '(^|/)conftest\.py$' \
+            --exclude 'tests/' \
+            --exclude 'migrations/' \
+            --check-untyped-defs \
+            --disable-error-code=import-untyped \
+            "${mypy_targets[@]}"
+
+      - name: Test Dify KnowledgeFS unit surface
+        working-directory: .
+        env:
+          COVERAGE_FILE: ${{ runner.temp }}/dify-knowledge-fs.coverage
+        run: |
+          set -euo pipefail
+          targets=()
+          while IFS= read -r -d '' path; do
+            targets+=("$path")
+          done < "${RUNNER_TEMP:?}/knowledge-fs-ci-targets/unit-test-files"
+          if ((${#targets[@]} == 0)); then
+            echo "Dify KnowledgeFS unit test target manifest is empty" >&2
+            exit 1
+          fi
+          uv run --project api --dev coverage run --branch --source=api -m pytest "${targets[@]}" --no-cov -q
+
+      - name: Enforce Dify KnowledgeFS focused coverage
+        working-directory: .
+        env:
+          COVERAGE_FILE: ${{ runner.temp }}/dify-knowledge-fs.coverage
+          KNOWLEDGE_FS_COVERAGE_BASE: >-
+            ${{ github.event.pull_request.base.sha || github.event.merge_group.base_sha || github.event.before || '' }}
+        run: |
+          set -euo pipefail
+          report="${RUNNER_TEMP:?}/dify-knowledge-fs-coverage.json"
+          uv run --project api --dev coverage json --show-contexts -o "$report"
+          uv run --project api --dev python api/dev/check_knowledge_fs_coverage.py \
+            --coverage-json "$report" \
+            --glue-manifest "${RUNNER_TEMP:?}/knowledge-fs-ci-targets/glue-files" \
+            --base "$KNOWLEDGE_FS_COVERAGE_BASE" \
+            --minimum 90 \
+            --glue-minimum 90
+
+      - name: Verify Dify KnowledgeFS contract
+        working-directory: .
+        run: uv run --project api python api/dev/generate_knowledge_fs_contract.py --check
+
+      - name: Verify Dify Agent dependency lock
+        working-directory: .
+        run: uv lock --project dify-agent --check
+
+      - name: Install Dify Agent gate dependencies
+        working-directory: .
+        run: uv sync --project dify-agent --locked --dev
+
+      - name: Lint Dify Agent KnowledgeFS integration
+        working-directory: ./dify-agent
+        run: |
+          uv run --project . --dev ruff format --check \
+            src/dify_agent/layers/dify_core_tools/client.py \
+            tests/local/dify_agent/layers/dify_core_tools/test_client.py
+          uv run --project . --dev ruff check \
+            src/dify_agent/layers/dify_core_tools/client.py \
+            tests/local/dify_agent/layers/dify_core_tools/test_client.py
+
+      - name: Type-check Dify Agent KnowledgeFS integration
+        working-directory: ./dify-agent
+        run: >-
+          uv run --project . --dev basedpyright --level error
+          src/dify_agent/layers/dify_core_tools/client.py
+          tests/local/dify_agent/layers/dify_core_tools/test_client.py
+
+      - name: Test Dify Agent KnowledgeFS integration
+        working-directory: ./dify-agent
+        run: >-
+          uv run --project . --dev python -m pytest
+          tests/local/dify_agent/layers/dify_core_tools/test_client.py
+          -q
+
+  skip:
+    name: Skip KnowledgeFS quality and contract gates
+    needs: check-changes
+    if: needs.check-changes.outputs.knowledge-fs != 'true' && github.event_name != 'workflow_dispatch'
+    runs-on: depot-ubuntu-24.04-4
+    steps:
+      - name: Report skipped KnowledgeFS checks
+        run: echo "No KnowledgeFS-related changes detected; skipping KnowledgeFS checks."
+
+  final:
+    name: KnowledgeFS CI
+    if: ${{ always() }}
+    needs:
+      - check-changes
+      - build
+      - quality
+      - skip
+    runs-on: depot-ubuntu-24.04-4
+    steps:
+      - name: Finalize KnowledgeFS CI status
+        env:
+          EVENT_NAME: ${{ github.event_name }}
+          BUILD_RESULT: ${{ needs.build.result }}
+          KNOWLEDGE_FS_CHANGED: ${{ needs.check-changes.outputs.knowledge-fs }}
+          QUALITY_RESULT: ${{ needs.quality.result }}
+          SKIP_RESULT: ${{ needs.skip.result }}
+        run: |
+          if [[ "$EVENT_NAME" == 'workflow_dispatch' || "$KNOWLEDGE_FS_CHANGED" == 'true' ]]; then
+            if [[ "$BUILD_RESULT" == 'success' && "$QUALITY_RESULT" == 'success' ]]; then
+              echo "KnowledgeFS build and checks ran successfully."
+              exit 0
+            fi
+
+            echo "KnowledgeFS build or checks failed: build=$BUILD_RESULT quality=$QUALITY_RESULT" >&2
+            exit 1
+          fi
+
+          if [[ "$SKIP_RESULT" == 'success' ]]; then
+            echo "KnowledgeFS checks were skipped because no related files changed."
+            exit 0
+          fi
+
+          echo "KnowledgeFS change detection or skip reporting failed with result: $SKIP_RESULT" >&2
+          exit 1

.github/workflows/labeler.yml

@@ -7,7 +7,7 @@ jobs:
     permissions:
       contents: read
       pull-requests: write
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - uses: actions/labeler@b8dd2d9be0f68b860e7dae5dae7d772984eacd6d # v6.2.0
         with:

.github/workflows/main-ci.yml

@@ -21,13 +21,13 @@ concurrency:
 jobs:
   pre_job:
     name: Skip Duplicate Checks
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     outputs:
       should_skip: ${{ steps.skip_check.outputs.should_skip || 'false' }}
     steps:
       - id: skip_check
         continue-on-error: true
-        uses: fkirc/skip-duplicate-actions@f75f66ce1886f00957d99748a42c724f4330bdcf # v5.3.1
+        uses: fkirc/skip-duplicate-actions@b974a9395958c231af965b70070979a577efa578 # v5.3.2
         with:
           cancel_others: 'true'
           concurrent_skipping: same_content_newer
@@ -37,7 +37,7 @@ jobs:
     name: Check Changed Files
     needs: pre_job
     if: needs.pre_job.outputs.should_skip != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     outputs:
       api-changed: ${{ steps.changes.outputs.api }}
       cli-changed: ${{ steps.changes.outputs.cli }}
@@ -164,7 +164,7 @@ jobs:
       - pre_job
       - check-changes
     if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.api-changed != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Report skipped API tests
         run: echo "No API-related changes detected; skipping API tests."
@@ -177,7 +177,7 @@ jobs:
       - check-changes
       - api-tests-run
       - api-tests-skip
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Finalize API Tests status
         env:
@@ -224,7 +224,7 @@ jobs:
       - pre_job
       - check-changes
     if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.cli-changed != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Report skipped CLI tests
         run: echo "No CLI-related changes detected; skipping CLI tests."
@@ -237,7 +237,7 @@ jobs:
       - check-changes
       - cli-tests-run
       - cli-tests-skip
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Finalize CLI Tests status
         env:
@@ -284,7 +284,7 @@ jobs:
       - pre_job
       - check-changes
     if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.web-changed != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Report skipped web tests
         run: echo "No web-related changes detected; skipping web tests."
@@ -297,7 +297,7 @@ jobs:
       - check-changes
       - web-tests-run
       - web-tests-skip
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Finalize Web Tests status
         env:
@@ -344,7 +344,7 @@ jobs:
       - pre_job
       - check-changes
     if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.e2e-changed != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Report skipped web full-stack e2e
         run: echo "No E2E-related changes detected; skipping web full-stack E2E."
@@ -357,7 +357,7 @@ jobs:
       - check-changes
       - web-e2e-run
       - web-e2e-skip
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Finalize Web Full-Stack E2E status
         env:
@@ -411,7 +411,7 @@ jobs:
       - pre_job
       - check-changes
     if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.vdb-changed != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Report skipped VDB tests
         run: echo "No VDB-related changes detected; skipping VDB tests."
@@ -424,7 +424,7 @@ jobs:
       - check-changes
       - vdb-tests-run
       - vdb-tests-skip
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Finalize VDB Tests status
         env:
@@ -470,7 +470,7 @@ jobs:
       - pre_job
       - check-changes
     if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.migration-changed != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Report skipped DB migration tests
         run: echo "No migration-related changes detected; skipping DB migration tests."
@@ -483,7 +483,7 @@ jobs:
       - check-changes
       - db-migration-test-run
       - db-migration-test-skip
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Finalize DB Migration Test status
         env:
@@ -530,7 +530,7 @@ jobs:
       - pre_job
       - check-changes
     if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.sandbox-runtime-changed != 'true'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Report skipped sandbox runtime tests
         run: echo "No sandbox-runtime-related changes detected; skipping sandbox runtime tests."
@@ -543,7 +543,7 @@ jobs:
       - check-changes
       - sandbox-runtime-tests-run
       - sandbox-runtime-tests-skip
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Finalize Sandbox Runtime Tests status
         env:

.github/workflows/post-merge.yml

@@ -14,7 +14,7 @@ concurrency:
 jobs:
   check-changes:
     name: Check Changed Files
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     outputs:
       external-e2e-changed: ${{ steps.changes.outputs.external_e2e }}
     steps:

.github/workflows/pyrefly-diff-comment.yml

@@ -12,7 +12,7 @@ permissions: {}
 jobs:
   comment:
     name: Comment PR with pyrefly diff
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     permissions:
       actions: read
       contents: read

.github/workflows/pyrefly-diff.yml

@@ -10,7 +10,7 @@ permissions:
 
 jobs:
   pyrefly-diff:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     permissions:
       contents: read
       issues: write

.github/workflows/pyrefly-type-coverage-comment.yml

@@ -12,7 +12,7 @@ permissions: {}
 jobs:
   comment:
     name: Comment PR with type coverage
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     permissions:
       actions: read
       contents: read

.github/workflows/pyrefly-type-coverage.yml

@@ -10,7 +10,7 @@ permissions:
 
 jobs:
   pyrefly-type-coverage:
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     permissions:
       contents: read
       issues: write

.github/workflows/sandbox-runtime-tests.yml

@@ -13,7 +13,7 @@ concurrency:
 jobs:
   sandbox-runtime-unit:
     name: Sandbox Runtime Unit Tests
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     defaults:
       run:
         shell: bash
@@ -27,7 +27,7 @@ jobs:
           persist-credentials: false
 
       - name: Setup Go
-        uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
+        uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7.0.0
         with:
           go-version-file: dify-agent-runtime/go.mod
           cache-dependency-path: dify-agent-runtime/go.sum
@@ -37,7 +37,7 @@ jobs:
 
   sandbox-runtime-lint:
     name: Sandbox Runtime Lint
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     defaults:
       run:
         shell: bash
@@ -51,20 +51,20 @@ jobs:
           persist-credentials: false
 
       - name: Setup Go
-        uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
+        uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7.0.0
         with:
           go-version-file: dify-agent-runtime/go.mod
           cache-dependency-path: dify-agent-runtime/go.sum
 
       - name: Run golangci-lint
-        uses: golangci/golangci-lint-action@4afd733a84b1f43292c63897423277bb7f4313a9 # v6.5.0
+        uses: golangci/golangci-lint-action@ba0d7d2ec06a0ea1cb5fa41b2e4a3ab91d21278a # v6.5.0
         with:
           working-directory: dify-agent-runtime
           version: latest
 
   sandbox-runtime-integration:
     name: Sandbox Runtime Integration Tests
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     defaults:
       run:
         shell: bash
@@ -78,7 +78,7 @@ jobs:
           persist-credentials: false
 
       - name: Setup Go
-        uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
+        uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7.0.0
         with:
           go-version-file: dify-agent-runtime/go.mod
           cache-dependency-path: dify-agent-runtime/go.sum

.github/workflows/semantic-pull-request.yml

@@ -16,7 +16,7 @@ jobs:
     name: Validate PR title
     permissions:
       pull-requests: read
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     steps:
       - name: Complete merge group check
         if: github.event_name == 'merge_group'

.github/workflows/stale.yml

@@ -12,7 +12,7 @@ on:
 jobs:
   stale:
 
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     permissions:
       issues: write
       pull-requests: write

.github/workflows/style.yml

@@ -19,7 +19,7 @@ permissions:
 jobs:
   python-style:
     name: Python Style
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
 
     steps:
       - name: Checkout code
@@ -83,7 +83,7 @@ jobs:
 
   web-style:
     name: Web Style
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     defaults:
       run:
         working-directory: ./web
@@ -182,7 +182,7 @@ jobs:
 
   superlinter:
     name: SuperLinter
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
 
     steps:
       - name: Checkout code

.github/workflows/tool-test-sdks.yaml

@@ -17,7 +17,7 @@ concurrency:
 jobs:
   build:
     name: unit test for Node.js SDK
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
 
     defaults:
       run:
@@ -29,7 +29,7 @@ jobs:
           persist-credentials: false
 
       - name: Use Node.js
-        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
+        uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7.0.0
         with:
           node-version: 22
           cache: ''

.github/workflows/translate-i18n-claude.yml

@@ -35,7 +35,7 @@ concurrency:
 jobs:
   translate:
     if: github.repository == 'langgenius/dify'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     timeout-minutes: 120
 
     steps:
@@ -158,7 +158,7 @@ jobs:
 
       - name: Run Claude Code for Translation Sync
         if: steps.context.outputs.CHANGED_FILES != ''
-        uses: anthropics/claude-code-action@e90deca47693f9457b72f2b53c17d7c445a87342 # v1.0.171
+        uses: anthropics/claude-code-action@af0559ee4f514d1ef21826982bed13f7edc3c35e # v1.0.178
         with:
           anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
           github_token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/trigger-i18n-sync.yml

@@ -16,7 +16,7 @@ concurrency:
 jobs:
   trigger:
     if: github.repository == 'langgenius/dify'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     timeout-minutes: 5
 
     steps:

.github/workflows/vdb-tests-full.yml

@@ -16,7 +16,7 @@ jobs:
   test:
     name: Full VDB Tests
     if: github.repository == 'langgenius/dify'
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     strategy:
       matrix:
         python-version:

.github/workflows/vdb-tests.yml

@@ -13,7 +13,7 @@ concurrency:
 jobs:
   test:
     name: VDB Smoke Tests
-    runs-on: depot-ubuntu-24.04
+    runs-on: depot-ubuntu-24.04-4
     strategy:
       matrix:
         python-version:

.gitignore

@@ -30,6 +30,11 @@ share/python-wheels/
 *.egg
 MANIFEST
 
+# KnowledgeFS is an independently rooted TypeScript workspace. Its admin `lib`
+# directory contains source files rather than Python build output.
+!/knowledge-fs/apps/admin/lib/
+!/knowledge-fs/apps/admin/lib/**
+
 # PyInstaller
 #  Usually these files are written by a python script from a template
 #  before PyInstaller builds the exe, so as to inject date/other infos into it.

api/.env.example

@@ -681,6 +681,31 @@ AGENT_BACKEND_STREAM_READ_TIMEOUT_SECONDS=30
 AGENT_BACKEND_STREAM_MAX_RECONNECTS=3
 AGENT_BACKEND_RUN_TIMEOUT_SECONDS=1200
 
+# KnowledgeFS (Dataset 2.0)
+KNOWLEDGE_FS_ENABLED=false
+# Production deployments require HTTPS; plain HTTP is limited to non-production or loopback.
+KNOWLEDGE_FS_BASE_URL=
+KNOWLEDGE_FS_DIRECT_ORIGIN=
+KNOWLEDGE_FS_LIFECYCLE_WORKER_ENABLED=false
+KNOWLEDGE_FS_INTEGRATED_PROVISION_READY=false
+KNOWLEDGE_FS_LEGACY_ACL_FREEZE_READY=false
+KNOWLEDGE_FS_LIFECYCLE_POLL_INTERVAL_SECONDS=15
+KNOWLEDGE_FS_LIFECYCLE_LEASE_SECONDS=60
+KNOWLEDGE_FS_LIFECYCLE_BATCH_SIZE=25
+# Legacy rollback-only HMAC; Capability v2 deployments leave this blank.
+KNOWLEDGE_FS_JWT_SECRET=
+KNOWLEDGE_FS_CAPABILITY_V2_ENABLED=false
+KNOWLEDGE_FS_CAPABILITY_V2_SIGNING_KID=
+KNOWLEDGE_FS_CAPABILITY_V2_PRIVATE_KEY_PEM=
+KNOWLEDGE_FS_CAPABILITY_V2_PREVIOUS_PUBLIC_JWKS=
+KNOWLEDGE_FS_CAPABILITY_V2_ISSUER=dify-control-plane
+KNOWLEDGE_FS_CAPABILITY_V2_AUDIENCE=knowledge-fs
+KNOWLEDGE_FS_CAPABILITY_V2_MAX_TTL_SECONDS=60
+KNOWLEDGE_FS_SSE_READ_TIMEOUT_SECONDS=300
+KNOWLEDGE_FS_TIMEOUT_SECONDS=10
+KNOWLEDGE_FS_JWKS_CACHE_MAX_AGE_SECONDS=300
+KNOWLEDGE_FS_PRODUCT_MAX_RESPONSE_BYTES=4194304
+
 # Marketplace configuration
 MARKETPLACE_ENABLED=true
 MARKETPLACE_API_URL=https://marketplace.dify.ai

api/app_factory.py

@@ -154,6 +154,7 @@ def initialize_extensions(app: DifyApp):
         ext_forward_refs,
         ext_hosting_provider,
         ext_import_modules,
+        ext_knowledge_fs_observability,
         ext_logging,
         ext_login,
         ext_logstore,
@@ -204,6 +205,7 @@ def initialize_extensions(app: DifyApp):
         ext_enterprise_telemetry,
         ext_request_logging,
         ext_session_factory,
+        ext_knowledge_fs_observability,
         ext_oauth_bearer,
     ]
     for ext in extensions:

api/commands/__init__.py

@@ -10,6 +10,7 @@ from .data_migration import (
     import_migration_data,
     migration_data_wizard,
 )
+from .knowledge_fs import knowledge_fs_control_space
 from .plugin import (
     backfill_plugin_auto_upgrade,
     extract_plugins,
@@ -75,6 +76,7 @@ __all__ = [
     "import_migration_data",
     "install_plugins",
     "install_rag_pipeline_plugins",
+    "knowledge_fs_control_space",
     "legacy_model_types",
     "migrate_annotation_vector_database",
     "migrate_data_for_plugin",

api/commands/knowledge_fs.py

@@ -0,0 +1,604 @@
+"""Operator commands for the independent KnowledgeFS control-plane."""
+
+from __future__ import annotations
+
+import json
+from collections.abc import Callable
+from datetime import datetime
+from functools import partial
+from pathlib import Path
+
+import click
+from pydantic import BaseModel, ValidationError
+
+from core.db.session_factory import session_factory
+from services.knowledge_fs.cleanup import (
+    CleanupApprovalInput,
+    CleanupCompletionEvidenceInput,
+    CleanupReadinessEvidenceInput,
+    CleanupStartInput,
+    KnowledgeFSCleanupError,
+    KnowledgeFSCleanupService,
+)
+from services.knowledge_fs.control_space_commands import KnowledgeFSControlSpaceCommandService
+from services.knowledge_fs.control_space_lifecycle import KnowledgeFSControlSpaceLifecycleError
+from services.knowledge_fs.control_space_management import (
+    KnowledgeFSControlSpaceManagementService,
+    KnowledgeFSControlSpaceRegistration,
+)
+from services.knowledge_fs.cutover import (
+    CutoverSmokeResultsInput,
+    FinalDeltaInput,
+    KnowledgeFSCutoverError,
+    KnowledgeFSWorkspaceCutoverService,
+    LegacyDependencyInput,
+    QuarantineResolutionInput,
+    ShadowAuthorizationObservationInput,
+    ShadowCompletionInput,
+    WorkspaceInventoryInput,
+)
+from services.knowledge_fs.orphan_reconciler import KnowledgeFSOrphanReconciler
+from services.knowledge_fs.remote_registry import get_knowledge_fs_lifecycle_remote
+
+
+@click.group("knowledge-fs-control-space")
+def knowledge_fs_control_space() -> None:
+    """Inspect and repair Dify-owned KnowledgeFS control-space state."""
+
+
+@knowledge_fs_control_space.command("dry-run")
+@click.option("--tenant-id", default=None)
+def dry_run(tenant_id: str | None) -> None:
+    report = _management_service().dry_run(tenant_id=tenant_id)
+    click.echo(json.dumps(report._asdict(), sort_keys=True))
+
+
+@knowledge_fs_control_space.command("inventory")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Create ledgers; omitted means read-only inventory.")
+def inventory(input_path: Path, apply: bool) -> None:
+    """Validate strict Workspace inventory JSONL and optionally create ledgers."""
+
+    service = _cutover_service()
+    for payload in _read_jsonl(input_path, WorkspaceInventoryInput):
+        report = _operator_call(partial(service.inventory, payload, apply=apply))
+        _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("register")
+@click.option("--tenant-id", required=True)
+@click.option("--owner-account-id", required=True)
+@click.option("--provisioning-key", required=True)
+@click.option("--knowledge-space-id", required=True)
+@click.option("--knowledge-space-revision", type=click.IntRange(min=0), required=True)
+def register(
+    tenant_id: str,
+    owner_account_id: str,
+    provisioning_key: str,
+    knowledge_space_id: str,
+    knowledge_space_revision: int,
+) -> None:
+    control_space, replayed = _management_service().register(
+        KnowledgeFSControlSpaceRegistration(
+            tenant_id,
+            owner_account_id,
+            provisioning_key,
+            knowledge_space_id,
+            knowledge_space_revision,
+        )
+    )
+    click.echo(json.dumps({"control_space_id": control_space.id, "replayed": replayed}, sort_keys=True))
+
+
+@knowledge_fs_control_space.command("backfill")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist registrations; omitted means dry-run.")
+def backfill(input_path: Path, apply: bool) -> None:
+    """Backfill strict Workspace inventory JSONL; dry-run unless --apply is explicit."""
+
+    service = _cutover_service()
+    for payload in _read_jsonl(input_path, WorkspaceInventoryInput):
+        report = _operator_call(partial(service.backfill, payload, apply=apply))
+        _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("quarantine-resolve")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist resolutions; omitted means dry-run.")
+def quarantine_resolve(input_path: Path, apply: bool) -> None:
+    """Resolve strict tenant-scoped quarantine JSONL with immutable operator evidence."""
+
+    service = _cutover_service()
+    for payload in _read_jsonl(input_path, QuarantineResolutionInput):
+        report = _operator_call(partial(service.resolve_quarantine, payload, apply=apply))
+        _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("shadow-start")
+@click.option("--tenant-id", required=True)
+@click.option("--expected-cas-version", type=click.IntRange(min=0), required=True)
+@click.option("--at", "started_at", default=None, help="Optional explicit timezone-aware shadow start.")
+def shadow_start(tenant_id: str, expected_cas_version: int, started_at: str | None) -> None:
+    service = _cutover_service()
+    _operator_call(
+        lambda: service.begin_shadow(
+            tenant_id=tenant_id,
+            expected_cas_version=expected_cas_version,
+            started_at=_parse_timestamp(started_at) if started_at is not None else None,
+        )
+    )
+    _echo_json(service.status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("shadow-report")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist observations; omitted means dry-run.")
+def shadow_report(input_path: Path, apply: bool) -> None:
+    observations = _read_jsonl(input_path, ShadowAuthorizationObservationInput)
+    report = _operator_call(lambda: _cutover_service().record_shadow_report(observations, apply=apply))
+    _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("shadow-complete")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist completion; omitted means dry-run.")
+def shadow_complete(input_path: Path, apply: bool) -> None:
+    payload = _read_one_jsonl(input_path, ShadowCompletionInput)
+    report = _operator_call(lambda: _cutover_service().complete_shadow(payload, apply=apply))
+    _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("issue-approve")
+@click.option("--tenant-id", required=True)
+@click.option("--issue-key", required=True)
+@click.option("--account-id", required=True)
+@click.option("--at", "approved_at", required=True)
+def issue_approve(tenant_id: str, issue_key: str, account_id: str, approved_at: str) -> None:
+    _operator_call(
+        lambda: _cutover_service().approve_issue_fail_closed(
+            tenant_id=tenant_id,
+            issue_key=issue_key,
+            account_id=account_id,
+            approved_at=_parse_timestamp(approved_at),
+        )
+    )
+    _echo_json(_cutover_service().status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("issue-resolve")
+@click.option("--tenant-id", required=True)
+@click.option("--issue-key", required=True)
+@click.option("--account-id", required=True)
+@click.option("--at", "resolved_at", required=True)
+def issue_resolve(tenant_id: str, issue_key: str, account_id: str, resolved_at: str) -> None:
+    _operator_call(
+        lambda: _cutover_service().resolve_issue(
+            tenant_id=tenant_id,
+            issue_key=issue_key,
+            account_id=account_id,
+            resolved_at=_parse_timestamp(resolved_at),
+        )
+    )
+    _echo_json(_cutover_service().status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("shadow-approve")
+@click.option("--tenant-id", required=True)
+@click.option("--diff-key", required=True)
+@click.option("--account-id", required=True)
+@click.option("--at", "approved_at", required=True)
+def shadow_approve(tenant_id: str, diff_key: str, account_id: str, approved_at: str) -> None:
+    _operator_call(
+        lambda: _cutover_service().approve_shadow_diff(
+            tenant_id=tenant_id,
+            diff_key=diff_key,
+            account_id=account_id,
+            approved_at=_parse_timestamp(approved_at),
+        )
+    )
+    _echo_json(_cutover_service().status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("shadow-resolve")
+@click.option("--tenant-id", required=True)
+@click.option("--diff-key", required=True)
+@click.option("--account-id", required=True)
+@click.option("--at", "resolved_at", required=True)
+def shadow_resolve(tenant_id: str, diff_key: str, account_id: str, resolved_at: str) -> None:
+    _operator_call(
+        lambda: _cutover_service().resolve_shadow_diff(
+            tenant_id=tenant_id,
+            diff_key=diff_key,
+            account_id=account_id,
+            resolved_at=_parse_timestamp(resolved_at),
+        )
+    )
+    _echo_json(_cutover_service().status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("legacy-dashboard")
+@click.option("--tenant-id", required=True)
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--checked-at", required=True)
+@click.option("--expected-cas-version", type=click.IntRange(min=0), default=None)
+@click.option("--apply", is_flag=True, default=False, help="Persist gate evidence; omitted means read-only report.")
+def legacy_dashboard(
+    tenant_id: str,
+    input_path: Path,
+    checked_at: str,
+    expected_cas_version: int | None,
+    apply: bool,
+) -> None:
+    dependencies = _read_jsonl(input_path, LegacyDependencyInput, allow_empty=True)
+    report = _operator_call(
+        lambda: _cutover_service().legacy_dependency_dashboard(
+            tenant_id=tenant_id,
+            dependencies=dependencies,
+            expected_cas_version=expected_cas_version,
+            checked_at=_parse_timestamp(checked_at),
+            apply=apply,
+        )
+    )
+    _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("legacy-check")
+@click.option("--tenant-id", required=True)
+def legacy_check(tenant_id: str) -> None:
+    status_report = _operator_call(lambda: _cutover_service().status(tenant_id=tenant_id))
+    passed = (
+        bool(status_report["legacy_dependency_ready"])
+        and status_report["open_issues"] == 0
+        and status_report["unresolved_cutover_quarantine"] == 0
+    )
+    _echo_json({"tenant_id": tenant_id, "passed": passed, "status": status_report})
+    if not passed:
+        raise click.exceptions.Exit(1)
+
+
+@knowledge_fs_control_space.command("freeze")
+@click.option("--tenant-id", required=True)
+@click.option("--expected-cas-version", type=click.IntRange(min=0), required=True)
+@click.option("--at", "freeze_at", required=True)
+def freeze(tenant_id: str, expected_cas_version: int, freeze_at: str) -> None:
+    service = _cutover_service()
+    _operator_call(
+        lambda: service.freeze(
+            tenant_id=tenant_id,
+            expected_cas_version=expected_cas_version,
+            freeze_at=_parse_timestamp(freeze_at),
+        )
+    )
+    _echo_json(service.status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("final-delta")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+def final_delta(input_path: Path) -> None:
+    payload = _read_one_jsonl(input_path, FinalDeltaInput)
+    service = _cutover_service()
+    _operator_call(lambda: service.apply_final_delta(payload))
+    _echo_json(service.status(tenant_id=str(payload.tenant_id)))
+
+
+@knowledge_fs_control_space.command("cutover")
+@click.option("--tenant-id", required=True)
+@click.option("--expected-cas-version", type=click.IntRange(min=0), required=True)
+@click.option("--at", "cutover_at", required=True)
+@click.option("--rollback-cutoff-at", required=True)
+def cutover(tenant_id: str, expected_cas_version: int, cutover_at: str, rollback_cutoff_at: str) -> None:
+    service = _cutover_service()
+    _operator_call(
+        lambda: service.cutover(
+            tenant_id=tenant_id,
+            expected_cas_version=expected_cas_version,
+            cutover_at=_parse_timestamp(cutover_at),
+            rollback_cutoff_at=_parse_timestamp(rollback_cutoff_at),
+        )
+    )
+    _echo_json(service.status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("smoke")
+@click.option("--tenant-id", required=True)
+@click.option("--expected-cas-version", type=click.IntRange(min=0), required=True)
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+def smoke(tenant_id: str, expected_cas_version: int, input_path: Path) -> None:
+    results = _read_one_jsonl(input_path, CutoverSmokeResultsInput)
+    service = _cutover_service()
+    _operator_call(
+        lambda: service.record_smoke_results(
+            tenant_id=tenant_id,
+            expected_cas_version=expected_cas_version,
+            results=results,
+        )
+    )
+    _echo_json(service.status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("observe")
+@click.option("--tenant-id", required=True)
+@click.option("--expected-cas-version", type=click.IntRange(min=0), required=True)
+@click.option("--started-at", default=None)
+@click.option("--window-ends-at", default=None)
+@click.option("--maximum-task-expires-at", default=None)
+@click.option("--observed-at", default=None)
+def observe(
+    tenant_id: str,
+    expected_cas_version: int,
+    started_at: str | None,
+    window_ends_at: str | None,
+    maximum_task_expires_at: str | None,
+    observed_at: str | None,
+) -> None:
+    service = _cutover_service()
+    if observed_at is not None:
+        if any(value is not None for value in (started_at, window_ends_at, maximum_task_expires_at)):
+            raise click.UsageError("--observed-at cannot be combined with observation start options")
+        _operator_call(
+            lambda: service.complete_observation(
+                tenant_id=tenant_id,
+                expected_cas_version=expected_cas_version,
+                observed_at=_parse_timestamp(observed_at),
+            )
+        )
+    else:
+        if started_at is None or window_ends_at is None or maximum_task_expires_at is None:
+            raise click.UsageError(
+                "observation start requires --started-at, --window-ends-at, and --maximum-task-expires-at"
+            )
+        _operator_call(
+            lambda: service.begin_observation(
+                tenant_id=tenant_id,
+                expected_cas_version=expected_cas_version,
+                started_at=_parse_timestamp(started_at),
+                window_ends_at=_parse_timestamp(window_ends_at),
+                maximum_task_expires_at=_parse_timestamp(maximum_task_expires_at),
+            )
+        )
+    _echo_json(service.status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("rollback")
+@click.option("--tenant-id", required=True)
+@click.option("--expected-cas-version", type=click.IntRange(min=0), required=True)
+@click.option("--at", "rolled_back_at", required=True)
+def rollback(tenant_id: str, expected_cas_version: int, rolled_back_at: str) -> None:
+    service = _cutover_service()
+    _operator_call(
+        lambda: service.rollback(
+            tenant_id=tenant_id,
+            expected_cas_version=expected_cas_version,
+            rolled_back_at=_parse_timestamp(rolled_back_at),
+        )
+    )
+    _echo_json(service.status(tenant_id=tenant_id))
+
+
+@knowledge_fs_control_space.command("status")
+@click.option("--tenant-id", required=True)
+def status(tenant_id: str) -> None:
+    _echo_json(_operator_call(lambda: _cutover_service().status(tenant_id=tenant_id)))
+
+
+@knowledge_fs_control_space.command("cleanup-request")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist readiness evidence; omitted means dry-run.")
+def cleanup_request(input_path: Path, apply: bool) -> None:
+    payload = _read_one_jsonl(input_path, CleanupReadinessEvidenceInput)
+    report = _cleanup_call(lambda: _cleanup_service().request_cleanup(payload, apply=apply))
+    _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("cleanup-approve")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist four-eyes approval; omitted means dry-run.")
+def cleanup_approve(input_path: Path, apply: bool) -> None:
+    payload = _read_one_jsonl(input_path, CleanupApprovalInput)
+    report = _cleanup_call(lambda: _cleanup_service().approve_cleanup(payload, apply=apply))
+    _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("cleanup-start")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist the irreversible fence; never runs deletion.")
+@click.option(
+    "--acknowledge-irreversible",
+    is_flag=True,
+    default=False,
+    help="Required with --apply; confirms rollback will be permanently closed.",
+)
+def cleanup_start(input_path: Path, apply: bool, acknowledge_irreversible: bool) -> None:
+    payload = _read_one_jsonl(input_path, CleanupStartInput)
+    if apply and not acknowledge_irreversible:
+        raise click.UsageError("--apply requires --acknowledge-irreversible")
+    report = _cleanup_call(lambda: _cleanup_service().start_cleanup(payload, apply=apply))
+    _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("cleanup-complete")
+@click.option("--input", "input_path", type=click.Path(path_type=Path, exists=True, dir_okay=False), required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist externally verified cleanup completion.")
+@click.option(
+    "--acknowledge-executed",
+    is_flag=True,
+    default=False,
+    help="Required with --apply; confirms the reviewed destructive bundle already executed.",
+)
+def cleanup_complete(input_path: Path, apply: bool, acknowledge_executed: bool) -> None:
+    payload = _read_one_jsonl(input_path, CleanupCompletionEvidenceInput)
+    if apply and not acknowledge_executed:
+        raise click.UsageError("--apply requires --acknowledge-executed")
+    report = _cleanup_call(lambda: _cleanup_service().complete_cleanup(payload, apply=apply))
+    _echo_json(report._asdict())
+
+
+@knowledge_fs_control_space.command("cleanup-status")
+@click.option("--tenant-id", required=True)
+@click.option("--request-id", required=True)
+def cleanup_status(tenant_id: str, request_id: str) -> None:
+    _echo_json(_cleanup_call(lambda: _cleanup_service().status(tenant_id=tenant_id, request_id=request_id)))
+
+
+@knowledge_fs_control_space.command("repair")
+@click.option("--tenant-id", required=True)
+@click.option("--control-space-id", required=True)
+@click.option("--expected-resource-version", type=click.IntRange(min=0), required=True)
+@click.option("--knowledge-space-id", required=True)
+@click.option("--knowledge-space-revision", type=click.IntRange(min=0), required=True)
+def repair(
+    tenant_id: str,
+    control_space_id: str,
+    expected_resource_version: int,
+    knowledge_space_id: str,
+    knowledge_space_revision: int,
+) -> None:
+    control_space = _management_service().repair_registration(
+        tenant_id=tenant_id,
+        control_space_id=control_space_id,
+        expected_resource_version=expected_resource_version,
+        knowledge_space_id=knowledge_space_id,
+        knowledge_space_revision=knowledge_space_revision,
+    )
+    click.echo(json.dumps({"control_space_id": control_space.id, "state": control_space.state.value}, sort_keys=True))
+
+
+@knowledge_fs_control_space.command("orphan-report")
+@click.option("--limit", type=click.IntRange(min=1, max=10_000), default=500, show_default=True)
+def orphan_report(limit: int) -> None:
+    report = KnowledgeFSOrphanReconciler(
+        session_factory.get_session_maker(),
+        get_knowledge_fs_lifecycle_remote(),
+    ).reconcile(limit=limit, apply_repairs=False)
+    click.echo(json.dumps(report._asdict(), sort_keys=True))
+
+
+@knowledge_fs_control_space.command("workspace-delete-request")
+@click.option("--tenant-id", required=True)
+@click.option("--apply", is_flag=True, default=False, help="Persist durable deletion intents; omitted means dry-run.")
+def workspace_delete_request(tenant_id: str, apply: bool) -> None:
+    """Route every KnowledgeFS Space through the canonical lifecycle deletion path."""
+
+    if not apply:
+        report = _management_service().dry_run(tenant_id=tenant_id)
+        _echo_json(
+            {
+                "apply": False,
+                "by_state": report.by_state,
+                "tenant_id": tenant_id,
+                "total": report.total,
+            }
+        )
+        return
+    results = _lifecycle_call(lambda: _lifecycle_service().request_workspace_cleanup(tenant_id=tenant_id))
+    _echo_json(
+        {
+            "apply": True,
+            "control_space_ids": [result.control_space.id for result in results],
+            "operation_ids": [result.outbox.operation_id for result in results if result.outbox is not None],
+            "tenant_id": tenant_id,
+        }
+    )
+
+
+@knowledge_fs_control_space.command("workspace-delete-finalize")
+@click.option("--tenant-id", required=True)
+@click.option("--apply", is_flag=True, default=False, help="Purge terminal local control-plane rows.")
+@click.option(
+    "--acknowledge-control-plane-purge",
+    is_flag=True,
+    default=False,
+    help="Required with --apply after every remote Space has reached deleted.",
+)
+def workspace_delete_finalize(tenant_id: str, apply: bool, acknowledge_control_plane_purge: bool) -> None:
+    """Release the Workspace FK only after all remote deletions are terminal."""
+
+    service = _lifecycle_service()
+    if not apply:
+        _lifecycle_call(lambda: service.assert_workspace_deletion_allowed(tenant_id=tenant_id))
+        _echo_json({"apply": False, "ready": True, "tenant_id": tenant_id})
+        return
+    if not acknowledge_control_plane_purge:
+        raise click.UsageError("--apply requires --acknowledge-control-plane-purge")
+    deleted = _lifecycle_call(lambda: service.finalize_workspace_deletion(tenant_id=tenant_id))
+    _echo_json({"apply": True, "purged_control_spaces": deleted, "tenant_id": tenant_id})
+
+
+def _read_jsonl[InputT: BaseModel](
+    input_path: Path, input_type: type[InputT], *, allow_empty: bool = False
+) -> tuple[InputT, ...]:
+    records: list[InputT] = []
+    for line_number, line in enumerate(input_path.read_text(encoding="utf-8").splitlines(), start=1):
+        if not line.strip():
+            continue
+        try:
+            records.append(input_type.model_validate_json(line))
+        except ValidationError as exc:
+            raise click.ClickException(f"invalid strict JSONL at line {line_number}: {exc}") from exc
+    if not records and not allow_empty:
+        raise click.ClickException("strict JSONL input must contain at least one record")
+    return tuple(records)
+
+
+def _read_one_jsonl[InputT: BaseModel](input_path: Path, input_type: type[InputT]) -> InputT:
+    records = _read_jsonl(input_path, input_type)
+    if len(records) != 1:
+        raise click.ClickException("this command requires exactly one JSONL record")
+    return records[0]
+
+
+def _parse_timestamp(value: str) -> datetime:
+    try:
+        parsed = datetime.fromisoformat(value)
+    except ValueError as exc:
+        raise click.ClickException(f"invalid ISO-8601 timestamp: {value}") from exc
+    if parsed.tzinfo is None:
+        raise click.ClickException("operator timestamps must include an explicit timezone")
+    return parsed
+
+
+def _operator_call[ResultT](operation: Callable[[], ResultT]) -> ResultT:
+    try:
+        return operation()
+    except KnowledgeFSCutoverError as exc:
+        raise click.ClickException(str(exc)) from exc
+
+
+def _cleanup_call[ResultT](operation: Callable[[], ResultT]) -> ResultT:
+    try:
+        return operation()
+    except KnowledgeFSCleanupError as exc:
+        raise click.ClickException(str(exc)) from exc
+
+
+def _lifecycle_call[ResultT](operation: Callable[[], ResultT]) -> ResultT:
+    try:
+        return operation()
+    except KnowledgeFSControlSpaceLifecycleError as exc:
+        raise click.ClickException(str(exc)) from exc
+
+
+def _echo_json(payload: object) -> None:
+    click.echo(json.dumps(payload, default=str, sort_keys=True))
+
+
+def _management_service() -> KnowledgeFSControlSpaceManagementService:
+    return KnowledgeFSControlSpaceManagementService(session_factory.get_session_maker())
+
+
+def _cutover_service() -> KnowledgeFSWorkspaceCutoverService:
+    return KnowledgeFSWorkspaceCutoverService(
+        session_factory.get_session_maker(),
+        remote_factory=get_knowledge_fs_lifecycle_remote,
+    )
+
+
+def _cleanup_service() -> KnowledgeFSCleanupService:
+    return KnowledgeFSCleanupService(session_factory.get_session_maker())
+
+
+def _lifecycle_service() -> KnowledgeFSControlSpaceCommandService:
+    return KnowledgeFSControlSpaceCommandService(session_factory.get_session_maker())
+
+
+__all__ = ["knowledge_fs_control_space"]

api/commands/retention.py

@@ -1,6 +1,8 @@
 import datetime
 import logging
+import re
 import time
+import uuid
 from collections.abc import Callable
 from typing import TypedDict
 
@@ -21,6 +23,7 @@ from tasks.remove_app_and_related_data_task import delete_draft_variables_batch
 logger = logging.getLogger(__name__)
 
 _HEX_PREFIXES = tuple("0123456789abcdef")
+_TARGET_MONTH_PATTERN = re.compile(r"^\d{4}-(0[1-9]|1[0-2])$")
 
 
 class WorkflowRunArchivePlanRow(TypedDict):
@@ -66,6 +69,7 @@ def _parse_tenant_prefixes(prefixes: str | None) -> list[str]:
 
 
 def _parse_comma_separated_ids(raw_ids: str | None, *, param_name: str) -> list[str] | None:
+    """Keep an omitted scope unset while rejecting an explicitly empty scope."""
     if raw_ids is None:
         return None
     parsed = sorted({raw_id.strip() for raw_id in raw_ids.split(",") if raw_id.strip()})
@@ -74,6 +78,27 @@ def _parse_comma_separated_ids(raw_ids: str | None, *, param_name: str) -> list[
     return parsed
 
 
+def _parse_archive_target_month(target_month: str) -> tuple[int, int]:
+    """Validate the V2 catalog month selector and return its numeric components."""
+    if not _TARGET_MONTH_PATTERN.fullmatch(target_month):
+        raise click.BadParameter("target-month must use YYYY-MM format", param_hint="--target-month")
+    year_text, month_text = target_month.split("-", maxsplit=1)
+    return int(year_text), int(month_text)
+
+
+def _parse_archive_catalog_cursor(after_catalog_id: str | None) -> str | None:
+    """Normalize the exclusive V2 catalog keyset cursor when one is provided."""
+    if after_catalog_id is None:
+        return None
+    try:
+        return str(uuid.UUID(after_catalog_id))
+    except ValueError as exc:
+        raise click.BadParameter(
+            "after-catalog-id must be a UUID returned by the same V2 operation and scope",
+            param_hint="--after-catalog-id",
+        ) from exc
+
+
 def _get_archive_candidate_tenant_ids_by_prefix(
     session: Session,
     prefix: str,
@@ -810,9 +835,11 @@ def backfill_workflow_run_archive_bundles(
         click.echo(click.style(f"  ... and {len(summary.errors) - 10} more failures", fg="red"))
 
 
-def _echo_bundle_archive_operation_summary(summary) -> None:
+def _echo_bundle_archive_operation_summary(summary, *, dry_run: bool) -> None:
     status = "completed successfully" if summary.bundles_failed == 0 else "completed with failures"
     fg = "green" if summary.bundles_failed == 0 else "red"
+    cursor_label = "preview_next_catalog_id" if dry_run else "next_catalog_id"
+    cursor_value = summary.preview_next_catalog_id if dry_run else summary.next_catalog_id
     click.echo(
         click.style(
             f"{summary.operation} {status}. "
@@ -821,10 +848,12 @@ def _echo_bundle_archive_operation_summary(summary) -> None:
             f"archive_bytes={summary.archive_bytes} duration={summary.elapsed_time:.2f}s "
             f"validation_time={summary.validation_time:.2f}s "
             f"runs_per_second={summary.runs_per_second:.2f} rows_per_second={summary.rows_per_second:.2f} "
-            f"bytes_per_second={summary.bytes_per_second:.2f}",
+            f"bytes_per_second={summary.bytes_per_second:.2f} {cursor_label}={cursor_value or 'none'}",
             fg=fg,
         )
     )
+    if dry_run:
+        click.echo(click.style("Dry-run cursor is preview-only; do not persist it for a destructive run.", fg="yellow"))
     click.echo(click.style("table,row_count", fg="white"))
     for table_name in [
         "workflow_runs",
@@ -842,7 +871,8 @@ def _echo_bundle_archive_operation_summary(summary) -> None:
                 click.style(
                     f"  bundle={result.bundle_id} tenant={result.tenant_id} runs={result.run_count} "
                     f"rows={result.row_count} archive_bytes={result.archive_bytes} "
-                    f"time={result.elapsed_time:.2f}s validation={result.validation_time:.2f}s",
+                    f"catalog_id={result.catalog_id} time={result.elapsed_time:.2f}s "
+                    f"validation={result.validation_time:.2f}s",
                     fg="white",
                 )
             )
@@ -850,7 +880,7 @@ def _echo_bundle_archive_operation_summary(summary) -> None:
             click.echo(
                 click.style(
                     f"  failed bundle={result.bundle_id} tenant={result.tenant_id} "
-                    f"object_prefix={result.object_prefix} error={result.error}",
+                    f"catalog_id={result.catalog_id} object_prefix={result.object_prefix} error={result.error}",
                     fg="red",
                 )
             )
@@ -867,25 +897,24 @@ def _echo_bundle_archive_operation_summary(summary) -> None:
 )
 @click.option("--run-id", required=False, help="Workflow run ID to restore.")
 @click.option(
-    "--start-from",
-    type=click.DateTime(formats=["%Y-%m-%d", "%Y-%m-%dT%H:%M:%S"]),
+    "--target-month",
+    metavar="YYYY-MM",
     default=None,
-    help="Optional lower bound (inclusive) for created_at; must be paired with --end-before.",
+    help="V2 catalog month to restore; required unless --run-id is used.",
 )
 @click.option(
-    "--end-before",
-    type=click.DateTime(formats=["%Y-%m-%d", "%Y-%m-%dT%H:%M:%S"]),
+    "--after-catalog-id",
     default=None,
-    help="Optional upper bound (exclusive) for created_at; must be paired with --start-from.",
+    help="Exclusive V2 cursor from the same restore month and tenant scope.",
 )
 @click.option("--workers", default=1, show_default=True, type=int, help="V1 --run-id compatibility only.")
-@click.option("--limit", type=int, default=100, show_default=True, help="Maximum number of V2 bundles to restore.")
+@click.option("--limit", type=click.IntRange(min=1), default=100, show_default=True, help="Maximum V2 catalog rows.")
 @click.option("--dry-run", is_flag=True, help="Preview without restoring.")
 def restore_workflow_runs(
     tenant_ids: str | None,
     run_id: str | None,
-    start_from: datetime.datetime | None,
-    end_before: datetime.datetime | None,
+    target_month: str | None,
+    after_catalog_id: str | None,
     workers: int,
     limit: int,
     dry_run: bool,
@@ -905,23 +934,20 @@ def restore_workflow_runs(
     from services.retention.workflow_run.bundle_archive_maintenance import WorkflowRunBundleArchiveMaintenance
     from services.retention.workflow_run.restore_archived_workflow_run import WorkflowRunRestore
 
-    parsed_tenant_ids = None
-    if tenant_ids:
-        parsed_tenant_ids = [tid.strip() for tid in tenant_ids.split(",") if tid.strip()]
-        if not parsed_tenant_ids:
-            raise click.BadParameter("tenant-ids must not be empty")
+    parsed_tenant_ids = _parse_comma_separated_ids(tenant_ids, param_name="tenant-ids")
 
-    if (start_from is None) ^ (end_before is None):
-        raise click.UsageError("--start-from and --end-before must be provided together.")
-    if run_id is None and (start_from is None or end_before is None):
-        raise click.UsageError("--start-from and --end-before are required for batch restore.")
     if workers < 1:
         raise click.BadParameter("workers must be at least 1")
+    if run_id is not None and (target_month is not None or after_catalog_id is not None):
+        raise click.UsageError("--target-month and --after-catalog-id are only valid for V2 batch restore.")
+    if run_id is None and target_month is None:
+        raise click.UsageError("--target-month is required for V2 batch restore.")
 
     start_time = datetime.datetime.now(datetime.UTC)
+    target_desc = f"workflow run {run_id}" if run_id else f"workflow archive catalog month {target_month}"
     click.echo(
         click.style(
-            f"Starting restore of workflow run {run_id} at {start_time.isoformat()}.",
+            f"Starting restore of {target_desc} at {start_time.isoformat()}.",
             fg="white",
         )
     )
@@ -955,17 +981,20 @@ def restore_workflow_runs(
         click.echo(
             click.style("--workers is ignored for V2 bundle restore; bundles are processed serially.", fg="yellow")
         )
-    assert start_from is not None
-    assert end_before is not None
+    assert target_month is not None
+    target_year, target_month_number = _parse_archive_target_month(target_month)
+    catalog_cursor = _parse_archive_catalog_cursor(after_catalog_id)
     bundle_restorer = WorkflowRunBundleArchiveMaintenance(dry_run=dry_run, strict_content_validation=True)
     summary = bundle_restorer.restore_batch(
         tenant_ids=parsed_tenant_ids,
-        start_date=start_from,
-        end_date=end_before,
+        target_year=target_year,
+        target_month=target_month_number,
+        after_catalog_id=catalog_cursor,
         limit=limit,
     )
-    _echo_bundle_archive_operation_summary(summary)
-    return
+    _echo_bundle_archive_operation_summary(summary, dry_run=dry_run)
+    if summary.bundles_failed:
+        raise click.exceptions.Exit(1)
 
 
 @click.command(
@@ -979,23 +1008,41 @@ def restore_workflow_runs(
 )
 @click.option("--run-id", required=False, help="Workflow run ID to delete.")
 @click.option(
-    "--start-from",
-    type=click.DateTime(formats=["%Y-%m-%d", "%Y-%m-%dT%H:%M:%S"]),
+    "--target-month",
+    metavar="YYYY-MM",
     default=None,
-    help="Optional lower bound (inclusive) for created_at; must be paired with --end-before.",
+    help="V2 catalog month to delete; required unless --run-id is used.",
 )
 @click.option(
-    "--end-before",
-    type=click.DateTime(formats=["%Y-%m-%d", "%Y-%m-%dT%H:%M:%S"]),
+    "--after-catalog-id",
     default=None,
-    help="Optional upper bound (exclusive) for created_at; must be paired with --start-from.",
+    help="Exclusive V2 cursor from the same delete month and tenant scope.",
+)
+@click.option(
+    "--run-shard-index",
+    default=None,
+    type=click.IntRange(min=0),
+    help="Zero-based archive shard index. Must be paired with --run-shard-total.",
+)
+@click.option(
+    "--run-shard-total",
+    default=None,
+    type=click.IntRange(min=1, max=16),
+    help="Total archive shard count. Must be paired with --run-shard-index.",
+)
+@click.option("--all-pages", is_flag=True, help="Process catalog pages until an empty page is reached.")
+@click.option(
+    "--limit",
+    type=click.IntRange(min=1),
+    default=100,
+    show_default=True,
+    help="Maximum V2 catalog rows per page.",
 )
-@click.option("--limit", type=int, default=100, show_default=True, help="Maximum number of V2 bundles to delete.")
 @click.option("--dry-run", is_flag=True, help="Preview without deleting.")
 @click.option(
     "--skip-bad-archives",
     is_flag=True,
-    help="Continue batch deletion when one archive object fails validation.",
+    help="V1 --run-id only: continue when one archive object fails validation.",
 )
 @click.option(
     "--restore-sample-interval",
@@ -1007,8 +1054,11 @@ def restore_workflow_runs(
 def delete_archived_workflow_runs(
     tenant_ids: str | None,
     run_id: str | None,
-    start_from: datetime.datetime | None,
-    end_before: datetime.datetime | None,
+    target_month: str | None,
+    after_catalog_id: str | None,
+    run_shard_index: int | None,
+    run_shard_total: int | None,
+    all_pages: bool,
     limit: int,
     dry_run: bool,
     skip_bad_archives: bool,
@@ -1018,26 +1068,38 @@ def delete_archived_workflow_runs(
     Delete archived workflow runs from the database.
 
     Batch delete uses V2 bundle metadata and validates object existence, manifest schema, object size, checksum, row
-    counts, and source/archive content checksums before deleting source rows. `--run-id` keeps the V1 per-run path.
+    counts, and source/archive content checksums before deleting source rows. Parallel workers may select one exact
+    archive shard; all-pages mode keeps only the current bounded page in memory. `--run-id` keeps the V1 per-run path.
     """
     from services.retention.workflow_run.bundle_archive_maintenance import WorkflowRunBundleArchiveMaintenance
     from services.retention.workflow_run.delete_archived_workflow_run import ArchivedWorkflowRunDeletion
 
-    parsed_tenant_ids = None
-    if tenant_ids:
-        parsed_tenant_ids = [tid.strip() for tid in tenant_ids.split(",") if tid.strip()]
-        if not parsed_tenant_ids:
-            raise click.BadParameter("tenant-ids must not be empty")
+    parsed_tenant_ids = _parse_comma_separated_ids(tenant_ids, param_name="tenant-ids")
 
-    if (start_from is None) ^ (end_before is None):
-        raise click.UsageError("--start-from and --end-before must be provided together.")
-    if run_id is None and (start_from is None or end_before is None):
-        raise click.UsageError("--start-from and --end-before are required for batch delete.")
     if restore_sample_interval < 0:
         raise click.BadParameter("restore-sample-interval must be >= 0")
+    if run_id is not None and (
+        target_month is not None
+        or after_catalog_id is not None
+        or run_shard_index is not None
+        or run_shard_total is not None
+        or all_pages
+    ):
+        raise click.UsageError(
+            "--target-month, --after-catalog-id, --run-shard-index, --run-shard-total, and --all-pages "
+            "are only valid for V2 batch delete."
+        )
+    if run_id is None and target_month is None:
+        raise click.UsageError("--target-month is required for V2 batch delete.")
+    if run_id is None and skip_bad_archives:
+        raise click.UsageError("--skip-bad-archives is not supported for V2 catalog batches; they fail fast.")
+    if (run_shard_index is None) ^ (run_shard_total is None):
+        raise click.UsageError("--run-shard-index and --run-shard-total must be provided together.")
+    if run_shard_index is not None and run_shard_total is not None and run_shard_index >= run_shard_total:
+        raise click.UsageError("--run-shard-index must be less than --run-shard-total.")
 
     start_time = datetime.datetime.now(datetime.UTC)
-    target_desc = f"workflow run {run_id}" if run_id else "workflow runs"
+    target_desc = f"workflow run {run_id}" if run_id else f"workflow archive catalog month {target_month}"
     click.echo(
         click.style(
             f"Starting delete of {target_desc} at {start_time.isoformat()}.",
@@ -1110,20 +1172,104 @@ def delete_archived_workflow_runs(
 
     if restore_sample_interval:
         click.echo(click.style("--restore-sample-interval is ignored for V2 bundle delete.", fg="yellow"))
-    assert start_from is not None
-    assert end_before is not None
-    bundle_deleter = WorkflowRunBundleArchiveMaintenance(
-        dry_run=dry_run,
-        strict_content_validation=True,
-        stop_on_error=not skip_bad_archives,
+    assert target_month is not None
+    target_year, target_month_number = _parse_archive_target_month(target_month)
+    catalog_cursor = _parse_archive_catalog_cursor(after_catalog_id)
+    shard = (
+        f"{run_shard_index:02d}-of-{run_shard_total:02d}"
+        if run_shard_index is not None and run_shard_total is not None
+        else None
     )
-    summary = bundle_deleter.delete_batch(
-        tenant_ids=parsed_tenant_ids,
-        start_date=start_from,
-        end_date=end_before,
-        limit=limit,
-    )
-    _echo_bundle_archive_operation_summary(summary)
+    bundle_deleter = WorkflowRunBundleArchiveMaintenance(dry_run=dry_run, strict_content_validation=True)
+    if run_shard_total is not None:
+        try:
+            bundle_deleter.validate_catalog_shards(
+                target_year=target_year,
+                target_month=target_month_number,
+                shard_total=run_shard_total,
+                tenant_ids=parsed_tenant_ids,
+            )
+        except ValueError as exc:
+            logger.exception(
+                "Archive catalog shard preflight failed: target_month=%s shard=%s",
+                target_month,
+                shard,
+            )
+            raise click.ClickException(
+                f"Archive catalog shard preflight failed for target_month={target_month} shard={shard}: {exc}"
+            ) from exc
+
+    initial_catalog_cursor = catalog_cursor
+    pages_processed = 0
+    bundles_succeeded = 0
+    runs_processed = 0
+    rows_processed = 0
+    archive_bytes = 0
+    while True:
+        summary = bundle_deleter.delete_batch(
+            tenant_ids=parsed_tenant_ids,
+            target_year=target_year,
+            target_month=target_month_number,
+            after_catalog_id=catalog_cursor,
+            limit=limit,
+            shard=shard,
+        )
+        _echo_bundle_archive_operation_summary(summary, dry_run=dry_run)
+        if summary.bundles_failed:
+            failed_result = next((result for result in summary.results if not result.success), None)
+            failed_catalog_id = failed_result.catalog_id if failed_result is not None else "unknown"
+            page_resume_cursor = summary.preview_next_catalog_id if dry_run else summary.next_catalog_id
+            resume_cursor = page_resume_cursor or catalog_cursor
+            if dry_run:
+                cursor_details = (
+                    f"preview_after_catalog_id={resume_cursor or 'none'} "
+                    f"destructive_retry_after_catalog_id={initial_catalog_cursor or 'none'}"
+                )
+            else:
+                cursor_details = f"resume_after_catalog_id={resume_cursor or 'none'}"
+            click.echo(
+                click.style(
+                    f"Delete stopped: target_month={target_month} shard={shard or 'all'} "
+                    f"failed_catalog_id={failed_catalog_id} "
+                    f"{cursor_details}",
+                    fg="red",
+                )
+            )
+            raise click.exceptions.Exit(1)
+
+        if not all_pages:
+            break
+        if summary.bundles_processed == 0:
+            break
+
+        pages_processed += 1
+        bundles_succeeded += summary.bundles_succeeded
+        runs_processed += summary.runs_processed
+        rows_processed += summary.rows_processed
+        archive_bytes += summary.archive_bytes
+        next_catalog_id = summary.preview_next_catalog_id if dry_run else summary.next_catalog_id
+        if next_catalog_id is None or (catalog_cursor is not None and next_catalog_id <= catalog_cursor):
+            click.echo(
+                click.style(
+                    f"Delete cursor did not advance: target_month={target_month} shard={shard or 'all'} "
+                    f"after_catalog_id={catalog_cursor or 'none'} next_catalog_id={next_catalog_id or 'none'}",
+                    fg="red",
+                )
+            )
+            raise click.exceptions.Exit(1)
+        catalog_cursor = next_catalog_id
+
+    if all_pages:
+        final_cursor_label = "preview_final_catalog_id" if dry_run else "final_catalog_id"
+        click.echo(
+            click.style(
+                f"Delete all-pages completed successfully. target_month={target_month} shard={shard or 'all'} "
+                f"pages={pages_processed} bundles_success={bundles_succeeded} runs={runs_processed} "
+                f"rows={rows_processed} archive_bytes={archive_bytes} "
+                f"{final_cursor_label}={catalog_cursor or 'none'}",
+                fg="green",
+            )
+        )
 
 
 def _find_orphaned_draft_variables(batch_size: int = 1000) -> list[str]:

api/configs/extra/__init__.py

@@ -1,5 +1,6 @@
 from configs.extra.agent_backend_config import AgentBackendConfig
 from configs.extra.archive_config import ArchiveStorageConfig
+from configs.extra.knowledge_fs_config import KnowledgeFSConfig
 from configs.extra.notion_config import NotionConfig
 from configs.extra.sentry_config import SentryConfig
 
@@ -8,6 +9,7 @@ class ExtraServiceConfig(
     # place the configs in alphabet order
     AgentBackendConfig,
     ArchiveStorageConfig,
+    KnowledgeFSConfig,
     NotionConfig,
     SentryConfig,
 ):

api/configs/extra/knowledge_fs_config.py

@@ -0,0 +1,134 @@
+"""Configuration for the optional KnowledgeFS control-plane integration."""
+
+from ipaddress import ip_address
+from urllib.parse import urlsplit
+
+from pydantic import Field, PositiveFloat, PositiveInt, SecretStr, field_validator, model_validator
+from pydantic_settings import BaseSettings
+
+
+class KnowledgeFSConfig(BaseSettings):
+    """Server-only KnowledgeFS connection and rollout settings."""
+
+    KNOWLEDGE_FS_ENABLED: bool = Field(
+        default=False,
+        description="Enable the private KnowledgeFS Console bridge.",
+    )
+    KNOWLEDGE_FS_LIFECYCLE_WORKER_ENABLED: bool = Field(
+        default=False,
+        description="Enable delivery of durable KnowledgeFS lifecycle commands after every rollout gate is ready.",
+    )
+    KNOWLEDGE_FS_INTEGRATED_PROVISION_READY: bool = Field(
+        default=False,
+        description="Confirm that the Capability-v2 integrated provision route is deployed and verified.",
+    )
+    KNOWLEDGE_FS_LEGACY_ACL_FREEZE_READY: bool = Field(
+        default=False,
+        description="Confirm that legacy KFS ACL mutation is frozen for integrated mode.",
+    )
+    KNOWLEDGE_FS_LIFECYCLE_POLL_INTERVAL_SECONDS: PositiveInt = Field(default=15, le=300)
+    KNOWLEDGE_FS_LIFECYCLE_LEASE_SECONDS: PositiveInt = Field(default=60, le=600)
+    KNOWLEDGE_FS_LIFECYCLE_BATCH_SIZE: PositiveInt = Field(default=25, le=1_000)
+    KNOWLEDGE_FS_BASE_URL: str | None = Field(default=None, description="KnowledgeFS gateway base URL.")
+    KNOWLEDGE_FS_DIRECT_ORIGIN: str | None = Field(
+        default=None,
+        description="Public KnowledgeFS origin returned with direct upload capabilities.",
+    )
+    KNOWLEDGE_FS_CAPABILITY_V2_ENABLED: bool = Field(
+        default=False,
+        description="Prepare resource-scoped Capability v2 issuance; disabled until rollout approval.",
+    )
+    KNOWLEDGE_FS_CAPABILITY_V2_SIGNING_KID: str | None = Field(
+        default=None,
+        description="Identifier for the current asymmetric Capability v2 signing key.",
+    )
+    KNOWLEDGE_FS_CAPABILITY_V2_PRIVATE_KEY_PEM: SecretStr | None = Field(
+        default=None,
+        description="Server-only PEM for the current Capability v2 RSA signing key.",
+    )
+    KNOWLEDGE_FS_CAPABILITY_V2_PREVIOUS_PUBLIC_JWKS: str | None = Field(
+        default=None,
+        description="Optional public-only JWKS JSON retained during key rotation overlap.",
+    )
+    KNOWLEDGE_FS_CAPABILITY_V2_ISSUER: str = Field(default="dify-control-plane", min_length=1)
+    KNOWLEDGE_FS_CAPABILITY_V2_AUDIENCE: str = Field(default="knowledge-fs", min_length=1)
+    KNOWLEDGE_FS_CAPABILITY_V2_MAX_TTL_SECONDS: PositiveInt = Field(default=60, le=60)
+    KNOWLEDGE_FS_JWKS_CACHE_MAX_AGE_SECONDS: PositiveInt = Field(default=300, le=86_400)
+    KNOWLEDGE_FS_PRODUCT_MAX_RESPONSE_BYTES: PositiveInt = Field(default=4 * 1024 * 1024, le=16 * 1024 * 1024)
+    KNOWLEDGE_FS_TIMEOUT_SECONDS: PositiveFloat = Field(default=10.0, le=60.0, allow_inf_nan=False)
+
+    @field_validator(
+        "KNOWLEDGE_FS_BASE_URL",
+        "KNOWLEDGE_FS_DIRECT_ORIGIN",
+        "KNOWLEDGE_FS_CAPABILITY_V2_SIGNING_KID",
+        "KNOWLEDGE_FS_CAPABILITY_V2_PRIVATE_KEY_PEM",
+        "KNOWLEDGE_FS_CAPABILITY_V2_PREVIOUS_PUBLIC_JWKS",
+        mode="before",
+    )
+    @classmethod
+    def normalize_optional_string(cls, value: object) -> object:
+        if isinstance(value, SecretStr):
+            normalized = value.get_secret_value().strip()
+            return SecretStr(normalized) if normalized else None
+        if isinstance(value, str):
+            normalized = value.strip()
+            return normalized or None
+        return value
+
+    @field_validator("KNOWLEDGE_FS_BASE_URL")
+    @classmethod
+    def validate_base_url(cls, value: str | None) -> str | None:
+        return cls._validate_origin(value, name="KNOWLEDGE_FS_BASE_URL")
+
+    @field_validator("KNOWLEDGE_FS_DIRECT_ORIGIN")
+    @classmethod
+    def validate_direct_origin(cls, value: str | None) -> str | None:
+        return cls._validate_origin(value, name="KNOWLEDGE_FS_DIRECT_ORIGIN")
+
+    @classmethod
+    def _validate_origin(cls, value: str | None, *, name: str) -> str | None:
+        if value is None:
+            return None
+        parsed = urlsplit(value)
+        if parsed.scheme not in {"http", "https"} or not parsed.netloc:
+            raise ValueError(f"{name} must be an absolute HTTP(S) URL")
+        try:
+            _ = parsed.port
+        except ValueError as exc:
+            raise ValueError(f"{name} must include a valid port") from exc
+        if parsed.username or parsed.password or parsed.query or parsed.fragment or parsed.path not in {"", "/"}:
+            raise ValueError(f"{name} must be an origin without credentials, path, query, or fragment")
+        return value.rstrip("/")
+
+    @model_validator(mode="after")
+    def validate_enabled_connection(self) -> "KnowledgeFSConfig":
+        if str(getattr(self, "DEPLOY_ENV", "")).strip().upper() == "PRODUCTION":
+            for name, value in (
+                ("KNOWLEDGE_FS_BASE_URL", self.KNOWLEDGE_FS_BASE_URL),
+                ("KNOWLEDGE_FS_DIRECT_ORIGIN", self.KNOWLEDGE_FS_DIRECT_ORIGIN),
+            ):
+                if value and not self._is_secure_or_loopback_origin(value):
+                    raise ValueError(f"{name} must use HTTPS in production unless it targets loopback")
+        if self.KNOWLEDGE_FS_ENABLED:
+            if not self.KNOWLEDGE_FS_BASE_URL:
+                raise ValueError("KnowledgeFS base URL is required when the integration is enabled")
+            if not self.KNOWLEDGE_FS_CAPABILITY_V2_ENABLED:
+                raise ValueError("KnowledgeFS product routes require Capability v2 when enabled")
+        if self.KNOWLEDGE_FS_CAPABILITY_V2_ENABLED and not (
+            self.KNOWLEDGE_FS_CAPABILITY_V2_SIGNING_KID and self.KNOWLEDGE_FS_CAPABILITY_V2_PRIVATE_KEY_PEM
+        ):
+            raise ValueError("Capability v2 signing kid and private key are required when issuance is enabled")
+        return self
+
+    @staticmethod
+    def _is_secure_or_loopback_origin(value: str) -> bool:
+        parsed = urlsplit(value)
+        if parsed.scheme == "https":
+            return True
+        hostname = (parsed.hostname or "").rstrip(".").lower()
+        if hostname == "localhost":
+            return True
+        try:
+            return ip_address(hostname).is_loopback
+        except ValueError:
+            return False

api/controllers/console/__init__.py

@@ -125,6 +125,7 @@ from .explore import (
     saved_message,
     trial,
 )
+from .knowledge_fs import resources as knowledge_fs_resources
 from .snippets import snippet_workflow, snippet_workflow_draft_variable
 from .socketio import workflow as socketio_workflow
 
@@ -195,6 +196,7 @@ __all__ = [
     "human_input_form",
     "init_validate",
     "installed_app",
+    "knowledge_fs_resources",
     "load_balancing_config",
     "login",
     "mcp_server",

api/controllers/console/app/workflow.py

@@ -105,6 +105,7 @@ class SyncDraftWorkflowPayload(BaseModel):
     graph: dict[str, Any]
     features: dict[str, Any]
     hash: str | None = None
+    is_collaborative: bool = Field(default=False, alias="_is_collaborative")
     environment_variables: list[dict[str, Any]] = Field(
         default_factory=list,
     )
@@ -610,6 +611,7 @@ class DraftWorkflowApi(Resource):
                 environment_variables=environment_variables,
                 conversation_variables=conversation_variables,
                 session=db.session(),
+                graph_only=args["is_collaborative"],
             )
         except WorkflowHashNotEqualError:
             raise DraftWorkflowNotSync()

api/controllers/console/explore/trial.py

@@ -47,7 +47,9 @@ from controllers.console.explore.error import (
     NotWorkflowAppError,
 )
 from controllers.console.explore.wraps import TrialAppResource, trial_feature_enable
-from controllers.console.wraps import with_current_user
+from controllers.console.files import FILE_UPLOAD_PARAMS, upload_file_from_request
+from controllers.console.remote_files import RemoteFileUploadPayload, upload_remote_file_from_request
+from controllers.console.wraps import cloud_edition_billing_resource_check, with_current_user
 from controllers.web.error import InvokeRateLimitError as InvokeRateLimitHttpError
 from core.app.app_config.common.parameters_mapping import get_parameters_from_feature_dict
 from core.app.apps.base_app_queue_manager import AppQueueManager
@@ -61,12 +63,13 @@ from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from fields.base import ResponseModel
 from fields.conversation_variable_fields import WorkflowConversationVariableResponse
+from fields.file_fields import FileResponse, FileWithSignedUrl
 from fields.message_fields import SuggestedQuestionsResponse
 from graphon.graph_engine.manager import GraphEngineManager
 from graphon.model_runtime.errors.invoke import InvokeError
 from libs import helper
 from libs.helper import dump_response, to_timestamp, uuid_value
-from models import Account
+from models import Account, App
 from models.account import TenantStatus
 from models.model import AppMode, Site, load_annotation_reply_config
 from models.workflow import Workflow
@@ -428,6 +431,36 @@ register_response_schema_models(
 simple_account_model = console_ns.models[TrialSimpleAccount.__name__]
 
 
+class TrialAppFileUploadApi(TrialAppResource):
+    @trial_feature_enable
+    @cloud_edition_billing_resource_check("documents")
+    @console_ns.doc(consumes=["multipart/form-data"], params=FILE_UPLOAD_PARAMS)
+    @console_ns.response(201, "File uploaded successfully", console_ns.models[FileResponse.__name__])
+    @with_current_user
+    def post(self, current_user: Account, app_model: App):
+        """Upload a file into the tenant that owns the trial app."""
+        upload_file = upload_file_from_request(
+            current_user=current_user,
+            resource_tenant_id=app_model.tenant_id,
+        )
+        return dump_response(FileResponse, upload_file), 201
+
+
+class TrialAppRemoteFileUploadApi(TrialAppResource):
+    @trial_feature_enable
+    @cloud_edition_billing_resource_check("documents")
+    @console_ns.expect(console_ns.models[RemoteFileUploadPayload.__name__])
+    @console_ns.response(201, "File uploaded successfully", console_ns.models[FileWithSignedUrl.__name__])
+    @with_current_user
+    def post(self, current_user: Account, app_model: App):
+        """Upload a remote file into the tenant that owns the trial app."""
+        remote_file = upload_remote_file_from_request(
+            current_user=current_user,
+            resource_tenant_id=app_model.tenant_id,
+        )
+        return remote_file.model_dump(mode="json"), 201
+
+
 class TrialAppWorkflowRunApi(TrialAppResource):
     @trial_feature_enable
     @console_ns.expect(console_ns.models[WorkflowRunRequest.__name__])
@@ -887,6 +920,18 @@ class DatasetListApi(Resource):
 
 console_ns.add_resource(TrialChatApi, "/trial-apps/<uuid:app_id>/chat-messages", endpoint="trial_app_chat_completion")
 
+console_ns.add_resource(
+    TrialAppFileUploadApi,
+    "/trial-apps/<uuid:app_id>/files/upload",
+    endpoint="trial_app_file_upload",
+)
+
+console_ns.add_resource(
+    TrialAppRemoteFileUploadApi,
+    "/trial-apps/<uuid:app_id>/remote-files/upload",
+    endpoint="trial_app_remote_file_upload",
+)
+
 console_ns.add_resource(
     TrialMessageSuggestedQuestionApi,
     "/trial-apps/<uuid:app_id>/messages/<uuid:message_id>/suggested-questions",

api/controllers/console/files.py

@@ -29,7 +29,7 @@ from extensions.ext_database import db
 from fields.file_fields import FileResponse, UploadConfig
 from libs.helper import dump_response
 from libs.login import login_required
-from models import Account
+from models import Account, UploadFile
 from services.file_service import FileService
 
 from . import console_ns
@@ -39,7 +39,7 @@ register_response_schema_models(console_ns, AllowedExtensionsResponse, TextConte
 
 PREVIEW_WORDS_LIMIT = 3000
 
-_FILE_UPLOAD_PARAMS = {
+FILE_UPLOAD_PARAMS = {
     "file": {
         "description": "File to upload",
         "in": "formData",
@@ -56,6 +56,43 @@ _FILE_UPLOAD_PARAMS = {
 }
 
 
+def upload_file_from_request(*, current_user: Account, resource_tenant_id: str | None = None) -> UploadFile:
+    """Validate the multipart request and persist the file under the requested resource tenant."""
+    source_str = request.form.get("source")
+    source: Literal["datasets"] | None = "datasets" if source_str == "datasets" else None
+
+    if "file" not in request.files:
+        raise NoFileUploadedError()
+
+    if len(request.files) > 1:
+        raise TooManyFilesError()
+    file = request.files["file"]
+
+    if not file.filename:
+        raise FilenameNotExistsError
+    if source == "datasets" and not current_user.is_dataset_editor:
+        raise Forbidden()
+
+    if source not in ("datasets", None):
+        source = None
+
+    try:
+        return FileService(db.engine).upload_file(
+            filename=file.filename,
+            content=file.stream.read(),
+            mimetype=file.mimetype,
+            user=current_user,
+            tenant_id=resource_tenant_id,
+            source=source,
+        )
+    except services.errors.file.FileTooLargeError as file_too_large_error:
+        raise FileTooLargeError(file_too_large_error.description)
+    except services.errors.file.UnsupportedFileTypeError:
+        raise UnsupportedFileTypeError()
+    except services.errors.file.BlockedFileExtensionError as blocked_extension_error:
+        raise BlockedFileExtensionError(blocked_extension_error.description)
+
+
 @console_ns.route("/files/upload")
 class FileApi(Resource):
     @setup_required
@@ -81,42 +118,11 @@ class FileApi(Resource):
     @login_required
     @account_initialization_required
     @cloud_edition_billing_resource_check("documents")
-    @console_ns.doc(consumes=["multipart/form-data"], params=_FILE_UPLOAD_PARAMS)
+    @console_ns.doc(consumes=["multipart/form-data"], params=FILE_UPLOAD_PARAMS)
     @console_ns.response(201, "File uploaded successfully", console_ns.models[FileResponse.__name__])
     @with_current_user
     def post(self, current_user: Account):
-        source_str = request.form.get("source")
-        source: Literal["datasets"] | None = "datasets" if source_str == "datasets" else None
-
-        if "file" not in request.files:
-            raise NoFileUploadedError()
-
-        if len(request.files) > 1:
-            raise TooManyFilesError()
-        file = request.files["file"]
-
-        if not file.filename:
-            raise FilenameNotExistsError
-        if source == "datasets" and not current_user.is_dataset_editor:
-            raise Forbidden()
-
-        if source not in ("datasets", None):
-            source = None
-
-        try:
-            upload_file = FileService(db.engine).upload_file(
-                filename=file.filename,
-                content=file.stream.read(),
-                mimetype=file.mimetype,
-                user=current_user,
-                source=source,
-            )
-        except services.errors.file.FileTooLargeError as file_too_large_error:
-            raise FileTooLargeError(file_too_large_error.description)
-        except services.errors.file.UnsupportedFileTypeError:
-            raise UnsupportedFileTypeError()
-        except services.errors.file.BlockedFileExtensionError as blocked_extension_error:
-            raise BlockedFileExtensionError(blocked_extension_error.description)
+        upload_file = upload_file_from_request(current_user=current_user)
 
         return dump_response(FileResponse, upload_file), 201
 

api/controllers/console/knowledge_fs/__init__.py

@@ -0,0 +1,5 @@
+"""Typed Dify-owned KnowledgeFS Console product API."""
+
+from . import resources
+
+__all__ = ["resources"]

api/controllers/console/knowledge_fs/error.py

@@ -0,0 +1,56 @@
+"""Stable, non-enumerating Console error contract for KnowledgeFS."""
+
+from libs.exception import BaseHTTPException
+
+
+class KnowledgeFSSpaceNotFoundHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_space_not_found"
+    description = "KnowledgeFS space was not found."
+    code = 404
+
+
+class KnowledgeFSOperationUnavailableHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_operation_unavailable"
+    description = "KnowledgeFS operation is not available."
+    code = 503
+
+
+class KnowledgeFSUpstreamUnavailableHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_upstream_unavailable"
+    description = "KnowledgeFS is unavailable."
+    code = 502
+
+
+class KnowledgeFSInvalidRequestHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_invalid_request"
+    description = "KnowledgeFS request is invalid."
+    code = 400
+
+
+class KnowledgeFSAccessDeniedHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_access_denied"
+    description = "KnowledgeFS operation is not allowed."
+    code = 403
+
+
+class KnowledgeFSRateLimitHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_rate_limit_exceeded"
+    description = "KnowledgeFS operation rate limit exceeded."
+    code = 429
+
+
+class KnowledgeFSQuotaExceededHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_quota_exceeded"
+    description = "KnowledgeFS operation quota exceeded."
+    code = 403
+
+
+__all__ = [
+    "KnowledgeFSAccessDeniedHTTPError",
+    "KnowledgeFSInvalidRequestHTTPError",
+    "KnowledgeFSOperationUnavailableHTTPError",
+    "KnowledgeFSQuotaExceededHTTPError",
+    "KnowledgeFSRateLimitHTTPError",
+    "KnowledgeFSSpaceNotFoundHTTPError",
+    "KnowledgeFSUpstreamUnavailableHTTPError",
+]

api/controllers/console/remote_files.py

@@ -46,6 +46,61 @@ class GetRemoteFileInfo(Resource):
         ).model_dump(mode="json")
 
 
+def upload_remote_file_from_request(
+    *,
+    current_user: Account,
+    resource_tenant_id: str | None = None,
+) -> FileWithSignedUrl:
+    """Validate the JSON request, fetch its remote file, and persist it under the requested tenant."""
+    payload = RemoteFileUploadPayload.model_validate(console_ns.payload)
+    url = payload.url
+
+    # Try to fetch remote file metadata/content first
+    try:
+        resp = remote_fetcher.make_request("HEAD", url=url)
+        if resp.status_code != httpx.codes.OK:
+            resp = remote_fetcher.make_request("GET", url=url, timeout=3, follow_redirects=True)
+        if resp.status_code != httpx.codes.OK:
+            # Normalize into a user-friendly error message expected by tests
+            raise RemoteFileUploadError(f"Failed to fetch file from {url}: {resp.text}")
+    except httpx.RequestError as e:
+        raise RemoteFileUploadError(f"Failed to fetch file from {url}: {str(e)}")
+
+    file_info = helpers.guess_file_info_from_response(resp)
+
+    # Enforce file size limit with 400 (Bad Request) per tests' expectation
+    if not FileService.is_file_size_within_limit(extension=file_info.extension, file_size=file_info.size):
+        raise FileTooLargeError()
+
+    # Load content if needed
+    content = resp.content if resp.request.method == "GET" else remote_fetcher.make_request("GET", url).content
+
+    try:
+        upload_file = FileService(db.engine).upload_file(
+            filename=file_info.filename,
+            content=content,
+            mimetype=file_info.mimetype,
+            user=current_user,
+            tenant_id=resource_tenant_id,
+            source_url=url,
+        )
+    except services.errors.file.FileTooLargeError as file_too_large_error:
+        raise FileTooLargeError(file_too_large_error.description)
+    except services.errors.file.UnsupportedFileTypeError:
+        raise UnsupportedFileTypeError()
+
+    return FileWithSignedUrl(
+        id=upload_file.id,
+        name=upload_file.name,
+        size=upload_file.size,
+        extension=upload_file.extension,
+        url=file_helpers.get_signed_file_url(upload_file_id=upload_file.id),
+        mime_type=upload_file.mime_type,
+        created_by=upload_file.created_by,
+        created_at=int(upload_file.created_at.timestamp()),
+    )
+
+
 @console_ns.route("/remote-files/upload")
 class RemoteFileUpload(Resource):
     @console_ns.expect(console_ns.models[RemoteFileUploadPayload.__name__])
@@ -53,53 +108,8 @@ class RemoteFileUpload(Resource):
     @login_required
     @with_current_user
     def post(self, current_user: Account):
-        payload = RemoteFileUploadPayload.model_validate(console_ns.payload)
-        url = payload.url
-
-        # Try to fetch remote file metadata/content first
-        try:
-            resp = remote_fetcher.make_request("HEAD", url=url)
-            if resp.status_code != httpx.codes.OK:
-                resp = remote_fetcher.make_request("GET", url=url, timeout=3, follow_redirects=True)
-            if resp.status_code != httpx.codes.OK:
-                # Normalize into a user-friendly error message expected by tests
-                raise RemoteFileUploadError(f"Failed to fetch file from {url}: {resp.text}")
-        except httpx.RequestError as e:
-            raise RemoteFileUploadError(f"Failed to fetch file from {url}: {str(e)}")
-
-        file_info = helpers.guess_file_info_from_response(resp)
-
-        # Enforce file size limit with 400 (Bad Request) per tests' expectation
-        if not FileService.is_file_size_within_limit(extension=file_info.extension, file_size=file_info.size):
-            raise FileTooLargeError()
-
-        # Load content if needed
-        content = resp.content if resp.request.method == "GET" else remote_fetcher.make_request("GET", url).content
-
-        try:
-            upload_file = FileService(db.engine).upload_file(
-                filename=file_info.filename,
-                content=content,
-                mimetype=file_info.mimetype,
-                user=current_user,
-                source_url=url,
-            )
-        except services.errors.file.FileTooLargeError as file_too_large_error:
-            raise FileTooLargeError(file_too_large_error.description)
-        except services.errors.file.UnsupportedFileTypeError:
-            raise UnsupportedFileTypeError()
-
-        # Success: return created resource with 201 status
+        remote_file = upload_remote_file_from_request(current_user=current_user)
         return (
-            FileWithSignedUrl(
-                id=upload_file.id,
-                name=upload_file.name,
-                size=upload_file.size,
-                extension=upload_file.extension,
-                url=file_helpers.get_signed_file_url(upload_file_id=upload_file.id),
-                mime_type=upload_file.mime_type,
-                created_by=upload_file.created_by,
-                created_at=int(upload_file.created_at.timestamp()),
-            ).model_dump(mode="json"),
+            remote_file.model_dump(mode="json"),
             201,
         )

api/controllers/console/socketio/workflow.py

@@ -98,6 +98,7 @@ def handle_collaboration_event(sid, data):
     6. workflow_update
     7. comments_update
     8. node_panel_presence
+    9. graph_view_state (session reports tab visibility; drives leader election)
     """
     return collaboration_service.relay_collaboration_event(sid, data)
 

api/controllers/console/workspace/rbac.py

@@ -346,7 +346,13 @@ class RBACRoleItemApi(Resource):
     def put(self, role_id):
         tenant_id, account_id = _current_ids()
         request = _payload(_RoleUpsertRequest)
-        role = svc.RBACService.Roles.update(tenant_id, account_id, str(role_id), request.to_mutation())
+        role = svc.RBACService.KnowledgeFSRoleMutations.update_role(
+            tenant_id,
+            account_id,
+            str(role_id),
+            request.to_mutation(),
+            session=db.session(),
+        )
         return _dump(role)
 
     @login_required
@@ -356,7 +362,12 @@ class RBACRoleItemApi(Resource):
     @console_ns.response(200, "Success", console_ns.models[svc.RBACRole.__name__])
     def delete(self, role_id):
         tenant_id, account_id = _current_ids()
-        svc.RBACService.Roles.delete(tenant_id, account_id, str(role_id))
+        svc.RBACService.KnowledgeFSRoleMutations.delete_role(
+            tenant_id,
+            account_id,
+            str(role_id),
+            session=db.session(),
+        )
         return {"result": "success"}
 
 
@@ -915,7 +926,7 @@ class RBACMemberRolesApi(Resource):
         tenant_id, account_id = _current_ids()
         request = _payload(_ReplaceMemberRolesRequest)
         return _dump(
-            svc.RBACService.MemberRoles.replace(
+            svc.RBACService.KnowledgeFSRoleMutations.replace_member_roles(
                 tenant_id,
                 account_id,
                 str(member_id),

api/controllers/inner_api/plugin/plugin.py

@@ -8,6 +8,7 @@ from controllers.inner_api.plugin.wraps import get_user_tenant, plugin_data
 from controllers.inner_api.wraps import plugin_inner_api_only
 from core.plugin.backwards_invocation.app import PluginAppBackwardsInvocation
 from core.plugin.backwards_invocation.base import BaseBackwardsInvocationResponse
+from core.plugin.backwards_invocation.datasource import PluginDatasourceBackwardsInvocation
 from core.plugin.backwards_invocation.encrypt import PluginEncrypter
 from core.plugin.backwards_invocation.model import PluginModelBackwardsInvocation
 from core.plugin.backwards_invocation.node import PluginNodeBackwardsInvocation
@@ -15,10 +16,12 @@ from core.plugin.backwards_invocation.tool import PluginToolBackwardsInvocation
 from core.plugin.entities.request import (
     RequestFetchAppInfo,
     RequestInvokeApp,
+    RequestInvokeDatasource,
     RequestInvokeEncrypt,
     RequestInvokeLLM,
     RequestInvokeLLMWithStructuredOutput,
     RequestInvokeModeration,
+    RequestInvokeMultimodalEmbedding,
     RequestInvokeParameterExtractorNode,
     RequestInvokeQuestionClassifierNode,
     RequestInvokeRerank,
@@ -27,6 +30,7 @@ from core.plugin.entities.request import (
     RequestInvokeTextEmbedding,
     RequestInvokeTool,
     RequestInvokeTTS,
+    RequestListModels,
     RequestRequestDownloadFile,
     RequestRequestUploadFile,
 )
@@ -118,6 +122,36 @@ class PluginInvokeTextEmbeddingApi(Resource):
             return jsonable_encoder(BaseBackwardsInvocationResponse(error=str(e)))
 
 
+@inner_api_ns.route("/invoke/multimodal-embedding")
+class PluginInvokeMultimodalEmbeddingApi(Resource):
+    @get_user_tenant
+    @setup_required
+    @plugin_inner_api_only
+    @plugin_data(payload_type=RequestInvokeMultimodalEmbedding)
+    @inner_api_ns.doc("plugin_invoke_multimodal_embedding")
+    @inner_api_ns.doc(description="Invoke multimodal embedding models through Dify model management")
+    @inner_api_ns.doc(
+        responses={
+            200: "Multimodal embedding successful",
+            401: "Unauthorized - invalid API key",
+            404: "Service not available",
+        }
+    )
+    def post(self, user_model: Account | EndUser, tenant_model: Tenant, payload: RequestInvokeMultimodalEmbedding):
+        try:
+            return jsonable_encoder(
+                BaseBackwardsInvocationResponse(
+                    data=PluginModelBackwardsInvocation.invoke_multimodal_embedding(
+                        user_id=user_model.id,
+                        tenant=tenant_model,
+                        payload=payload,
+                    )
+                )
+            )
+        except Exception as e:
+            return jsonable_encoder(BaseBackwardsInvocationResponse(error=str(e)))
+
+
 @inner_api_ns.route("/invoke/rerank")
 class PluginInvokeRerankApi(Resource):
     @get_user_tenant
@@ -144,6 +178,70 @@ class PluginInvokeRerankApi(Resource):
             return jsonable_encoder(BaseBackwardsInvocationResponse(error=str(e)))
 
 
+@inner_api_ns.route("/invoke/model-catalog")
+class PluginModelCatalogApi(Resource):
+    @get_user_tenant
+    @setup_required
+    @plugin_inner_api_only
+    @plugin_data(payload_type=RequestListModels)
+    @inner_api_ns.doc("plugin_model_catalog")
+    @inner_api_ns.doc(description="List tenant-active models managed by Dify")
+    @inner_api_ns.doc(
+        responses={
+            200: "Model catalog lookup successful",
+            401: "Unauthorized - invalid API key",
+            404: "Service not available",
+        }
+    )
+    def post(self, user_model: Account | EndUser, tenant_model: Tenant, payload: RequestListModels):
+        try:
+            return jsonable_encoder(
+                BaseBackwardsInvocationResponse(
+                    data=PluginModelBackwardsInvocation.list_models(
+                        tenant_id=tenant_model.id,
+                        user_id=user_model.id,
+                        payload=payload,
+                    )
+                )
+            )
+        except Exception as e:
+            return jsonable_encoder(BaseBackwardsInvocationResponse(error=str(e)))
+
+
+@inner_api_ns.route("/invoke/datasource")
+class PluginInvokeDatasourceApi(Resource):
+    """Invoke an installed datasource with credentials resolved inside Dify."""
+
+    @get_user_tenant
+    @setup_required
+    @plugin_inner_api_only
+    @plugin_data(payload_type=RequestInvokeDatasource)
+    @inner_api_ns.doc("plugin_invoke_datasource")
+    @inner_api_ns.doc(description="Invoke datasource plugins through Dify credential management")
+    @inner_api_ns.doc(
+        responses={
+            200: "Datasource invocation successful (streaming response)",
+            401: "Unauthorized - invalid API key",
+            404: "Datasource provider, datasource, or credential not found",
+        }
+    )
+    def post(
+        self,
+        user_model: Account | EndUser,
+        tenant_model: Tenant,
+        payload: RequestInvokeDatasource,
+    ):
+        response = PluginDatasourceBackwardsInvocation.invoke(
+            user_id=user_model.id,
+            tenant=tenant_model,
+            payload=payload,
+        )
+        return length_prefixed_response(
+            0xF,
+            PluginDatasourceBackwardsInvocation.convert_to_event_stream(response),
+        )
+
+
 @inner_api_ns.route("/invoke/tts")
 class PluginInvokeTTSApi(Resource):
     @get_user_tenant

api/controllers/service_api/__init__.py

@@ -38,6 +38,7 @@ from .dataset import (
 )
 from .dataset.rag_pipeline import rag_pipeline_workflow
 from .end_user import end_user
+from .knowledge_fs import resources as knowledge_fs_resources
 from .workspace import models
 
 __all__ = [
@@ -54,6 +55,7 @@ __all__ = [
     "hit_testing",
     "human_input_form",
     "index",
+    "knowledge_fs_resources",
     "message",
     "metadata",
     "models",

api/controllers/service_api/knowledge_fs/__init__.py

@@ -0,0 +1,5 @@
+"""KnowledgeFS-specific Service API authenticated by resource credentials."""
+
+from . import resources
+
+__all__ = ["resources"]

api/controllers/service_api/knowledge_fs/error.py

@@ -0,0 +1,49 @@
+"""Stable KnowledgeFS Service API error contract."""
+
+from libs.exception import BaseHTTPException
+
+
+class KnowledgeFSInvalidCredentialHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_invalid_credential"
+    description = "Invalid KnowledgeFS service credential."
+    code = 401
+
+
+class KnowledgeFSServiceOperationUnavailableHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_operation_unavailable"
+    description = "KnowledgeFS operation is not available."
+    code = 503
+
+
+class KnowledgeFSServiceUpstreamUnavailableHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_upstream_unavailable"
+    description = "KnowledgeFS is unavailable."
+    code = 502
+
+
+class KnowledgeFSServiceInvalidRequestHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_invalid_request"
+    description = "KnowledgeFS request is invalid."
+    code = 400
+
+
+class KnowledgeFSServiceRateLimitHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_rate_limit_exceeded"
+    description = "KnowledgeFS operation rate limit exceeded."
+    code = 429
+
+
+class KnowledgeFSServiceQuotaExceededHTTPError(BaseHTTPException):
+    error_code = "knowledge_fs_quota_exceeded"
+    description = "KnowledgeFS operation quota exceeded."
+    code = 403
+
+
+__all__ = [
+    "KnowledgeFSInvalidCredentialHTTPError",
+    "KnowledgeFSServiceInvalidRequestHTTPError",
+    "KnowledgeFSServiceOperationUnavailableHTTPError",
+    "KnowledgeFSServiceQuotaExceededHTTPError",
+    "KnowledgeFSServiceRateLimitHTTPError",
+    "KnowledgeFSServiceUpstreamUnavailableHTTPError",
+]

api/core/agent/base_agent_runner.py

@@ -14,7 +14,10 @@ from core.app.apps.base_app_queue_manager import AppQueueManager
 from core.app.apps.base_app_runner import AppRunner
 from core.app.entities.app_invoke_entities import (
     AgentChatAppGenerateEntity,
+    DifyRunContext,
+    InvokeFrom,
     ModelConfigWithCredentialsEntity,
+    UserFrom,
 )
 from core.app.file_access import DatabaseFileAccessController
 from core.callback_handler.agent_tool_callback_handler import DifyAgentCallbackHandler
@@ -148,6 +151,15 @@ class BaseAgentRunner(AppRunner):
             user_id=self.user_id,
             invoke_from=self.application_generate_entity.invoke_from,
         )
+        invoke_from = self.application_generate_entity.invoke_from
+        if isinstance(invoke_from, InvokeFrom):
+            tool_entity.runtime.dify_run_context = DifyRunContext(
+                tenant_id=self.tenant_id,
+                app_id=self.app_config.app_id,
+                user_id=self.user_id,
+                user_from=(UserFrom.ACCOUNT if invoke_from.runs_as_account() else UserFrom.END_USER),
+                invoke_from=invoke_from,
+            )
         assert tool_entity.entity.description
         message_tool = PromptMessageTool(
             name=tool.tool_name,

api/core/app/apps/agent_app/runtime_request_builder.py

@@ -159,6 +159,7 @@ class AgentAppRuntimeRequestBuilder:
                     user_from=cast(DifyExecutionContextUserFrom, context.dify_context.user_from.value),
                     invoke_from=cast(DifyExecutionContextInvokeFrom, context.dify_context.invoke_from.value),
                     agent_mode="agent_app",
+                    trace_id=context.dify_context.trace_session_id,
                 ),
                 # ENG-616: expand slash-menu mention tokens to canonical names so
                 # no frontend-internal {{#…#}} marker ever reaches the model.

api/core/datasource/__base/datasource_provider.py

@@ -38,6 +38,11 @@ class DatasourcePluginProviderController(ABC):
         ):
             raise ToolProviderCredentialValidationError("Invalid credentials")
 
+    def validate_credentials(self, user_id: str, credentials: dict[str, Any]) -> None:
+        """Validate credential shape and value against this installed provider declaration."""
+        self.validate_credentials_format(credentials)
+        self._validate_credentials(user_id, credentials)
+
     @property
     def provider_type(self) -> DatasourceProviderType:
         """

api/core/helper/ssrf_proxy.py

@@ -47,6 +47,24 @@ class MaxRetriesExceededError(ValueError):
     pass
 
 
+class ResponseLimitError(ValueError):
+    """Base error for responses that cannot be safely bounded."""
+
+    pass
+
+
+class ResponseTooLargeError(ResponseLimitError):
+    """Raised when an identity response exceeds the configured byte limit."""
+
+    pass
+
+
+class UnsupportedResponseEncodingError(ResponseLimitError):
+    """Raised when response encoding prevents safe decoded-size enforcement."""
+
+    pass
+
+
 request_error = httpx.RequestError
 max_retries_exceeded_error = MaxRetriesExceededError
 
@@ -142,7 +160,31 @@ def _inject_trace_headers(headers: Headers | None) -> Headers:
     return headers
 
 
-def make_request(method: str, url: str, max_retries: int = SSRF_DEFAULT_MAX_RETRIES, **kwargs: Any) -> httpx.Response:
+def make_request(
+    method: str,
+    url: str,
+    max_retries: int = SSRF_DEFAULT_MAX_RETRIES,
+    stream_response: bool = False,
+    **kwargs: Any,
+) -> httpx.Response:
+    """Send one SSRF-protected request with optional streaming.
+
+    Args:
+        method: HTTP method sent through the configured SSRF client.
+        url: Absolute request URL.
+        max_retries: Number of retry attempts after the initial request.
+        stream_response: Return an open streaming response that the caller must close.
+        **kwargs: Additional keyword arguments forwarded to ``httpx.Client``.
+
+    Returns:
+        A buffered response, or an open response when ``stream_response`` is true.
+
+    Raises:
+        ToolSSRFError: The configured SSRF proxy rejects the destination.
+        MaxRetriesExceededError: All configured request attempts fail.
+        httpx.RequestError: A request fails while retries are disabled.
+        ValueError: The SSL verification option or request headers are invalid.
+    """
     # Convert requests-style allow_redirects to httpx-style follow_redirects
     if "allow_redirects" in kwargs:
         allow_redirects = kwargs.pop("allow_redirects")
@@ -175,6 +217,11 @@ def make_request(method: str, url: str, max_retries: int = SSRF_DEFAULT_MAX_RETR
     # When using a forward proxy, httpx may override the Host header based on the URL.
     # We extract and preserve any explicitly set Host header to support virtual hosting.
     user_provided_host = _get_user_provided_host_header(headers)
+    send_kwargs: dict[str, Any] = {}
+    if "auth" in kwargs:
+        send_kwargs["auth"] = kwargs.pop("auth")
+    if "follow_redirects" in kwargs:
+        send_kwargs["follow_redirects"] = kwargs.pop("follow_redirects")
 
     retries = 0
     while retries <= max_retries:
@@ -185,7 +232,11 @@ def make_request(method: str, url: str, max_retries: int = SSRF_DEFAULT_MAX_RETR
             if user_provided_host is not None:
                 headers["host"] = user_provided_host
             kwargs["headers"] = headers
-            response = client.request(method=method, url=url, **kwargs)
+            request = client.build_request(method=method, url=url, **kwargs)
+            if stream_response:
+                response = client.send(request, stream=True, **send_kwargs)
+            else:
+                response = client.send(request, **send_kwargs)
 
             # Check for SSRF protection by Squid proxy
             if response.status_code in (401, 403):
@@ -195,6 +246,7 @@ def make_request(method: str, url: str, max_retries: int = SSRF_DEFAULT_MAX_RETR
 
                 # Squid typically identifies itself in Server or Via headers
                 if "squid" in server_header or "squid" in via_header:
+                    response.close()
                     raise ToolSSRFError(
                         f"Access to '{url}' was blocked by SSRF protection. "
                         f"The URL may point to a private or local network address. "
@@ -208,6 +260,7 @@ def make_request(method: str, url: str, max_retries: int = SSRF_DEFAULT_MAX_RETR
                     response.status_code,
                     url,
                 )
+                response.close()
 
         except httpx.RequestError as e:
             logger.warning("Request to URL %s failed on attempt %s: %s", url, retries + 1, e)
@@ -220,6 +273,42 @@ def make_request(method: str, url: str, max_retries: int = SSRF_DEFAULT_MAX_RETR
     raise MaxRetriesExceededError(f"Reached maximum retries ({max_retries}) for URL {url}")
 
 
+def buffer_response(response: httpx.Response, *, max_response_bytes: int) -> httpx.Response:
+    """Consume one open identity response under a decoded byte limit and close its stream."""
+    if max_response_bytes <= 0:
+        raise ValueError("max_response_bytes must be positive")
+
+    try:
+        content_encoding = response.headers.get("content-encoding", "identity").strip().lower()
+        if content_encoding not in {"", "identity"}:
+            raise UnsupportedResponseEncodingError(f"content encoding {content_encoding} cannot be safely bounded")
+        content = bytearray()
+        for chunk in response.iter_bytes():
+            if len(content) + len(chunk) > max_response_bytes:
+                raise ResponseTooLargeError(f"response exceeded {max_response_bytes} bytes")
+            content.extend(chunk)
+        decoded_headers = {
+            name: value
+            for name, value in response.headers.items()
+            if name.lower() not in {"content-encoding", "content-length", "transfer-encoding"}
+        }
+        try:
+            request = response.request
+        except RuntimeError:
+            request = None
+        return httpx.Response(
+            response.status_code,
+            headers=decoded_headers,
+            content=bytes(content),
+            request=request,
+            extensions=response.extensions,
+            history=response.history,
+            default_encoding=response.default_encoding,
+        )
+    finally:
+        response.close()
+
+
 def get(url: str, max_retries: int = SSRF_DEFAULT_MAX_RETRIES, **kwargs: Any) -> httpx.Response:
     return make_request("GET", url, max_retries=max_retries, **kwargs)
 

api/core/plugin/backwards_invocation/datasource.py

@@ -0,0 +1,110 @@
+"""Datasource backward invocation through Dify's tenant-bound runtime.
+
+This module is the internal service boundary used by trusted callers such as
+KnowledgeFS. It resolves installed provider declarations and Dify-owned
+credential references before reaching ``PluginDatasourceManager``; callers
+must never provide raw datasource credentials or a plugin-daemon API key.
+"""
+
+from collections.abc import Generator
+from typing import Any, cast
+
+from pydantic import BaseModel
+
+from core.datasource.datasource_manager import DatasourceManager
+from core.datasource.entities.datasource_entities import (
+    OnlineDriveBrowseFilesRequest,
+    OnlineDriveDownloadFileRequest,
+)
+from core.datasource.online_document.online_document_plugin import OnlineDocumentDatasourcePlugin
+from core.datasource.online_drive.online_drive_plugin import OnlineDriveDatasourcePlugin
+from core.datasource.website_crawl.website_crawl_plugin import WebsiteCrawlDatasourcePlugin
+from core.plugin.backwards_invocation.base import BaseBackwardsInvocation
+from core.plugin.entities.request import RequestInvokeDatasource
+from models.account import Tenant
+from models.provider_ids import DatasourceProviderID
+from services.datasource_provider_service import DatasourceProviderService
+
+
+class PluginDatasourceBackwardsInvocation(BaseBackwardsInvocation):
+    """Resolve and invoke a datasource without exposing credential material to the caller."""
+
+    @classmethod
+    def invoke(
+        cls,
+        *,
+        user_id: str,
+        tenant: Tenant,
+        payload: RequestInvokeDatasource,
+    ) -> Generator[BaseModel | dict[str, Any], None, None]:
+        """Yield datasource messages for one validated inner-runtime request."""
+        provider_id = DatasourceProviderID(payload.provider)
+        canonical_provider_id = str(provider_id)
+        controller = DatasourceManager.get_datasource_plugin_provider(
+            provider_id=canonical_provider_id,
+            tenant_id=tenant.id,
+            datasource_type=payload.datasource_type,
+        )
+        if controller.entity.provider_type != payload.datasource_type:
+            raise ValueError("Datasource provider type mismatch")
+
+        # Resolving the datasource from the installed declaration prevents a caller
+        # from dispatching an arbitrary datasource name under a valid plugin ID.
+        runtime = controller.get_datasource(payload.datasource)
+        credentials = DatasourceProviderService().get_datasource_credentials(
+            tenant_id=tenant.id,
+            provider=provider_id.provider_name,
+            plugin_id=provider_id.plugin_id,
+            credential_id=payload.credential_id,
+        )
+        if controller.need_credentials and not credentials:
+            raise ValueError("Datasource credential not found")
+
+        if payload.operation == "validate_credentials":
+            controller.validate_credentials(user_id=user_id, credentials=credentials)
+            yield {"result": True}
+            return
+
+        runtime.runtime.credentials = credentials
+        provider_type = runtime.datasource_provider_type()
+
+        match payload.operation:
+            case "get_website_crawl":
+                website = cast(WebsiteCrawlDatasourcePlugin, runtime)
+                yield from website.get_website_crawl(
+                    user_id=user_id,
+                    datasource_parameters=payload.datasource_parameters,
+                    provider_type=provider_type,
+                )
+            case "get_online_document_pages":
+                document = cast(OnlineDocumentDatasourcePlugin, runtime)
+                yield from document.get_online_document_pages(
+                    user_id=user_id,
+                    datasource_parameters=payload.datasource_parameters,
+                    provider_type=provider_type,
+                )
+            case "get_online_document_page_content":
+                if payload.page is None:
+                    raise ValueError("Online-document page input is required")
+                document = cast(OnlineDocumentDatasourcePlugin, runtime)
+                yield from document.get_online_document_page_content(
+                    user_id=user_id,
+                    datasource_parameters=payload.page,
+                    provider_type=provider_type,
+                )
+            case "online_drive_browse_files":
+                drive = cast(OnlineDriveDatasourcePlugin, runtime)
+                yield from drive.online_drive_browse_files(
+                    user_id=user_id,
+                    request=OnlineDriveBrowseFilesRequest.model_validate(payload.request),
+                    provider_type=provider_type,
+                )
+            case "online_drive_download_file":
+                drive = cast(OnlineDriveDatasourcePlugin, runtime)
+                yield from drive.online_drive_download_file(
+                    user_id=user_id,
+                    request=OnlineDriveDownloadFileRequest.model_validate(payload.request),
+                    provider_type=provider_type,
+                )
+            case _:
+                raise ValueError(f"Unsupported datasource operation: {payload.operation}")

api/core/plugin/backwards_invocation/model.py

@@ -1,22 +1,31 @@
 import tempfile
 from binascii import hexlify, unhexlify
-from collections.abc import Generator
+from collections.abc import Generator, Mapping
+from enum import Enum
 from typing import Any
 
+from pydantic import BaseModel
+
 from core.app.llm import deduct_llm_quota
 from core.llm_generator.output_parser.structured_output import invoke_llm_with_structured_output
 from core.model_manager import ModelManager
 from core.plugin.backwards_invocation.base import BaseBackwardsInvocation
 from core.plugin.entities.request import (
+    InvokableModelCatalogItem,
+    InvokableModelCatalogPage,
     RequestInvokeLLM,
     RequestInvokeLLMWithStructuredOutput,
     RequestInvokeModeration,
+    RequestInvokeMultimodalEmbedding,
     RequestInvokeRerank,
     RequestInvokeSpeech2Text,
     RequestInvokeSummary,
     RequestInvokeTextEmbedding,
     RequestInvokeTTS,
+    RequestListModels,
 )
+from core.plugin.impl.model_runtime_factory import create_plugin_provider_manager
+from core.plugin.plugin_service import PluginService
 from core.tools.entities.tool_entities import ToolProviderType
 from core.tools.utils.model_invocation_utils import ModelInvocationUtils
 from graphon.model_runtime.entities.llm_entities import (
@@ -33,6 +42,20 @@ from graphon.model_runtime.entities.message_entities import (
 )
 from graphon.model_runtime.entities.model_entities import ModelType
 from models.account import Tenant
+from models.provider_ids import ModelProviderID
+
+
+def _json_compatible(value: Any) -> Any:
+    """Convert model-runtime metadata into stable JSON-compatible values."""
+    if isinstance(value, BaseModel):
+        return value.model_dump(mode="json")
+    if isinstance(value, Enum):
+        return value.value
+    if isinstance(value, Mapping):
+        return {str(_json_compatible(key)): _json_compatible(child) for key, child in value.items()}
+    if isinstance(value, list | tuple | set):
+        return [_json_compatible(child) for child in value]
+    return value
 
 
 class PluginModelBackwardsInvocation(BaseBackwardsInvocation):
@@ -183,7 +206,30 @@ class PluginModelBackwardsInvocation(BaseBackwardsInvocation):
         )
 
         # invoke model
-        response = model_instance.invoke_text_embedding(texts=payload.texts)
+        response = model_instance.invoke_text_embedding(texts=payload.texts, input_type=payload.input_type)
+
+        return response
+
+    @classmethod
+    def invoke_multimodal_embedding(
+        cls,
+        user_id: str,
+        tenant: Tenant,
+        payload: RequestInvokeMultimodalEmbedding,
+    ):
+        """Invoke multimodal embedding through the tenant-bound model instance."""
+        model_instance = cls._get_bound_model_instance(
+            tenant_id=tenant.id,
+            user_id=user_id,
+            provider=payload.provider,
+            model_type=payload.model_type,
+            model=payload.model,
+        )
+
+        response = model_instance.invoke_multimodal_embedding(
+            multimodel_documents=[document.model_dump(exclude_none=True) for document in payload.documents],
+            input_type=payload.input_type,
+        )
 
         return response
 
@@ -210,6 +256,67 @@ class PluginModelBackwardsInvocation(BaseBackwardsInvocation):
 
         return response
 
+    @classmethod
+    def list_models(
+        cls,
+        tenant_id: str,
+        user_id: str,
+        payload: RequestListModels,
+    ) -> InvokableModelCatalogPage:
+        """List only models that are active for the tenant's Dify configuration."""
+        provider_manager = create_plugin_provider_manager(tenant_id=tenant_id, user_id=user_id)
+        active_models = provider_manager.get_configurations(tenant_id).get_models(
+            model_type=payload.model_type,
+            only_active=True,
+        )
+
+        installed_identities: dict[str, str] = {}
+        for plugin in PluginService.list(tenant_id):
+            existing = installed_identities.get(plugin.plugin_id)
+            if existing is not None and existing != plugin.plugin_unique_identifier:
+                raise ValueError(f"Ambiguous installed identity for model plugin {plugin.plugin_id}")
+            installed_identities[plugin.plugin_id] = plugin.plugin_unique_identifier
+
+        requested_provider = str(ModelProviderID(payload.provider)) if payload.provider else None
+        matched_models = [
+            model
+            for model in active_models
+            if (requested_provider is None or model.provider.provider == requested_provider)
+            and (payload.model is None or model.model == payload.model)
+        ]
+        matched_models.sort(key=lambda model: (model.provider.provider, model.model))
+
+        page_models = matched_models[payload.offset : payload.offset + payload.limit]
+        items: list[InvokableModelCatalogItem] = []
+        for model in page_models:
+            provider_id = ModelProviderID(model.provider.provider)
+            unique_identifier = installed_identities.get(provider_id.plugin_id)
+            if unique_identifier is None:
+                raise ValueError(f"Installed identity not found for active model plugin {provider_id.plugin_id}")
+            items.append(
+                InvokableModelCatalogItem(
+                    plugin_id=provider_id.plugin_id,
+                    plugin_unique_identifier=unique_identifier,
+                    provider=provider_id.provider_name,
+                    model=model.model,
+                    model_type=model.model_type,
+                    capabilities={
+                        "deprecated": model.deprecated,
+                        "features": _json_compatible(model.features or []),
+                        "fetchFrom": _json_compatible(model.fetch_from),
+                        "modelProperties": _json_compatible(model.model_properties),
+                        "modelType": model.model_type.value,
+                        "status": _json_compatible(model.status),
+                    },
+                )
+            )
+
+        next_offset = payload.offset + len(page_models)
+        return InvokableModelCatalogPage(
+            items=items,
+            next_offset=next_offset if next_offset < len(matched_models) else None,
+        )
+
     @classmethod
     def invoke_tts(cls, user_id: str, tenant: Tenant, payload: RequestInvokeTTS):
         """

api/core/plugin/entities/request.py

@@ -6,6 +6,11 @@ from typing import Any, Literal
 from flask import Response
 from pydantic import BaseModel, ConfigDict, Field, field_validator, model_validator
 
+from core.datasource.entities.datasource_entities import (
+    DatasourceProviderType,
+    GetOnlineDocumentPageContentRequest,
+)
+from core.entities.embedding_type import EmbeddingInputType
 from core.entities.provider_entities import BasicProviderConfig
 from core.plugin.utils.http_parser import deserialize_response
 from core.workflow.file_reference import is_canonical_file_reference
@@ -54,6 +59,61 @@ class RequestInvokeTool(BaseModel):
     credential_id: str | None = None
 
 
+DatasourceInvocationOperation = Literal[
+    "get_online_document_page_content",
+    "get_online_document_pages",
+    "get_website_crawl",
+    "online_drive_browse_files",
+    "online_drive_download_file",
+    "validate_credentials",
+]
+
+
+class RequestInvokeDatasource(BaseModel):
+    """Invoke one installed datasource using a Dify-owned credential reference.
+
+    Raw credentials are intentionally not part of this contract. ``tenant_id`` and
+    ``user_id`` are consumed by the inner-API request context, while the remaining
+    fields select an installed provider declaration and an operation-specific input.
+    """
+
+    tenant_id: str = Field(min_length=1, max_length=512)
+    user_id: str = Field(min_length=1, max_length=512)
+    provider: str = Field(min_length=1, max_length=768)
+    datasource: str = Field(min_length=1, max_length=256)
+    datasource_type: DatasourceProviderType
+    credential_id: str = Field(min_length=1, max_length=512)
+    operation: DatasourceInvocationOperation
+    datasource_parameters: dict[str, Any] = Field(default_factory=dict)
+    page: GetOnlineDocumentPageContentRequest | None = None
+    request: dict[str, Any] | None = None
+
+    model_config = ConfigDict(extra="forbid")
+
+    @model_validator(mode="after")
+    def validate_operation_payload(self) -> "RequestInvokeDatasource":
+        expected_type = {
+            "get_online_document_page_content": DatasourceProviderType.ONLINE_DOCUMENT,
+            "get_online_document_pages": DatasourceProviderType.ONLINE_DOCUMENT,
+            "get_website_crawl": DatasourceProviderType.WEBSITE_CRAWL,
+            "online_drive_browse_files": DatasourceProviderType.ONLINE_DRIVE,
+            "online_drive_download_file": DatasourceProviderType.ONLINE_DRIVE,
+            "validate_credentials": self.datasource_type,
+        }[self.operation]
+        if self.datasource_type != expected_type:
+            raise ValueError(f"{self.operation} requires datasource_type {expected_type.value}")
+
+        page_required = self.operation == "get_online_document_page_content"
+        if page_required != (self.page is not None):
+            raise ValueError("page is required only for get_online_document_page_content")
+
+        request_required = self.operation in {"online_drive_browse_files", "online_drive_download_file"}
+        if request_required != (self.request is not None):
+            raise ValueError("request is required only for online-drive operations")
+
+        return self
+
+
 class BaseRequestInvokeModel(BaseModel):
     provider: str
     model: str
@@ -115,6 +175,25 @@ class RequestInvokeTextEmbedding(BaseRequestInvokeModel):
 
     model_type: ModelType = ModelType.TEXT_EMBEDDING
     texts: list[str]
+    input_type: EmbeddingInputType = EmbeddingInputType.DOCUMENT
+
+
+class MultimodalEmbeddingDocument(BaseModel):
+    """A document accepted by a multimodal text-embedding model."""
+
+    content: str
+    content_type: str
+    file_id: str | None = None
+
+    model_config = ConfigDict(extra="forbid")
+
+
+class RequestInvokeMultimodalEmbedding(BaseRequestInvokeModel):
+    """Request to invoke a multimodal text-embedding model."""
+
+    model_type: ModelType = ModelType.TEXT_EMBEDDING
+    documents: list[MultimodalEmbeddingDocument] = Field(min_length=1)
+    input_type: EmbeddingInputType = EmbeddingInputType.DOCUMENT
 
 
 class RequestInvokeRerank(BaseRequestInvokeModel):
@@ -125,8 +204,40 @@ class RequestInvokeRerank(BaseRequestInvokeModel):
     model_type: ModelType = ModelType.RERANK
     query: str
     docs: list[str]
-    score_threshold: float
-    top_n: int
+    score_threshold: float | None = None
+    top_n: int | None = None
+
+
+class RequestListModels(BaseModel):
+    """Tenant-scoped query for models that Dify can invoke."""
+
+    model_type: Literal[ModelType.LLM, ModelType.TEXT_EMBEDDING, ModelType.RERANK]
+    provider: str | None = None
+    model: str | None = None
+    offset: int = Field(default=0, ge=0)
+    limit: int = Field(default=50, ge=1, le=100)
+
+    model_config = ConfigDict(protected_namespaces=())
+
+
+class InvokableModelCatalogItem(BaseModel):
+    """Installed identity and active Dify capability metadata for one model."""
+
+    plugin_id: str
+    plugin_unique_identifier: str
+    provider: str
+    model: str
+    model_type: ModelType
+    capabilities: dict[str, Any] = Field(default_factory=dict)
+
+    model_config = ConfigDict(protected_namespaces=())
+
+
+class InvokableModelCatalogPage(BaseModel):
+    """Offset page returned by the internal model catalog endpoint."""
+
+    items: list[InvokableModelCatalogItem] = Field(default_factory=list)
+    next_offset: int | None = None
 
 
 class RequestInvokeTTS(BaseRequestInvokeModel):

api/core/rbac/entities.py

@@ -10,6 +10,7 @@ class RBACResourceScope(StrEnum):
 
     APP = "app"
     DATASET = "dataset"
+    KNOWLEDGE_FS = "knowledge_space"
     WORKSPACE = "workspace"
 
 
@@ -57,6 +58,15 @@ class RBACPermission(StrEnum):
     DATASET_EXTERNAL_CONNECT = "dataset_external_connect"
     DATASET_IMPORT_EXPORT_DSL = "dataset_import_export_dsl"
 
+    KNOWLEDGE_FS_READ = "knowledge_space_read"
+    KNOWLEDGE_FS_CREATE = "knowledge_space_create"
+    KNOWLEDGE_FS_EDIT = "knowledge_space_edit"
+    KNOWLEDGE_FS_DELETE = "knowledge_space_delete"
+    KNOWLEDGE_FS_ACCESS_CONFIG = "knowledge_space_access_config"
+    KNOWLEDGE_FS_API_KEY_MANAGE = "knowledge_space_api_key_manage"
+    KNOWLEDGE_FS_DOCUMENT_WRITE = "knowledge_space_document_write"
+    KNOWLEDGE_FS_QUERY = "knowledge_space_query"
+
     WORKSPACE_MEMBER_MANAGE = "workspace_member_manage"
     WORKSPACE_ROLE_MANAGE = "workspace_role_manage"
     API_EXTENSION_MANAGE = "api_extension_manage"

api/core/tools/__base/tool_runtime.py

@@ -2,7 +2,7 @@ from typing import Any
 
 from pydantic import BaseModel, Field
 
-from core.app.entities.app_invoke_entities import InvokeFrom
+from core.app.entities.app_invoke_entities import DifyRunContext, InvokeFrom
 from core.plugin.entities.plugin_daemon import CredentialType
 from core.tools.entities.tool_entities import ToolInvokeFrom
 
@@ -20,6 +20,7 @@ class ToolRuntime(BaseModel):
     tool_id: str | None = None
     invoke_from: InvokeFrom | None = None
     tool_invoke_from: ToolInvokeFrom | None = None
+    dify_run_context: DifyRunContext | None = Field(default=None, exclude=True, repr=False)
     credentials: dict[str, Any] = Field(default_factory=dict)
     credential_type: CredentialType = Field(default=CredentialType.API_KEY)
     runtime_parameters: dict[str, Any] = Field(default_factory=dict)

api/core/tools/builtin_tool/_position.yaml

@@ -1,4 +1,5 @@
 - audio
 - code
+- knowledge_fs
 - time
 - webscraper

api/core/tools/builtin_tool/providers/knowledge_fs/_assets/icon.svg

@@ -0,0 +1,6 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" fill="none">
+  <rect width="64" height="64" rx="14" fill="#155EEF"/>
+  <path d="M17 17h20c5.5 0 10 4.5 10 10v20H27c-5.5 0-10-4.5-10-10V17Z" fill="white" fill-opacity=".96"/>
+  <path d="M26 27h12M26 34h12M26 41h7" stroke="#155EEF" stroke-width="4" stroke-linecap="round"/>
+</svg>
+

api/core/tools/builtin_tool/providers/knowledge_fs/knowledge_fs.py

@@ -0,0 +1,9 @@
+from typing import Any, override
+
+from core.tools.builtin_tool.provider import BuiltinToolProviderController
+
+
+class KnowledgeFSProvider(BuiltinToolProviderController):
+    @override
+    def _validate_credentials(self, user_id: str, credentials: dict[str, Any]) -> None:
+        _ = (user_id, credentials)

api/core/tools/builtin_tool/providers/knowledge_fs/knowledge_fs.yaml

@@ -0,0 +1,13 @@
+identity:
+  author: Dify
+  name: knowledge_fs
+  label:
+    en_US: KnowledgeFS
+    zh_Hans: KnowledgeFS
+  description:
+    en_US: Run explicitly bound KnowledgeFS operations from an Agent or Workflow.
+    zh_Hans: 从 Agent 或 Workflow 执行显式绑定的 KnowledgeFS 操作。
+  icon: icon.svg
+  tags:
+    - rag
+

api/core/tools/builtin_tool/providers/knowledge_fs/tools/create_research_task.py

@@ -0,0 +1,60 @@
+from collections.abc import Generator
+from typing import Any, override
+
+from pydantic import ValidationError
+from sqlalchemy.orm import Session, sessionmaker
+
+from core.tools.builtin_tool.tool import BuiltinTool
+from core.tools.entities.tool_entities import ToolInvokeFrom, ToolInvokeMessage
+from core.tools.errors import ToolInvokeError
+from models.knowledge_fs import KnowledgeFSAppSpaceJoinType
+from services.knowledge_fs.app_execution_capability import KnowledgeResourceRef
+from services.knowledge_fs.product_dto import KnowledgeFSResearchTaskCreatePayload
+from services.knowledge_fs.runtime import create_knowledge_fs_runtime
+
+
+class KnowledgeFSCreateResearchTaskTool(BuiltinTool):
+    @override
+    def _invoke(
+        self,
+        session: Session,
+        user_id: str,
+        tool_parameters: dict[str, Any],
+        conversation_id: str | None = None,
+        app_id: str | None = None,
+        message_id: str | None = None,
+    ) -> Generator[ToolInvokeMessage, None, None]:
+        _ = (user_id, conversation_id, app_id, message_id)
+        run_context = self.runtime.dify_run_context
+        if run_context is None or self.runtime.tenant_id != run_context.tenant_id:
+            raise ToolInvokeError("KnowledgeFS requires a trusted Dify run context")
+        match self.runtime.tool_invoke_from:
+            case ToolInvokeFrom.AGENT:
+                caller_kind = KnowledgeFSAppSpaceJoinType.AGENT
+            case ToolInvokeFrom.WORKFLOW:
+                caller_kind = KnowledgeFSAppSpaceJoinType.WORKFLOW
+            case _:
+                raise ToolInvokeError("KnowledgeFS is only available to Agent and Workflow callers")
+
+        try:
+            resource = KnowledgeResourceRef.model_validate(tool_parameters.get("resource"))
+            payload_data: dict[str, object] = {
+                "query": tool_parameters.get("query"),
+            }
+            mode = tool_parameters.get("mode")
+            if mode:
+                payload_data["mode"] = mode
+            payload = KnowledgeFSResearchTaskCreatePayload.model_validate(payload_data)
+            runtime = create_knowledge_fs_runtime(sessionmaker(bind=session.get_bind(), expire_on_commit=False))
+            response = runtime.app_capabilities.create_research_task(
+                run_context=run_context,
+                caller_kind=caller_kind,
+                resource=resource,
+                payload=payload,
+            )
+        except ToolInvokeError:
+            raise
+        except (ValidationError, RuntimeError, ValueError) as exc:
+            raise ToolInvokeError(str(exc)) from exc
+
+        yield self.create_json_message(response.model_dump(mode="json", by_alias=True))

api/core/tools/builtin_tool/providers/knowledge_fs/tools/create_research_task.yaml

@@ -0,0 +1,75 @@
+identity:
+  name: create_research_task
+  author: Dify
+  label:
+    en_US: Create KnowledgeFS Research Task
+    zh_Hans: 创建 KnowledgeFS Research 任务
+description:
+  human:
+    en_US: Create a Research task in an explicitly bound KnowledgeFS space.
+    zh_Hans: 在显式绑定的 KnowledgeFS 空间中创建 Research 任务。
+  llm: Create a durable research task using an explicitly configured KnowledgeFS resource.
+parameters:
+  - name: resource
+    type: object
+    required: true
+    label:
+      en_US: KnowledgeFS resource
+      zh_Hans: KnowledgeFS 资源
+    human_description:
+      en_US: A typed KnowledgeFS control-space reference configured by the app author.
+      zh_Hans: 由应用作者配置的类型化 KnowledgeFS control-space 引用。
+    form: form
+    input_schema:
+      type: object
+      additionalProperties: false
+      properties:
+        kind:
+          type: string
+          const: knowledge_fs
+        control_space_id:
+          type: string
+          minLength: 1
+      required:
+        - kind
+        - control_space_id
+  - name: query
+    type: string
+    required: true
+    label:
+      en_US: Research query
+      zh_Hans: Research 查询
+    human_description:
+      en_US: The question the Research task should investigate.
+      zh_Hans: Research 任务需要调查的问题。
+    llm_description: The question to investigate with KnowledgeFS.
+    form: llm
+  - name: mode
+    type: select
+    required: false
+    label:
+      en_US: Mode
+      zh_Hans: 模式
+    human_description:
+      en_US: Optional KnowledgeFS retrieval mode.
+      zh_Hans: 可选的 KnowledgeFS 检索模式。
+    llm_description: Optional retrieval mode. Use auto unless the task needs a specific mode.
+    form: llm
+    options:
+      - value: auto
+        label:
+          en_US: Auto
+          zh_Hans: 自动
+      - value: fast
+        label:
+          en_US: Fast
+          zh_Hans: 快速
+      - value: deep
+        label:
+          en_US: Deep
+          zh_Hans: 深度
+      - value: research
+        label:
+          en_US: Research
+          zh_Hans: Research
+

api/core/workflow/node_runtime.py

@@ -493,6 +493,7 @@ class DifyToolNodeRuntime(ToolNodeRuntimeProtocol):
                 self._run_context.invoke_from,
                 variable_pool,
             )
+            tool_runtime.runtime.dify_run_context = self._run_context
         except ToolNodeError:
             raise
         except Exception as exc:

api/core/workflow/nodes/agent_v2/runtime_request_builder.py

@@ -250,6 +250,7 @@ class WorkflowAgentRuntimeRequestBuilder:
                     agent_config_version_kind="snapshot",
                     agent_mode=self._agent_backend_agent_mode(context.dify_context.invoke_from),
                     invoke_from=cast(DifyExecutionContextInvokeFrom, context.dify_context.invoke_from.value),
+                    trace_id=context.dify_context.trace_session_id,
                 ),
                 agent_soul_prompt=soul_prompt or None,
                 workflow_node_job_prompt=workflow_job_prompt,

api/dev/check_knowledge_fs_coverage.py

@@ -0,0 +1,346 @@
+"""Enforce focused KnowledgeFS coverage without hiding integration-critical glue.
+
+The primary threshold aggregates statement and branch coverage for every Dify
+module owned by the KnowledgeFS integration. Large pre-existing Dify modules
+that only contain narrow integration hooks are checked with changed-line
+coverage instead, so unrelated legacy code cannot dilute or inflate the gate.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import logging
+import os
+import re
+import subprocess
+from dataclasses import dataclass
+from pathlib import Path
+from typing import TypedDict, cast
+
+WORKSPACE_ROOT = Path(__file__).resolve().parents[2]
+NON_CORE_COVERAGE_ALLOWLIST = frozenset(
+    {
+        "api/dev/check_knowledge_fs_coverage.py",
+        "api/dev/generate_knowledge_fs_contract.py",
+        "api/dev/knowledge_fs_product_contract.py",
+        "api/migrations/versions/2026_07_21_1200-a4e7c2f91b30_add_knowledge_fs_control_plane.py",
+        "api/migrations/versions/2026_07_21_1300-b7f2a9d41c60_add_knowledge_fs_cutover.py",
+        "api/migrations/versions/2026_07_21_1400-c8e31b7d52a4_add_knowledge_fs_cleanup_authorization.py",
+    }
+)
+HUNK_HEADER = re.compile(r"^@@ -\d+(?:,\d+)? \+(\d+)(?:,\d+)? @@")
+logger = logging.getLogger(__name__)
+
+
+class CoverageSummary(TypedDict):
+    """Coverage.py counts required by the aggregate gate."""
+
+    covered_lines: int
+    num_statements: int
+    covered_branches: int
+    num_branches: int
+
+
+class CoverageFile(TypedDict):
+    """Per-file coverage data emitted by ``coverage json``."""
+
+    executed_lines: list[int]
+    missing_lines: list[int]
+    summary: CoverageSummary
+
+
+class CoverageReport(TypedDict):
+    """Relevant top-level shape of a coverage.py JSON report."""
+
+    files: dict[str, CoverageFile]
+
+
+@dataclass(frozen=True, slots=True)
+class CoverageTotals:
+    """Covered and measurable units for one gate surface."""
+
+    covered: int
+    total: int
+
+    @property
+    def percent(self) -> float:
+        return 100.0 if self.total == 0 else self.covered * 100 / self.total
+
+
+class CoverageGateError(RuntimeError):
+    """Raised when coverage input is incomplete or below its threshold."""
+
+
+def main() -> None:
+    """Validate focused module coverage and changed integration glue."""
+    logging.basicConfig(level=logging.INFO, format="%(message)s")
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--coverage-json", type=Path, required=True)
+    parser.add_argument("--glue-manifest", type=Path, required=True)
+    parser.add_argument("--workspace-root", type=Path, default=WORKSPACE_ROOT)
+    parser.add_argument("--base", default="")
+    parser.add_argument("--minimum", type=float, default=90.0)
+    parser.add_argument("--glue-minimum", type=float, default=90.0)
+    args = parser.parse_args()
+
+    workspace_root = args.workspace_root.resolve()
+    report = load_coverage_report(args.coverage_json)
+    core_totals = validate_core_coverage(report, workspace_root=workspace_root, minimum=args.minimum)
+    base = resolve_diff_base(workspace_root, args.base)
+    glue_paths = load_glue_coverage_paths(args.glue_manifest, workspace_root=workspace_root)
+    changed_lines = collect_changed_glue_lines(workspace_root, base, glue_paths=glue_paths)
+    glue_totals = validate_changed_glue_coverage(
+        report,
+        changed_lines=changed_lines,
+        minimum=args.glue_minimum,
+    )
+    logger.info(
+        "Dify KnowledgeFS coverage passed: core lines+branches %.2f%% (%d/%d); changed glue lines %.2f%% (%d/%d)",
+        core_totals.percent,
+        core_totals.covered,
+        core_totals.total,
+        glue_totals.percent,
+        glue_totals.covered,
+        glue_totals.total,
+    )
+
+
+def load_coverage_report(path: Path) -> CoverageReport:
+    """Load the detailed JSON report used by both coverage checks."""
+    if not path.is_file():
+        raise CoverageGateError(f"coverage JSON does not exist: {path}")
+    document = json.loads(path.read_text())
+    if not isinstance(document, dict) or not isinstance(document.get("files"), dict):
+        raise CoverageGateError(f"coverage JSON has no files object: {path}")
+    return cast(CoverageReport, document)
+
+
+def load_glue_coverage_paths(path: Path, *, workspace_root: Path) -> tuple[str, ...]:
+    """Load the workflow's authoritative NUL-delimited integration touchpoints."""
+    if not path.is_file():
+        raise CoverageGateError(f"KnowledgeFS glue manifest does not exist: {path}")
+    try:
+        paths = tuple(item.decode() for item in path.read_bytes().split(b"\0") if item)
+    except UnicodeDecodeError as error:
+        raise CoverageGateError(f"KnowledgeFS glue manifest is not UTF-8: {path}") from error
+    if not paths:
+        raise CoverageGateError("Dify KnowledgeFS glue coverage target set is empty")
+    if len(paths) != len(set(paths)):
+        raise CoverageGateError("Dify KnowledgeFS glue coverage manifest contains duplicate paths")
+    invalid_paths = [
+        candidate
+        for candidate in paths
+        if not candidate.startswith("api/")
+        or not candidate.endswith(".py")
+        or not (workspace_root / candidate).is_file()
+    ]
+    if invalid_paths:
+        raise CoverageGateError(f"Dify KnowledgeFS glue coverage paths are invalid: {', '.join(invalid_paths)}")
+    return paths
+
+
+def is_core_coverage_path(path: str) -> bool:
+    """Return whether a repository-relative path belongs to the focused aggregate."""
+    if not path.endswith(".py"):
+        return False
+    if path in {
+        "api/commands/knowledge_fs.py",
+        "api/configs/extra/knowledge_fs_config.py",
+        "api/extensions/ext_knowledge_fs_observability.py",
+        "api/services/knowledge_fs_capability.py",
+    }:
+        return True
+    if path.startswith(
+        (
+            "api/controllers/console/knowledge_fs/",
+            "api/controllers/service_api/knowledge_fs/",
+            "api/core/tools/builtin_tool/providers/knowledge_fs/",
+            "api/services/knowledge_fs/",
+        )
+    ):
+        return True
+    filename = path.rsplit("/", maxsplit=1)[-1]
+    return (
+        path.startswith("api/models/")
+        and filename.startswith("knowledge_fs")
+        or path.startswith("api/repositories/")
+        and "knowledge_fs" in filename
+        or path.startswith("api/tasks/")
+        and "knowledge_fs" in filename
+    )
+
+
+def discover_core_coverage_paths(workspace_root: Path) -> tuple[str, ...]:
+    """Classify every KnowledgeFS-named production file or fail closed."""
+    named_paths = discover_knowledge_fs_production_paths(workspace_root)
+    core_paths = {path for path in named_paths if is_core_coverage_path(path)}
+    unclassified_paths = set(named_paths) - core_paths - NON_CORE_COVERAGE_ALLOWLIST
+    if unclassified_paths:
+        raise CoverageGateError(
+            "unclassified Dify KnowledgeFS production files must join the core coverage scope or explicit allowlist: "
+            + ", ".join(sorted(unclassified_paths))
+        )
+    if not core_paths:
+        raise CoverageGateError("Dify KnowledgeFS core coverage target set is empty")
+    return tuple(sorted(core_paths))
+
+
+def discover_knowledge_fs_production_paths(workspace_root: Path) -> tuple[str, ...]:
+    """Mirror the workflow's dynamic KnowledgeFS filename discovery."""
+    api_root = workspace_root / "api"
+    if not api_root.is_dir():
+        raise CoverageGateError(f"Dify API directory does not exist: {api_root}")
+    paths: set[str] = set()
+    for directory, child_directories, filenames in os.walk(api_root):
+        current_directory = Path(directory)
+        if current_directory == api_root:
+            child_directories[:] = [name for name in child_directories if name not in {".venv", "storage", "tests"}]
+        child_directories[:] = [name for name in child_directories if name != "__pycache__"]
+        for filename in filenames:
+            path = (current_directory / filename).relative_to(workspace_root).as_posix()
+            if filename.endswith(".py") and "knowledge_fs" in path:
+                paths.add(path)
+    if not paths:
+        raise CoverageGateError("Dify KnowledgeFS production target set is empty")
+    return tuple(sorted(paths))
+
+
+def validate_core_coverage(
+    report: CoverageReport,
+    *,
+    workspace_root: Path,
+    minimum: float,
+) -> CoverageTotals:
+    """Require the exact combined line-and-branch percentage for all core files."""
+    paths = discover_core_coverage_paths(workspace_root)
+    missing_paths = [path for path in paths if path not in report["files"]]
+    if missing_paths:
+        raise CoverageGateError(f"coverage report is missing core files: {', '.join(missing_paths)}")
+
+    covered = 0
+    total = 0
+    for path in paths:
+        summary = report["files"][path]["summary"]
+        covered += summary["covered_lines"] + summary["covered_branches"]
+        total += summary["num_statements"] + summary["num_branches"]
+    if total == 0:
+        raise CoverageGateError("Dify KnowledgeFS core coverage has no measurable statements or branches")
+
+    totals = CoverageTotals(covered=covered, total=total)
+    _require_minimum(totals, minimum=minimum, label="Dify KnowledgeFS core line-and-branch coverage")
+    return totals
+
+
+def resolve_diff_base(workspace_root: Path, preferred: str) -> str:
+    """Resolve an explicit event base, falling back to the previous commit for manual runs."""
+    base = preferred.strip()
+    if not base or set(base) == {"0"}:
+        base = "HEAD^"
+    result = subprocess.run(
+        ["git", "cat-file", "-e", f"{base}^{{commit}}"],
+        cwd=workspace_root,
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    if result.returncode != 0:
+        detail = result.stderr.strip() or "commit is unavailable"
+        raise CoverageGateError(f"cannot resolve coverage diff base {base}: {detail}")
+    return base
+
+
+def collect_changed_glue_lines(
+    workspace_root: Path,
+    base: str,
+    *,
+    glue_paths: tuple[str, ...],
+) -> dict[str, set[int]]:
+    """Return added line numbers in the narrow Dify modules touched by this integration."""
+    result = subprocess.run(
+        [
+            "git",
+            "diff",
+            "--no-ext-diff",
+            "--no-color",
+            "--unified=0",
+            base,
+            "--",
+            *glue_paths,
+        ],
+        cwd=workspace_root,
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    if result.returncode != 0:
+        detail = result.stderr.strip() or "git diff failed"
+        raise CoverageGateError(f"cannot collect KnowledgeFS glue diff from {base}: {detail}")
+    return parse_added_lines(result.stdout)
+
+
+def parse_added_lines(diff: str) -> dict[str, set[int]]:
+    """Parse repository paths and added-side line numbers from a zero-context Git diff."""
+    changed_lines: dict[str, set[int]] = {}
+    current_path: str | None = None
+    current_line: int | None = None
+    for raw_line in diff.splitlines():
+        if raw_line.startswith("diff --git "):
+            current_line = None
+            continue
+        if raw_line.startswith("+++ "):
+            candidate = raw_line[4:]
+            current_path = candidate[2:] if candidate.startswith("b/") else None
+            if current_path is not None:
+                changed_lines.setdefault(current_path, set())
+            current_line = None
+            continue
+        if raw_line.startswith("@@ "):
+            match = HUNK_HEADER.match(raw_line)
+            current_line = int(match.group(1)) if match is not None else None
+            continue
+        if current_path is None or current_line is None:
+            continue
+        if raw_line.startswith("+"):
+            changed_lines[current_path].add(current_line)
+            current_line += 1
+        elif raw_line.startswith("-") or raw_line.startswith("\\"):
+            continue
+        else:
+            current_line += 1
+    return changed_lines
+
+
+def validate_changed_glue_coverage(
+    report: CoverageReport,
+    *,
+    changed_lines: dict[str, set[int]],
+    minimum: float,
+) -> CoverageTotals:
+    """Require added executable glue lines to be exercised by the focused unit suite."""
+    covered = 0
+    total = 0
+    for path, lines in sorted(changed_lines.items()):
+        if not lines:
+            continue
+        file_coverage = report["files"].get(path)
+        if file_coverage is None:
+            raise CoverageGateError(f"coverage report is missing changed glue file: {path}")
+        executed_lines = set(file_coverage["executed_lines"])
+        executable_lines = executed_lines | set(file_coverage["missing_lines"])
+        changed_executable_lines = lines & executable_lines
+        covered += len(changed_executable_lines & executed_lines)
+        total += len(changed_executable_lines)
+
+    totals = CoverageTotals(covered=covered, total=total)
+    _require_minimum(totals, minimum=minimum, label="Dify KnowledgeFS changed-glue line coverage")
+    return totals
+
+
+def _require_minimum(totals: CoverageTotals, *, minimum: float, label: str) -> None:
+    if totals.percent + 1e-12 < minimum:
+        raise CoverageGateError(f"{label} is {totals.percent:.2f}%; minimum {minimum:.2f}%")
+
+
+if __name__ == "__main__":
+    main()

api/dev/generate_knowledge_fs_contract.py

@@ -0,0 +1,732 @@
+"""Pin the in-repository KnowledgeFS contract and validate every Dify product operation.
+
+The lock is intentionally independent of the enclosing Dify commit: it records the staged ``knowledge-fs/`` tree,
+the complete generated OpenAPI document, both explicit product-operation manifests, and the active Capability v2
+profile and deterministic public-key vector. The full OpenAPI hash
+covers request/response schemas, status codes, security, deprecation, and stream metadata. Field-level validation
+cross-checks the Dify product registry, Python Capability issuer, TypeScript request guard, and exported OpenAPI;
+each product operation must be ready or an explicit gap, and KFS-only activation remains explicitly internal.
+
+Contract export reads the working tree only after proving it matches the staged KnowledgeFS index. This keeps the
+OpenAPI bytes and auth manifest aligned with the exact subtree tree ID that will be reviewed and committed.
+"""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import json
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+from typing import Any, Literal, TypedDict, cast
+
+import jwt
+from cryptography.hazmat.primitives.asymmetric.rsa import RSAPublicKey
+from jwt.algorithms import RSAAlgorithm
+
+API_ROOT = Path(__file__).resolve().parents[1]
+if str(API_ROOT) not in sys.path:
+    sys.path.insert(0, str(API_ROOT))
+
+from dev.knowledge_fs_product_contract import (
+    capability_operation_runtime_contracts,
+    parse_capability_operation_policy,
+    parse_product_operation_gap_manifest,
+    parse_product_operation_manifest,
+    product_operation_runtime_contracts,
+    validate_product_operation_contracts,
+)
+
+WORKSPACE_ROOT = API_ROOT.parent
+KNOWLEDGE_FS_DIRECTORY = "knowledge-fs"
+CAPABILITY_V2_AUTH_MANIFEST_RELATIVE_PATH = Path("contracts/dify-capability-v2-auth-profile.json")
+CAPABILITY_V2_AUTH_TEST_VECTOR_RELATIVE_PATH = Path("contracts/dify-capability-v2-test-vector.json")
+UPSTREAM_PROVENANCE_RELATIVE_PATH = Path("upstream-provenance.json")
+LOCK_RELATIVE_PATH = Path("api/knowledge-fs-contract.lock.json")
+PRODUCT_OPERATIONS_RELATIVE_PATH = Path("api/knowledge-fs-product-operations.json")
+PRODUCT_OPERATION_GAPS_RELATIVE_PATH = Path("api/knowledge-fs-product-operation-gaps.json")
+OPENAPI_METHODS = ("delete", "get", "head", "options", "patch", "post", "put", "trace")
+PROXY_METHODS = frozenset({"delete", "get", "patch", "post", "put"})
+LOCK_SCHEMA_VERSION = 5
+
+
+class ContractDeclaration(TypedDict):
+    """KnowledgeFS transport contract declared by one Dify Console registry entry."""
+
+    operation_id: str
+    method: str
+    path: str
+    required_scope: str | None
+    response_kind: str
+    max_response_bytes: int
+    request_headers: tuple[str, ...]
+    response_headers: tuple[str, ...]
+    response_media_types: tuple[str, ...]
+
+
+class ContractLock(TypedDict):
+    """Content-addressed contract inputs that must move together."""
+
+    schemaVersion: int
+    subtreeTree: str
+    openapiSha256: str
+    capabilityV2AuthManifestSha256: str
+    capabilityV2AuthTestVectorSha256: str
+    productOperationManifestSha256: str
+    productOperationGapManifestSha256: str
+
+
+type DeclarationField = Literal[
+    "method",
+    "path",
+    "required_scope",
+    "response_kind",
+    "max_response_bytes",
+    "request_headers",
+    "response_headers",
+    "response_media_types",
+]
+
+DECLARATION_FIELDS: tuple[DeclarationField, ...] = (
+    "method",
+    "path",
+    "required_scope",
+    "response_kind",
+    "max_response_bytes",
+    "request_headers",
+    "response_headers",
+    "response_media_types",
+)
+
+
+def main() -> None:
+    """Update or verify the monorepo pin and validate Dify product declarations."""
+    parser = argparse.ArgumentParser()
+    mode = parser.add_mutually_exclusive_group(required=True)
+    mode.add_argument("--check", action="store_true")
+    mode.add_argument("--update-lock", action="store_true")
+    parser.add_argument("--workspace-root", type=Path, default=WORKSPACE_ROOT)
+    args = parser.parse_args()
+
+    workspace_root = args.workspace_root.resolve()
+    knowledge_fs_root = workspace_root / KNOWLEDGE_FS_DIRECTORY
+    lock_path = workspace_root / LOCK_RELATIVE_PATH
+    capability_v2_auth_manifest_path = knowledge_fs_root / CAPABILITY_V2_AUTH_MANIFEST_RELATIVE_PATH
+    capability_v2_auth_test_vector_path = knowledge_fs_root / CAPABILITY_V2_AUTH_TEST_VECTOR_RELATIVE_PATH
+    product_operations_path = workspace_root / PRODUCT_OPERATIONS_RELATIVE_PATH
+    product_operation_gaps_path = workspace_root / PRODUCT_OPERATION_GAPS_RELATIVE_PATH
+    upstream_provenance_path = knowledge_fs_root / UPSTREAM_PROVENANCE_RELATIVE_PATH
+    ensure_clean_knowledge_fs_worktree(workspace_root)
+    ensure_contract_inputs_exist(
+        knowledge_fs_root=knowledge_fs_root,
+        capability_v2_auth_manifest_path=capability_v2_auth_manifest_path,
+        capability_v2_auth_test_vector_path=capability_v2_auth_test_vector_path,
+        product_operations_path=product_operations_path,
+        product_operation_gaps_path=product_operation_gaps_path,
+        upstream_provenance_path=upstream_provenance_path,
+    )
+
+    subtree_tree = staged_subtree_tree(workspace_root)
+    capability_v2_auth_manifest_content = capability_v2_auth_manifest_path.read_bytes()
+    capability_v2_auth_test_vector_content = capability_v2_auth_test_vector_path.read_bytes()
+    product_operation_manifest_content = product_operations_path.read_bytes()
+    product_operation_gap_manifest_content = product_operation_gaps_path.read_bytes()
+    capability_v2_auth_manifest = load_json_object(capability_v2_auth_manifest_path)
+    validate_capability_v2_auth_manifest(capability_v2_auth_manifest)
+    validate_capability_v2_auth_test_vector(
+        load_json_object(capability_v2_auth_test_vector_path),
+        capability_v2_auth_manifest,
+    )
+    validate_upstream_provenance(load_json_object(upstream_provenance_path))
+    product_manifest = parse_product_operation_manifest(load_json_object(product_operations_path))
+    product_gap_manifest = parse_product_operation_gap_manifest(load_json_object(product_operation_gaps_path))
+    product_runtime_operations = product_operation_runtime_contracts()
+    capability_runtime_operations = capability_operation_runtime_contracts()
+    declarations = console_contract_declarations()
+
+    with tempfile.TemporaryDirectory(prefix="dify-knowledge-fs-contract-") as directory:
+        openapi_path = Path(directory) / "knowledge-fs.openapi.json"
+        capability_policy_path = Path(directory) / "dify-capability-v2-operations.json"
+        subprocess.run(
+            ["pnpm", "openapi:export", "--", "--output", str(openapi_path)],
+            cwd=knowledge_fs_root,
+            check=True,
+        )
+        subprocess.run(
+            ["pnpm", "capability:export", "--", "--output", str(capability_policy_path)],
+            cwd=knowledge_fs_root,
+            check=True,
+        )
+        openapi_content = openapi_path.read_bytes()
+        capability_policy = parse_capability_operation_policy(load_json_object(capability_policy_path))
+
+    document: dict[str, Any] = json.loads(openapi_content)
+    validate_product_operation_contracts(
+        capability_operations=capability_runtime_operations,
+        capability_policy=capability_policy,
+        document=document,
+        gap_manifest=product_gap_manifest,
+        manifest=product_manifest,
+        product_operations=product_runtime_operations,
+    )
+    validate_declarations(document, declarations)
+
+    expected_lock: ContractLock = {
+        "schemaVersion": LOCK_SCHEMA_VERSION,
+        "subtreeTree": subtree_tree,
+        "openapiSha256": sha256(openapi_content),
+        "capabilityV2AuthManifestSha256": sha256(capability_v2_auth_manifest_content),
+        "capabilityV2AuthTestVectorSha256": sha256(capability_v2_auth_test_vector_content),
+        "productOperationManifestSha256": sha256(product_operation_manifest_content),
+        "productOperationGapManifestSha256": sha256(product_operation_gap_manifest_content),
+    }
+
+    if args.update_lock:
+        lock_path.write_text(json.dumps(expected_lock, indent=2) + "\n")
+        return
+
+    received_lock = parse_contract_lock(load_json_object(lock_path))
+    lock_fields = (
+        (
+            "capabilityV2AuthManifestSha256",
+            received_lock["capabilityV2AuthManifestSha256"],
+            expected_lock["capabilityV2AuthManifestSha256"],
+        ),
+        (
+            "capabilityV2AuthTestVectorSha256",
+            received_lock["capabilityV2AuthTestVectorSha256"],
+            expected_lock["capabilityV2AuthTestVectorSha256"],
+        ),
+        (
+            "productOperationManifestSha256",
+            received_lock["productOperationManifestSha256"],
+            expected_lock["productOperationManifestSha256"],
+        ),
+        (
+            "productOperationGapManifestSha256",
+            received_lock["productOperationGapManifestSha256"],
+            expected_lock["productOperationGapManifestSha256"],
+        ),
+        ("schemaVersion", received_lock["schemaVersion"], expected_lock["schemaVersion"]),
+        ("subtreeTree", received_lock["subtreeTree"], expected_lock["subtreeTree"]),
+        ("openapiSha256", received_lock["openapiSha256"], expected_lock["openapiSha256"]),
+    )
+    for field, received_value, expected_value in lock_fields:
+        if received_value != expected_value:
+            raise RuntimeError(
+                f"KnowledgeFS contract lock field {field} drifted: "
+                f"expected {expected_value!r}, received {received_value!r}. "
+                "Run --update-lock intentionally after reviewing the staged subtree and contract changes."
+            )
+
+
+def ensure_contract_inputs_exist(
+    *,
+    knowledge_fs_root: Path,
+    capability_v2_auth_manifest_path: Path,
+    capability_v2_auth_test_vector_path: Path,
+    product_operations_path: Path,
+    product_operation_gaps_path: Path,
+    upstream_provenance_path: Path,
+) -> None:
+    """Fail with a stable error before invoking package tooling when a contract input is absent."""
+    required_paths = (
+        knowledge_fs_root / "package.json",
+        capability_v2_auth_manifest_path,
+        capability_v2_auth_test_vector_path,
+        product_operations_path,
+        product_operation_gaps_path,
+        upstream_provenance_path,
+    )
+    missing_paths = [path for path in required_paths if not path.is_file()]
+    if missing_paths:
+        missing = ", ".join(str(path) for path in missing_paths)
+        raise RuntimeError(f"KnowledgeFS contract input is missing: {missing}")
+
+
+def ensure_clean_knowledge_fs_worktree(workspace_root: Path) -> None:
+    """Require exported KnowledgeFS files to exactly match the staged index tree.
+
+    Staged changes are expected during intentional lock updates. Unstaged tracked changes and untracked files are
+    rejected because the export process reads the working tree while the tree ID is calculated from the index.
+    """
+    subtree_path = f"{KNOWLEDGE_FS_DIRECTORY}/"
+    unstaged = subprocess.run(
+        ["git", "diff", "--quiet", "--", subtree_path],
+        cwd=workspace_root,
+        check=False,
+    )
+    if unstaged.returncode > 1:
+        raise RuntimeError("git diff failed while validating the staged KnowledgeFS subtree")
+    untracked = run(
+        "git",
+        "ls-files",
+        "--others",
+        "--exclude-standard",
+        "--",
+        subtree_path,
+        cwd=workspace_root,
+    ).strip()
+    if unstaged.returncode != 0 or untracked:
+        raise RuntimeError(
+            "knowledge-fs/ contains unstaged or untracked changes; stage or remove them before contract export"
+        )
+
+
+def staged_subtree_tree(workspace_root: Path) -> str:
+    """Return the Git tree object for the staged ``knowledge-fs/`` subtree."""
+    return run("git", "write-tree", f"--prefix={KNOWLEDGE_FS_DIRECTORY}/", cwd=workspace_root).strip()
+
+
+def load_json_object(path: Path) -> dict[str, Any]:
+    """Load a JSON object and reject arrays/scalars at contract boundaries."""
+    value = json.loads(path.read_text())
+    if not isinstance(value, dict):
+        raise ValueError(f"KnowledgeFS contract file must contain a JSON object: {path}")
+    return cast(dict[str, Any], value)
+
+
+def parse_contract_lock(value: dict[str, Any]) -> ContractLock:
+    """Validate the compact, non-self-referential contract lock schema."""
+    expected_fields = {
+        "capabilityV2AuthManifestSha256",
+        "capabilityV2AuthTestVectorSha256",
+        "openapiSha256",
+        "productOperationGapManifestSha256",
+        "productOperationManifestSha256",
+        "schemaVersion",
+        "subtreeTree",
+    }
+    if set(value) != expected_fields:
+        raise ValueError(f"KnowledgeFS contract lock fields must be exactly {sorted(expected_fields)}")
+    if value.get("schemaVersion") != LOCK_SCHEMA_VERSION:
+        raise ValueError(f"KnowledgeFS contract lock schemaVersion must be {LOCK_SCHEMA_VERSION}")
+    for field in (
+        "capabilityV2AuthManifestSha256",
+        "capabilityV2AuthTestVectorSha256",
+        "openapiSha256",
+        "productOperationGapManifestSha256",
+        "productOperationManifestSha256",
+        "subtreeTree",
+    ):
+        field_value = value.get(field)
+        expected_length = 40 if field == "subtreeTree" else 64
+        if (
+            not isinstance(field_value, str)
+            or len(field_value) != expected_length
+            or any(character not in "0123456789abcdef" for character in field_value)
+        ):
+            raise ValueError(f"KnowledgeFS contract lock field {field} has an invalid digest")
+    return cast(ContractLock, value)
+
+
+def validate_required_product_operations(document: dict[str, Any], required_operation_ids: list[str]) -> None:
+    """Fail when the pinned KFS OpenAPI omits an operation required by the Dify product."""
+    available_operation_ids = [
+        operation_id
+        for path_item in document.get("paths", {}).values()
+        for method in OPENAPI_METHODS
+        if isinstance(path_item, dict)
+        for operation in (path_item.get(method),)
+        if isinstance(operation, dict)
+        for operation_id in (operation.get("operationId"),)
+        if isinstance(operation_id, str) and operation_id
+    ]
+    for operation_id in required_operation_ids:
+        count = available_operation_ids.count(operation_id)
+        if count != 1:
+            raise ValueError(
+                f"KnowledgeFS OpenAPI required product operation {operation_id} must occur exactly once; found {count}"
+            )
+
+
+def validate_capability_v2_auth_manifest(value: dict[str, Any]) -> None:
+    """Validate the active production RS256 profile consumed by both Dify and KnowledgeFS."""
+    expected_fields = {
+        "active",
+        "audience",
+        "callerProfiles",
+        "claimBindings",
+        "issuer",
+        "lifecycle",
+        "maxTtlSeconds",
+        "productionReady",
+        "profileId",
+        "protectedHeader",
+        "requiredClaims",
+        "resourceContract",
+        "runtimeAssembly",
+        "schemaVersion",
+        "signatureAlgorithms",
+        "tokenKind",
+    }
+    if set(value) != expected_fields:
+        raise ValueError(f"KnowledgeFS Capability v2 auth manifest fields must be exactly {sorted(expected_fields)}")
+    fixed_values = {
+        "active": True,
+        "audience": "knowledge-fs",
+        "issuer": "dify-control-plane",
+        "lifecycle": "active",
+        "maxTtlSeconds": 60,
+        "productionReady": True,
+        "profileId": "dify-capability-v2",
+        "schemaVersion": 3,
+        "signatureAlgorithms": ["RS256"],
+        "tokenKind": "jwt",
+    }
+    for field, expected in fixed_values.items():
+        if value.get(field) != expected:
+            raise ValueError(f"KnowledgeFS Capability v2 auth manifest field {field} must be {expected!r}")
+    if value.get("protectedHeader") != {
+        "algorithm": "RS256",
+        "keyIdClaim": "kid",
+        "keyIdRequired": True,
+        "type": "JWT",
+    }:
+        raise ValueError("KnowledgeFS Capability v2 protected header contract is invalid")
+    required_claims = [
+        "action",
+        "actor",
+        "aud",
+        "authz_revision",
+        "azp",
+        "caller_kind",
+        "cap_ver",
+        "content_policy_revision",
+        "content_scope_ids",
+        "control_space_id",
+        "exp",
+        "grant_id",
+        "iat",
+        "iss",
+        "jti",
+        "namespace_id",
+        "nbf",
+        "resource",
+        "sub",
+        "trace_id",
+    ]
+    if value.get("requiredClaims") != required_claims:
+        raise ValueError("KnowledgeFS Capability v2 required claims are invalid")
+    if value.get("claimBindings") != {
+        "action": "action",
+        "callerKind": "caller_kind",
+        "controlSpace": "control_space_id",
+        "namespace": "namespace_id",
+        "resource": "resource",
+        "resourceParent": "resource.parent_id",
+        "subject": "sub",
+    }:
+        raise ValueError("KnowledgeFS Capability v2 claim bindings are invalid")
+    if value.get("resourceContract") != {
+        "fields": ["id", "parent_id", "type"],
+        "parentForbiddenFor": ["namespace", "knowledge_space"],
+        "parentRequiredFor": ["document", "job", "query", "research_task", "source", "upload_session"],
+    }:
+        raise ValueError("KnowledgeFS Capability v2 resource contract is invalid")
+    if value.get("callerProfiles") != {
+        "agent": {"authorizedParty": "dify-agent", "subjectPrefix": "dify-app:"},
+        "interactive": {"authorizedParty": "dify-console", "subjectPrefix": "dify-account:"},
+        "internal_worker": {"authorizedParty": "dify-worker", "subjectPrefix": "dify-worker:"},
+        "mcp": {"authorizedParty": "dify-mcp", "subjectPrefix": "dify-mcp-session:"},
+        "service": {"authorizedParty": "dify-service-api", "subjectPrefix": "dify-kfs-credential:"},
+        "workflow": {"authorizedParty": "dify-workflow", "subjectPrefix": "dify-app:"},
+    }:
+        raise ValueError("KnowledgeFS Capability v2 caller profiles are invalid")
+    if value.get("runtimeAssembly") != {
+        "failClosed": True,
+        "keySelection": "kid",
+        "maximumPublishedKeys": 3,
+        "verificationKeySource": "jwks",
+    }:
+        raise ValueError("KnowledgeFS Capability v2 runtime assembly is invalid")
+
+
+def validate_capability_v2_auth_test_vector(
+    value: dict[str, Any],
+    manifest: dict[str, Any],
+) -> None:
+    """Verify the deterministic public-key vector and every security-sensitive binding."""
+    expected_fields = {
+        "algorithm",
+        "audience",
+        "expectedClaims",
+        "expectedPrincipal",
+        "issuer",
+        "operation",
+        "profileId",
+        "protectedHeader",
+        "publicJwk",
+        "schemaVersion",
+        "testOnly",
+        "token",
+        "ttlSeconds",
+    }
+    if set(value) != expected_fields:
+        raise ValueError(f"KnowledgeFS Capability v2 test vector fields must be exactly {sorted(expected_fields)}")
+    if (
+        value.get("schemaVersion") != 2
+        or value.get("profileId") != manifest.get("profileId")
+        or value.get("testOnly") is not True
+        or value.get("algorithm") != "RS256"
+        or value.get("issuer") != manifest.get("issuer")
+        or value.get("audience") != manifest.get("audience")
+        or value.get("ttlSeconds") != manifest.get("maxTtlSeconds")
+    ):
+        raise ValueError("KnowledgeFS Capability v2 test vector does not match the active profile")
+    protected_header = value.get("protectedHeader")
+    if not isinstance(protected_header, dict) or set(protected_header) != {"alg", "kid", "typ"}:
+        raise ValueError("KnowledgeFS Capability v2 test vector protected header is invalid")
+    kid = protected_header.get("kid")
+    if protected_header.get("alg") != "RS256" or protected_header.get("typ") != "JWT" or not _is_non_blank(kid):
+        raise ValueError("KnowledgeFS Capability v2 test vector protected header is invalid")
+    public_jwk = value.get("publicJwk")
+    if not isinstance(public_jwk, dict) or set(public_jwk) != {"alg", "e", "kid", "kty", "n", "use"}:
+        raise ValueError("KnowledgeFS Capability v2 test vector public JWK is invalid")
+    if (
+        public_jwk.get("alg") != "RS256"
+        or public_jwk.get("kid") != kid
+        or public_jwk.get("kty") != "RSA"
+        or public_jwk.get("use") != "sig"
+        or not _is_non_blank(public_jwk.get("e"))
+        or not _is_non_blank(public_jwk.get("n"))
+    ):
+        raise ValueError("KnowledgeFS Capability v2 test vector public JWK is invalid")
+
+    claims = value.get("expectedClaims")
+    required_claims = manifest.get("requiredClaims")
+    if not isinstance(claims, dict) or not isinstance(required_claims, list) or set(claims) != set(required_claims):
+        raise ValueError("KnowledgeFS Capability v2 test vector claims do not match the active profile")
+    operation = value.get("operation")
+    if not isinstance(operation, dict) or set(operation) != {"action", "method", "operationId", "requestPath"}:
+        raise ValueError("KnowledgeFS Capability v2 test vector operation is invalid")
+    resource = claims.get("resource")
+    if not isinstance(resource, dict) or set(resource) != {"id", "parent_id", "type"}:
+        raise ValueError("KnowledgeFS Capability v2 test vector resource is invalid")
+    expected_operation = {
+        "action": "documents.read",
+        "method": "GET",
+        "operationId": "getDocument",
+        "requestPath": "/knowledge-spaces/space-contract-vector/documents/document-contract-vector",
+    }
+    if operation != expected_operation:
+        raise ValueError("KnowledgeFS Capability v2 test vector operation binding is invalid")
+    exact_claims = {
+        "action": operation["action"],
+        "aud": value["audience"],
+        "caller_kind": "interactive",
+        "cap_ver": 2,
+        "control_space_id": "control-space-contract-vector",
+        "iss": value["issuer"],
+        "namespace_id": "workspace-contract-vector",
+        "resource": {
+            "id": "document-contract-vector",
+            "parent_id": "space-contract-vector",
+            "type": "document",
+        },
+        "sub": "dify-account:account-contract-vector",
+    }
+    for field, expected in exact_claims.items():
+        if claims.get(field) != expected:
+            raise ValueError(f"KnowledgeFS Capability v2 test vector claim {field} is invalid")
+    if claims.get("actor") != claims["sub"] or claims.get("azp") != "dify-console":
+        raise ValueError("KnowledgeFS Capability v2 test vector caller binding is invalid")
+    issued_at = claims.get("iat")
+    not_before = claims.get("nbf")
+    expires_at = claims.get("exp")
+    if (
+        not isinstance(issued_at, int)
+        or isinstance(issued_at, bool)
+        or not_before != issued_at
+        or not isinstance(expires_at, int)
+        or isinstance(expires_at, bool)
+        or expires_at - issued_at != value["ttlSeconds"]
+    ):
+        raise ValueError("KnowledgeFS Capability v2 test vector TTL is invalid")
+    expected_principal = {
+        "callerKind": claims["caller_kind"],
+        "subject": {
+            "scopes": ["knowledge-spaces:read"],
+            "subjectId": claims["sub"],
+            "tenantId": claims["namespace_id"],
+        },
+    }
+    if value.get("expectedPrincipal") != expected_principal:
+        raise ValueError("KnowledgeFS Capability v2 test vector principal is invalid")
+
+    token = value.get("token")
+    if not isinstance(token, str) or not _is_non_blank(token):
+        raise ValueError("KnowledgeFS Capability v2 test vector token is invalid")
+    try:
+        verification_key = RSAAlgorithm.from_jwk(public_jwk)
+        if not isinstance(verification_key, RSAPublicKey):
+            raise ValueError("Capability vector verification key is not RSA public material")
+        header = jwt.get_unverified_header(token)
+        decoded_claims = jwt.decode(
+            token,
+            verification_key,
+            algorithms=["RS256"],
+            audience=cast(str, value["audience"]),
+            issuer=cast(str, value["issuer"]),
+            options={"verify_exp": False, "verify_iat": False, "verify_nbf": False},
+        )
+    except (jwt.PyJWTError, TypeError, ValueError) as exc:
+        raise ValueError("KnowledgeFS Capability v2 test vector signature is invalid") from exc
+    if header != protected_header or decoded_claims != claims:
+        raise ValueError("KnowledgeFS Capability v2 test vector token content drifted")
+
+
+def _is_non_blank(value: object) -> bool:
+    return isinstance(value, str) and bool(value.strip()) and value == value.strip()
+
+
+def validate_upstream_provenance(value: dict[str, Any]) -> None:
+    """Validate the imported-source provenance that is itself covered by the subtree tree ID."""
+    expected_fields = {"commit", "release", "repository", "schemaVersion"}
+    if set(value) != expected_fields:
+        raise ValueError(f"KnowledgeFS upstream provenance fields must be exactly {sorted(expected_fields)}")
+    if value.get("schemaVersion") != 1:
+        raise ValueError("KnowledgeFS upstream provenance must use schemaVersion 1")
+    repository = value.get("repository")
+    commit = value.get("commit")
+    if not isinstance(repository, str) or not repository.startswith("https://"):
+        raise ValueError("KnowledgeFS upstream provenance repository must be an HTTPS URL")
+    if (
+        not isinstance(commit, str)
+        or len(commit) != 40
+        or any(character not in "0123456789abcdef" for character in commit)
+    ):
+        raise ValueError("KnowledgeFS upstream provenance commit must be a lowercase full Git SHA")
+    if value.get("release") is not None and not isinstance(value["release"], str):
+        raise ValueError("KnowledgeFS upstream provenance release must be null or a string")
+
+
+def validate_declarations(document: dict[str, Any], declarations: tuple[ContractDeclaration, ...]) -> None:
+    """Validate Dify Console declarations against matching pinned OpenAPI operations."""
+    operations_by_id: dict[str, list[tuple[str, str, dict[str, Any], dict[str, Any]]]] = {}
+    for path, path_item in document.get("paths", {}).items():
+        for method in OPENAPI_METHODS:
+            operation = path_item.get(method)
+            if operation is None:
+                continue
+            operation_id = operation.get("operationId")
+            if isinstance(operation_id, str) and operation_id:
+                operations_by_id.setdefault(operation_id, []).append((method, path, path_item, operation))
+
+    declared_ids: set[str] = set()
+    for declaration in declarations:
+        operation_id = declaration["operation_id"]
+        if operation_id in declared_ids:
+            raise ValueError(f"Dify Console registry has duplicate operationId: {operation_id}")
+        declared_ids.add(operation_id)
+
+        matches = operations_by_id.get(operation_id, [])
+        if not matches:
+            raise ValueError(f"KnowledgeFS OpenAPI has no operationId: {operation_id}")
+        if len(matches) > 1:
+            raise ValueError(f"KnowledgeFS OpenAPI has duplicate operationId: {operation_id}")
+
+        method, path, path_item, operation = matches[0]
+        if not path.startswith("/"):
+            raise ValueError(f"KnowledgeFS OpenAPI path must be absolute: {path}")
+        if method not in PROXY_METHODS:
+            raise ValueError(f"KnowledgeFS proxy does not support {method.upper()} {path}")
+        expected: ContractDeclaration = {
+            "operation_id": operation_id,
+            "method": method.upper(),
+            "path": path[1:],
+            "required_scope": required_scope(document, operation),
+            "response_kind": response_kind(operation),
+            "max_response_bytes": required_max_response_bytes(operation),
+            "request_headers": request_header_names(path_item, operation),
+            "response_headers": response_header_names(operation),
+            "response_media_types": response_media_types(operation),
+        }
+        for field in DECLARATION_FIELDS:
+            expected_value = expected[field]
+            received_value = declaration[field]
+            if received_value != expected_value:
+                raise ValueError(
+                    f"KnowledgeFS operation {operation_id} field {field} drifted: "
+                    f"expected {expected_value!r}, received {received_value!r}"
+                )
+
+
+def console_contract_declarations() -> tuple[ContractDeclaration, ...]:
+    """The P9 backend exposes only typed product controllers; the raw Console proxy is removed."""
+
+    return ()
+
+
+def response_kind(operation: dict[str, Any]) -> str:
+    media_types = response_media_types(operation)
+    if "text/event-stream" in media_types:
+        return "stream"
+    if "application/octet-stream" in media_types:
+        return "binary"
+    return "buffered"
+
+
+def response_media_types(operation: dict[str, Any]) -> tuple[str, ...]:
+    media_types: set[str] = set()
+    for status, response in operation.get("responses", {}).items():
+        if status == "2XX" or (len(status) == 3 and status.startswith("2") and status.isdigit()):
+            media_types.update(response.get("content", {}))
+    return tuple(sorted(media_types))
+
+
+def required_scope(document: dict[str, Any], operation: dict[str, Any]) -> str | None:
+    scope = operation.get("x-knowledge-fs-required-scope")
+    security = operation["security"] if "security" in operation else document.get("security")
+    if security == []:
+        if scope is not None:
+            raise ValueError(f"KnowledgeFS public operation must not declare a required scope: {scope}")
+        return None
+    if security != [{"bearerAuth": []}]:
+        raise ValueError(f"KnowledgeFS operation effective security must be exactly bearerAuth: {security}")
+    if scope in ("knowledge-spaces:read", "knowledge-spaces:write"):
+        return scope
+    raise ValueError(f"KnowledgeFS operation has no supported required scope: {scope}")
+
+
+def required_max_response_bytes(operation: dict[str, Any]) -> int:
+    value = operation.get("x-knowledge-fs-max-response-bytes")
+    if not isinstance(value, int) or isinstance(value, bool) or value <= 0:
+        raise ValueError(f"KnowledgeFS operation has no valid response byte limit: {value}")
+    return value
+
+
+def request_header_names(path_item: dict[str, Any], operation: dict[str, Any]) -> tuple[str, ...]:
+    names: set[str] = set()
+    for parameter in [*path_item.get("parameters", []), *operation.get("parameters", [])]:
+        if "$ref" in parameter:
+            raise ValueError(f"KnowledgeFS request header references are not supported: {parameter['$ref']}")
+        if parameter.get("in") == "header":
+            names.add(parameter["name"].lower())
+    return tuple(sorted(names))
+
+
+def response_header_names(operation: dict[str, Any]) -> tuple[str, ...]:
+    return tuple(
+        sorted(
+            {
+                name.lower()
+                for response in operation.get("responses", {}).values()
+                for name in response.get("headers", {})
+            }
+        )
+    )
+
+
+def sha256(content: bytes) -> str:
+    return hashlib.sha256(content).hexdigest()
+
+
+def run(*command: str, cwd: Path) -> str:
+    return subprocess.run(command, cwd=cwd, check=True, capture_output=True, text=True).stdout
+
+
+if __name__ == "__main__":
+    main()

api/dev/knowledge_fs_product_contract.py

@@ -0,0 +1,494 @@
+"""Cross-service KnowledgeFS product-operation contract validation.
+
+The checked product manifest is the reviewable boundary, while the Dify product registry, Dify
+Capability issuer registry, exported KnowledgeFS OpenAPI, and exported TypeScript request-guard
+policy remain executable sources of truth. A product operation must appear exactly once as ready
+or as an explicit gap; internal KFS-only operations require a named exclusion.
+"""
+
+from __future__ import annotations
+
+from collections.abc import Iterable, Mapping
+from typing import Any, Literal, NamedTuple, TypedDict, cast
+
+
+class ProductOperationLimits(TypedDict):
+    kfsMaxResponseBytes: int
+    productMaxRequestBytes: int
+    productMaxResponseBytes: int
+
+
+class ProductOperationStream(TypedDict):
+    kfsResponseKind: str
+    productKind: str
+
+
+class ProductOperationManifestEntry(TypedDict):
+    action: str
+    kfsOperationId: str
+    limits: ProductOperationLimits
+    method: str
+    path: str
+    productOperationId: str
+    resource: str
+    stream: ProductOperationStream
+    transport: str
+
+
+type ProductOperationManifestField = Literal[
+    "action",
+    "kfsOperationId",
+    "limits",
+    "method",
+    "path",
+    "productOperationId",
+    "resource",
+    "stream",
+    "transport",
+]
+
+PRODUCT_OPERATION_MANIFEST_FIELDS: tuple[ProductOperationManifestField, ...] = (
+    "action",
+    "kfsOperationId",
+    "limits",
+    "method",
+    "path",
+    "productOperationId",
+    "resource",
+    "stream",
+    "transport",
+)
+
+
+class ProductOperationManifest(TypedDict):
+    operations: list[ProductOperationManifestEntry]
+    schemaVersion: int
+
+
+class ProductOperationGapEntry(ProductOperationManifestEntry):
+    reason: str
+    reasonCode: str
+    replacementProductOperationIds: list[str]
+
+
+class InternalKfsOperationExclusion(TypedDict):
+    kfsOperationId: str
+    reason: str
+    reasonCode: str
+
+
+class ProductOperationGapManifest(TypedDict):
+    gaps: list[ProductOperationGapEntry]
+    internalKfsOperationExclusions: list[InternalKfsOperationExclusion]
+    schemaVersion: int
+
+
+class ProductOperationRuntimeContract(NamedTuple):
+    action: str
+    kfs_operation_id: str
+    max_request_bytes: int
+    max_response_bytes: int
+    method: str
+    path: str
+    product_operation_id: str
+    ready: bool
+    resource: str
+    stream_kind: str
+    transport: str
+
+
+class CapabilityOperationRuntimeContract(NamedTuple):
+    action: str
+    allowed_caller_kinds: tuple[str, ...]
+    method: str
+    operation_id: str
+    path: str
+    resource: str
+
+
+class CapabilityOperationPolicy(NamedTuple):
+    action: str
+    allowed_caller_kinds: tuple[str, ...]
+    method: str
+    operation_id: str
+    parent_resource_binding: dict[str, object] | None
+    path: str
+    resource: str
+    resource_binding: dict[str, object]
+
+
+def product_operation_runtime_contracts() -> tuple[ProductOperationRuntimeContract, ...]:
+    """Read the complete typed Dify product registry without normalizing away gaps."""
+    from services.knowledge_fs.product_operations import (
+        KNOWLEDGE_FS_PRODUCT_OPERATIONS,
+        is_product_operation_ready,
+    )
+
+    contracts: list[ProductOperationRuntimeContract] = []
+    for product_operation_id, operation in KNOWLEDGE_FS_PRODUCT_OPERATIONS.items():
+        if operation.capability_operation_id is None or operation.kfs_path is None or operation.action is None:
+            raise ValueError(f"Dify product operation {product_operation_id} has no Capability contract")
+        contracts.append(
+            ProductOperationRuntimeContract(
+                action=operation.action,
+                kfs_operation_id=operation.capability_operation_id,
+                max_request_bytes=operation.max_request_bytes,
+                max_response_bytes=operation.max_response_bytes,
+                method=operation.method,
+                path=operation.kfs_path,
+                product_operation_id=product_operation_id,
+                ready=is_product_operation_ready(product_operation_id),
+                resource=operation.resource_resolver,
+                stream_kind=operation.stream_kind,
+                transport=operation.transport,
+            )
+        )
+    return tuple(contracts)
+
+
+def capability_operation_runtime_contracts() -> tuple[CapabilityOperationRuntimeContract, ...]:
+    """Read the Python issuer policy used to mint each operation Capability."""
+    from services.knowledge_fs_capability import KNOWLEDGE_FS_CAPABILITY_OPERATIONS
+
+    return tuple(
+        CapabilityOperationRuntimeContract(
+            action=operation.action,
+            allowed_caller_kinds=operation.allowed_caller_kinds,
+            method=operation.method,
+            operation_id=operation_id,
+            path=operation.path,
+            resource=operation.resource_type,
+        )
+        for operation_id, operation in KNOWLEDGE_FS_CAPABILITY_OPERATIONS.items()
+    )
+
+
+def parse_product_operation_manifest(value: dict[str, Any]) -> ProductOperationManifest:
+    """Parse the ready product-operation inventory with a closed schema."""
+    if set(value) != {"operations", "schemaVersion"} or value.get("schemaVersion") != 2:
+        raise ValueError("KnowledgeFS product operation manifest must use schemaVersion 2")
+    operations = value.get("operations")
+    if not isinstance(operations, list) or not operations:
+        raise ValueError("KnowledgeFS product operation manifest operations must be a non-empty list")
+    return {
+        "operations": [_parse_product_entry(entry, gap=False) for entry in operations],
+        "schemaVersion": 2,
+    }
+
+
+def parse_product_operation_gap_manifest(value: dict[str, Any]) -> ProductOperationGapManifest:
+    """Parse explicit product gaps and KFS-only operation exclusions with a closed schema."""
+    expected_fields = {"gaps", "internalKfsOperationExclusions", "schemaVersion"}
+    if set(value) != expected_fields or value.get("schemaVersion") != 2:
+        raise ValueError("KnowledgeFS product operation gap manifest must use schemaVersion 2")
+    gaps = value.get("gaps")
+    exclusions = value.get("internalKfsOperationExclusions")
+    if not isinstance(gaps, list) or not gaps or not isinstance(exclusions, list):
+        raise ValueError("KnowledgeFS product operation gap manifest lists are invalid")
+    parsed_exclusions: list[InternalKfsOperationExclusion] = []
+    for raw in exclusions:
+        if not isinstance(raw, dict) or set(raw) != {"kfsOperationId", "reason", "reasonCode"}:
+            raise ValueError("KnowledgeFS internal KFS operation exclusion fields are invalid")
+        for field in ("kfsOperationId", "reason", "reasonCode"):
+            _required_string(raw.get(field), f"internal exclusion {field}")
+        parsed_exclusions.append(cast(InternalKfsOperationExclusion, raw))
+    return {
+        "gaps": [cast(ProductOperationGapEntry, _parse_product_entry(entry, gap=True)) for entry in gaps],
+        "internalKfsOperationExclusions": parsed_exclusions,
+        "schemaVersion": 2,
+    }
+
+
+def parse_capability_operation_policy(value: dict[str, Any]) -> tuple[CapabilityOperationPolicy, ...]:
+    """Parse the runtime-exported TypeScript request-guard registry."""
+    if set(value) != {"operations", "schemaVersion"} or value.get("schemaVersion") != 1:
+        raise ValueError("KnowledgeFS Capability operation policy must use schemaVersion 1")
+    operations = value.get("operations")
+    if not isinstance(operations, list) or not operations:
+        raise ValueError("KnowledgeFS Capability operation policy must contain operations")
+    parsed: list[CapabilityOperationPolicy] = []
+    for raw in operations:
+        expected_fields = {
+            "action",
+            "allowedCallerKinds",
+            "method",
+            "operationId",
+            "parentResourceBinding",
+            "path",
+            "resourceBinding",
+            "resourceType",
+        }
+        if not isinstance(raw, dict) or set(raw) != expected_fields:
+            raise ValueError("KnowledgeFS Capability operation policy fields are invalid")
+        callers = raw.get("allowedCallerKinds")
+        resource_binding = raw.get("resourceBinding")
+        parent_binding = raw.get("parentResourceBinding")
+        if (
+            not isinstance(callers, list)
+            or not callers
+            or not all(isinstance(caller, str) and caller for caller in callers)
+            or not isinstance(resource_binding, dict)
+            or not resource_binding
+            or (parent_binding is not None and not isinstance(parent_binding, dict))
+        ):
+            raise ValueError("KnowledgeFS Capability operation request binding is invalid")
+        for field in ("action", "method", "operationId", "path", "resourceType"):
+            _required_string(raw.get(field), f"Capability operation {field}")
+        parsed.append(
+            CapabilityOperationPolicy(
+                action=cast(str, raw["action"]),
+                allowed_caller_kinds=tuple(callers),
+                method=cast(str, raw["method"]),
+                operation_id=cast(str, raw["operationId"]),
+                parent_resource_binding=cast(dict[str, object] | None, parent_binding),
+                path=cast(str, raw["path"]),
+                resource=cast(str, raw["resourceType"]),
+                resource_binding=cast(dict[str, object], resource_binding),
+            )
+        )
+    _unique((operation.operation_id for operation in parsed), "Capability operation policy")
+    return tuple(parsed)
+
+
+def validate_product_operation_contracts(
+    *,
+    capability_operations: tuple[CapabilityOperationRuntimeContract, ...],
+    capability_policy: tuple[CapabilityOperationPolicy, ...],
+    document: dict[str, Any],
+    gap_manifest: ProductOperationGapManifest,
+    manifest: ProductOperationManifest,
+    product_operations: tuple[ProductOperationRuntimeContract, ...],
+) -> None:
+    """Require exact agreement across manifests, Dify registries, KFS guard policy, and OpenAPI."""
+    ready_entries = {entry["productOperationId"]: entry for entry in manifest["operations"]}
+    gap_entries = {entry["productOperationId"]: entry for entry in gap_manifest["gaps"]}
+    if len(ready_entries) != len(manifest["operations"]):
+        raise ValueError("KnowledgeFS product operation manifest contains duplicate product ids")
+    if len(gap_entries) != len(gap_manifest["gaps"]):
+        raise ValueError("KnowledgeFS product operation gap manifest contains duplicate product ids")
+    if set(ready_entries) & set(gap_entries):
+        raise ValueError("KnowledgeFS product operation cannot be both ready and a gap")
+    runtime_by_product = {operation.product_operation_id: operation for operation in product_operations}
+    if set(ready_entries) | set(gap_entries) != set(runtime_by_product):
+        missing = sorted(set(runtime_by_product) - set(ready_entries) - set(gap_entries))
+        extra = sorted((set(ready_entries) | set(gap_entries)) - set(runtime_by_product))
+        raise ValueError(f"KnowledgeFS product operation completeness drifted: missing={missing}, extra={extra}")
+
+    python_capabilities = {operation.operation_id: operation for operation in capability_operations}
+    guard_capabilities = {operation.operation_id: operation for operation in capability_policy}
+    openapi_operations = _openapi_operations(document)
+    for product_operation_id, runtime in runtime_by_product.items():
+        entry = ready_entries.get(product_operation_id) or gap_entries[product_operation_id]
+        if runtime.ready != (product_operation_id in ready_entries):
+            state = "ready" if runtime.ready else "gap"
+            raise ValueError(f"Dify product operation {product_operation_id} must be declared as {state}")
+        _validate_product_entry(entry, runtime, openapi_operations, python_capabilities, guard_capabilities)
+        if product_operation_id in gap_entries:
+            _required_string(gap_entries[product_operation_id]["reason"], "product gap reason")
+            _required_string(gap_entries[product_operation_id]["reasonCode"], "product gap reasonCode")
+            replacements = gap_entries[product_operation_id]["replacementProductOperationIds"]
+            if not replacements or any(replacement not in ready_entries for replacement in replacements):
+                raise ValueError(f"KnowledgeFS product gap {product_operation_id} has invalid replacements")
+
+    exclusions = gap_manifest["internalKfsOperationExclusions"]
+    exclusion_by_id = {entry["kfsOperationId"]: entry for entry in exclusions}
+    if len(exclusion_by_id) != len(exclusions):
+        raise ValueError("KnowledgeFS internal KFS operation exclusions contain duplicate ids")
+    lifecycle_operation_ids = {
+        "activateDifyWorkspaceIntegration",
+        "freezeDifyWorkspaceIntegration",
+    }
+    if set(exclusion_by_id) != lifecycle_operation_ids:
+        raise ValueError("KnowledgeFS Dify integration lifecycle operations must be explicit internal exclusions")
+    product_capability_ids = {operation.kfs_operation_id for operation in product_operations}
+    for operation_id in exclusion_by_id:
+        if operation_id in product_capability_ids:
+            raise ValueError(f"Internal KFS operation {operation_id} cannot also be a product operation")
+        _validate_capability_alignment(
+            operation_id,
+            openapi_operations,
+            python_capabilities,
+            guard_capabilities,
+        )
+        python_operation = python_capabilities[operation_id]
+        if python_operation.allowed_caller_kinds != ("internal_worker",) or python_operation.resource != "namespace":
+            raise ValueError("Dify integration lifecycle operations must remain internal-worker namespace-only")
+
+
+def _parse_product_entry(raw: object, *, gap: bool) -> ProductOperationManifestEntry:
+    base_fields = {
+        "action",
+        "kfsOperationId",
+        "limits",
+        "method",
+        "path",
+        "productOperationId",
+        "resource",
+        "stream",
+        "transport",
+    }
+    expected_fields = base_fields | ({"reason", "reasonCode", "replacementProductOperationIds"} if gap else set())
+    if not isinstance(raw, dict) or set(raw) != expected_fields:
+        raise ValueError("KnowledgeFS product operation entry fields are invalid")
+    for field in ("action", "kfsOperationId", "method", "path", "productOperationId", "resource", "transport"):
+        _required_string(raw.get(field), f"product operation {field}")
+    limits = raw.get("limits")
+    stream = raw.get("stream")
+    if not isinstance(limits, dict) or set(limits) != {
+        "kfsMaxResponseBytes",
+        "productMaxRequestBytes",
+        "productMaxResponseBytes",
+    }:
+        raise ValueError("KnowledgeFS product operation limits are invalid")
+    if not all(isinstance(limit, int) and not isinstance(limit, bool) and limit >= 0 for limit in limits.values()):
+        raise ValueError("KnowledgeFS product operation limits must be nonnegative integers")
+    if not isinstance(stream, dict) or set(stream) != {"kfsResponseKind", "productKind"}:
+        raise ValueError("KnowledgeFS product operation stream contract is invalid")
+    _required_string(stream.get("kfsResponseKind"), "KFS response kind")
+    _required_string(stream.get("productKind"), "product stream kind")
+    if gap:
+        _required_string(raw.get("reason"), "product gap reason")
+        _required_string(raw.get("reasonCode"), "product gap reasonCode")
+        replacements = raw.get("replacementProductOperationIds")
+        if not isinstance(replacements, list) or not all(isinstance(item, str) and item for item in replacements):
+            raise ValueError("KnowledgeFS product operation gap replacements are invalid")
+    return cast(ProductOperationManifestEntry, raw)
+
+
+def _validate_product_entry(
+    entry: ProductOperationManifestEntry,
+    runtime: ProductOperationRuntimeContract,
+    openapi_operations: Mapping[str, tuple[str, str, dict[str, Any]]],
+    python_capabilities: Mapping[str, CapabilityOperationRuntimeContract],
+    guard_capabilities: Mapping[str, CapabilityOperationPolicy],
+) -> None:
+    _validate_capability_alignment(
+        runtime.kfs_operation_id,
+        openapi_operations,
+        python_capabilities,
+        guard_capabilities,
+    )
+    _, _, openapi_operation = openapi_operations[runtime.kfs_operation_id]
+    expected: ProductOperationManifestEntry = {
+        "action": runtime.action,
+        "kfsOperationId": runtime.kfs_operation_id,
+        "limits": {
+            "kfsMaxResponseBytes": _openapi_max_response_bytes(openapi_operation),
+            "productMaxRequestBytes": runtime.max_request_bytes,
+            "productMaxResponseBytes": runtime.max_response_bytes,
+        },
+        "method": runtime.method,
+        "path": runtime.path,
+        "productOperationId": runtime.product_operation_id,
+        "resource": runtime.resource,
+        "stream": {
+            "kfsResponseKind": _openapi_response_kind(openapi_operation),
+            "productKind": runtime.stream_kind,
+        },
+        "transport": runtime.transport,
+    }
+    for field in PRODUCT_OPERATION_MANIFEST_FIELDS:
+        expected_value = expected[field]
+        if entry[field] != expected_value:
+            raise ValueError(
+                f"KnowledgeFS product operation {runtime.product_operation_id} field {field} drifted: "
+                f"expected {expected_value!r}, received {entry[field]!r}"
+            )
+
+
+def _validate_capability_alignment(
+    operation_id: str,
+    openapi_operations: Mapping[str, tuple[str, str, dict[str, Any]]],
+    python_capabilities: Mapping[str, CapabilityOperationRuntimeContract],
+    guard_capabilities: Mapping[str, CapabilityOperationPolicy],
+) -> None:
+    python_operation = python_capabilities.get(operation_id)
+    guard_operation = guard_capabilities.get(operation_id)
+    openapi_operation = openapi_operations.get(operation_id)
+    if python_operation is None or guard_operation is None or openapi_operation is None:
+        raise ValueError(
+            f"KnowledgeFS operation {operation_id} must exist in Python issuer, TypeScript guard, and OpenAPI"
+        )
+    method, path, _ = openapi_operation
+    expected = (
+        python_operation.action,
+        python_operation.method,
+        python_operation.path,
+        python_operation.resource,
+    )
+    received_guard = (guard_operation.action, guard_operation.method, guard_operation.path, guard_operation.resource)
+    if received_guard != expected:
+        raise ValueError(f"KnowledgeFS operation {operation_id} TypeScript guard drifted")
+    if (method, path) != (python_operation.method, python_operation.path):
+        raise ValueError(f"KnowledgeFS operation {operation_id} OpenAPI method/path drifted")
+    if guard_operation.allowed_caller_kinds != python_operation.allowed_caller_kinds:
+        raise ValueError(f"KnowledgeFS operation {operation_id} caller policy drifted")
+
+
+def _openapi_operations(document: dict[str, Any]) -> dict[str, tuple[str, str, dict[str, Any]]]:
+    operations: dict[str, tuple[str, str, dict[str, Any]]] = {}
+    duplicates: set[str] = set()
+    for path, path_item in document.get("paths", {}).items():
+        if not isinstance(path, str) or not isinstance(path_item, dict):
+            continue
+        for method in ("delete", "get", "patch", "post", "put"):
+            operation = path_item.get(method)
+            if not isinstance(operation, dict):
+                continue
+            operation_id = operation.get("operationId")
+            if not isinstance(operation_id, str) or not operation_id:
+                continue
+            if operation_id in operations:
+                duplicates.add(operation_id)
+            operations[operation_id] = (method.upper(), path, operation)
+    if duplicates:
+        raise ValueError(f"KnowledgeFS OpenAPI contains duplicate operation ids: {sorted(duplicates)}")
+    return operations
+
+
+def _openapi_response_kind(operation: dict[str, Any]) -> Literal["binary", "buffered", "stream"]:
+    media_types = {
+        media_type
+        for status, response in operation.get("responses", {}).items()
+        if status == "2XX" or (len(status) == 3 and status.startswith("2") and status.isdigit())
+        for media_type in response.get("content", {})
+    }
+    if "text/event-stream" in media_types:
+        return "stream"
+    if "application/octet-stream" in media_types:
+        return "binary"
+    return "buffered"
+
+
+def _openapi_max_response_bytes(operation: dict[str, Any]) -> int:
+    value = operation.get("x-knowledge-fs-max-response-bytes")
+    if not isinstance(value, int) or isinstance(value, bool) or value <= 0:
+        raise ValueError("KnowledgeFS product OpenAPI operation has no positive response byte limit")
+    return value
+
+
+def _required_string(value: object, field: str) -> str:
+    if not isinstance(value, str) or not value.strip() or value != value.strip():
+        raise ValueError(f"KnowledgeFS {field} must be a non-empty trimmed string")
+    return value
+
+
+def _unique(values: Iterable[str], label: str) -> None:
+    sequence = tuple(values)
+    if len(sequence) != len(set(sequence)):
+        raise ValueError(f"KnowledgeFS {label} contains duplicate ids")
+
+
+__all__ = [
+    "CapabilityOperationPolicy",
+    "CapabilityOperationRuntimeContract",
+    "ProductOperationGapManifest",
+    "ProductOperationManifest",
+    "ProductOperationRuntimeContract",
+    "capability_operation_runtime_contracts",
+    "parse_capability_operation_policy",
+    "parse_product_operation_gap_manifest",
+    "parse_product_operation_manifest",
+    "product_operation_runtime_contracts",
+    "validate_product_operation_contracts",
+]

api/extensions/ext_celery.py

@@ -164,6 +164,14 @@ def init_app(app: DifyApp) -> Celery:
 
     # if you add a new task, please add the switch to CeleryScheduleTasksConfig
     beat_schedule: dict[str, CeleryBeatScheduleEntry] = {}
+    from services.knowledge_fs.lifecycle_readiness import get_configured_knowledge_fs_lifecycle_worker_readiness
+
+    if get_configured_knowledge_fs_lifecycle_worker_readiness().ready:
+        imports.append("tasks.knowledge_fs_lifecycle_tasks")
+        beat_schedule["knowledge_fs_lifecycle_worker"] = {
+            "task": "tasks.knowledge_fs_lifecycle_tasks.run_knowledge_fs_lifecycle_worker",
+            "schedule": timedelta(seconds=dify_config.KNOWLEDGE_FS_LIFECYCLE_POLL_INTERVAL_SECONDS),
+        }
     if dify_config.ENABLE_CLEAN_EMBEDDING_CACHE_TASK:
         imports.append("schedule.clean_embedding_cache_task")
         beat_schedule["clean_embedding_cache_task"] = {

api/extensions/ext_commands.py

@@ -27,6 +27,7 @@ def init_app(app: DifyApp):
         import_migration_data,
         install_plugins,
         install_rag_pipeline_plugins,
+        knowledge_fs_control_space,
         migrate_data_for_plugin,
         migrate_dataset_permissions_to_rbac,
         migrate_member_roles_to_rbac,
@@ -88,6 +89,7 @@ def init_app(app: DifyApp):
         export_migration_data_template,
         import_migration_data,
         migration_data_wizard,
+        knowledge_fs_control_space,
     ]
     for cmd in cmds_to_register:
         app.cli.add_command(cmd)

api/extensions/ext_knowledge_fs_observability.py

@@ -0,0 +1,24 @@
+"""Production assembly for KnowledgeFS database-backed metrics."""
+
+from __future__ import annotations
+
+import logging
+
+from flask import Flask
+
+from core.db.session_factory import session_factory
+from repositories.sqlalchemy_knowledge_fs_control_space_state_metrics import (
+    SQLAlchemyKnowledgeFSControlSpaceStateCountReader,
+)
+from services.knowledge_fs.observability import get_knowledge_fs_operational_metrics
+
+logger = logging.getLogger(__name__)
+
+
+def init_app(app: Flask) -> None:
+    del app
+    try:
+        reader = SQLAlchemyKnowledgeFSControlSpaceStateCountReader(session_factory.get_session_maker())
+        get_knowledge_fs_operational_metrics().register_control_space_state_gauge(reader)
+    except Exception:
+        logger.exception("KnowledgeFS operational metric registration failed; continuing application startup")

api/knowledge-fs-contract.lock.json

@@ -0,0 +1,9 @@
+{
+  "schemaVersion": 5,
+  "subtreeTree": "04f9a6189c215f4fcccdccbb3ae7fc810b927463",
+  "openapiSha256": "df702b15769f79c561e34d72472a2ec8dddcdefe4bd4c82858f06a98cd6eae8c",
+  "capabilityV2AuthManifestSha256": "fc0a47e23cce12544882f0298522b4933002e892b84ce1815df7e81d36a7a0c7",
+  "capabilityV2AuthTestVectorSha256": "ae0de37b1ff05c40f905cf17a7b410d8971acacf64db07d5ee3d6fecfa559ce3",
+  "productOperationManifestSha256": "5ceef85419dfba9c803732f78f31c803a0836e883972bb67f26ae43b0c78d341",
+  "productOperationGapManifestSha256": "ccbae37fe658177a77529822211a2cf02e72b5815cbc9b7e7e65e6817ba5e0a9"
+}

api/knowledge-fs-product-operation-gaps.json

@@ -0,0 +1,44 @@
+{
+  "schemaVersion": 2,
+  "gaps": [
+    {
+      "productOperationId": "createDocument",
+      "kfsOperationId": "uploadDocument",
+      "method": "POST",
+      "path": "/knowledge-spaces/{id}/documents",
+      "action": "documents.create",
+      "resource": "knowledge_space",
+      "transport": "multipart",
+      "stream": {
+        "productKind": "buffered-multipart",
+        "kfsResponseKind": "buffered"
+      },
+      "limits": {
+        "productMaxRequestBytes": 0,
+        "productMaxResponseBytes": 0,
+        "kfsMaxResponseBytes": 1048576
+      },
+      "reasonCode": "NON_HOMOMORPHIC_MULTIPART_TRANSPORT",
+      "reason": "Legacy buffered multipart document creation has no homomorphic Dify product transport; use Capability v2 direct-upload operations.",
+      "replacementProductOperationIds": [
+        "createUploadSession",
+        "presignUploadSessionPart",
+        "uploadSmallFile",
+        "completeUploadSession",
+        "abortUploadSession"
+      ]
+    }
+  ],
+  "internalKfsOperationExclusions": [
+    {
+      "kfsOperationId": "activateDifyWorkspaceIntegration",
+      "reasonCode": "INTERNAL_CONTROL_PLANE_ONLY",
+      "reason": "Workspace activation is an internal cutover control-plane operation and is never exposed as a product operation."
+    },
+    {
+      "kfsOperationId": "freezeDifyWorkspaceIntegration",
+      "reasonCode": "INTERNAL_CONTROL_PLANE_ONLY",
+      "reason": "Workspace maintenance freeze is an internal cutover control-plane operation and is never exposed as a product operation."
+    }
+  ]
+}

api/knowledge-fs-product-operations.json

@@ -0,0 +1,53 @@
+{
+  "schemaVersion": 2,
+  "operations": [
+    {"productOperationId":"batchSpaceSummaries","kfsOperationId":"batchKnowledgeSpaceProductSummaries","method":"POST","path":"/internal/knowledge-spaces/product-summaries/batch","action":"knowledge_spaces.status.batch","resource":"namespace","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":65536,"productMaxResponseBytes":1048576,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getSpace","kfsOperationId":"getKnowledgeSpace","method":"GET","path":"/knowledge-spaces/{id}","action":"knowledge_spaces.read","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":262144,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"updateSpace","kfsOperationId":"updateKnowledgeSpace","method":"PATCH","path":"/knowledge-spaces/{id}","action":"knowledge_spaces.update","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":32768,"productMaxResponseBytes":262144,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getSettings","kfsOperationId":"getKnowledgeSpaceProductSettings","method":"GET","path":"/knowledge-spaces/{id}/product-settings","action":"knowledge_spaces.settings.read","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":262144,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"updateSettings","kfsOperationId":"updateKnowledgeSpaceProductSettings","method":"PATCH","path":"/knowledge-spaces/{id}/product-settings","action":"knowledge_spaces.settings.update","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":65536,"productMaxResponseBytes":262144,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listDocuments","kfsOperationId":"listDocuments","method":"GET","path":"/knowledge-spaces/{id}/documents","action":"documents.list","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":2097152,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getDocument","kfsOperationId":"getDocument","method":"GET","path":"/knowledge-spaces/{id}/documents/{documentId}","action":"documents.read","resource":"document","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getDocumentOutline","kfsOperationId":"getDocumentOutline","method":"GET","path":"/knowledge-spaces/{id}/documents/{documentId}/outline","action":"documents.outline.read","resource":"document","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listDocumentRevisions","kfsOperationId":"listDocumentRevisions","method":"GET","path":"/knowledge-spaces/{id}/documents/{documentId}/revisions","action":"documents.revisions.list","resource":"document","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":2097152,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"updateDocumentMetadata","kfsOperationId":"patchDocumentMetadata","method":"PATCH","path":"/knowledge-spaces/{id}/documents/{documentId}/metadata","action":"documents.metadata.update","resource":"document","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":131072,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listDocumentChunks","kfsOperationId":"listDocumentChunks","method":"GET","path":"/knowledge-spaces/{id}/documents/{documentId}/revisions/{revision}/chunks","action":"documents.chunks.list","resource":"document","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getDocumentChunk","kfsOperationId":"getDocumentChunk","method":"GET","path":"/knowledge-spaces/{id}/documents/{documentId}/revisions/{revision}/chunks/{chunkId}","action":"documents.chunks.read","resource":"document","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":1048576,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"deleteDocument","kfsOperationId":"requestDocumentDeletion","method":"DELETE","path":"/knowledge-spaces/{id}/documents/{documentId}","action":"documents.delete","resource":"document","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":32768,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"bulkDeleteDocuments","kfsOperationId":"requestBulkDocumentDeletion","method":"DELETE","path":"/knowledge-spaces/{id}/documents/bulk","action":"documents.bulk.delete","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":1048576,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"reindexDocuments","kfsOperationId":"bulkReindexDocuments","method":"POST","path":"/knowledge-spaces/{id}/documents/bulk/reindex","action":"documents.bulk.reindex","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":1048576,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getCompilationJob","kfsOperationId":"getDocumentCompilationJob","method":"GET","path":"/jobs/{id}","action":"document_jobs.read","resource":"job","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"cancelCompilationJob","kfsOperationId":"cancelDocumentCompilationJob","method":"DELETE","path":"/jobs/{id}","action":"document_jobs.cancel","resource":"job","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"retryCompilationJob","kfsOperationId":"retryDocumentCompilationJob","method":"POST","path":"/jobs/{id}/retry","action":"document_jobs.retry","resource":"job","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getBulkJob","kfsOperationId":"getBulkOperation","method":"GET","path":"/bulk-jobs/{id}","action":"bulk_jobs.read","resource":"job","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":2097152,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listSources","kfsOperationId":"listKnowledgeSpaceSources","method":"GET","path":"/knowledge-spaces/{id}/sources","action":"sources.list","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":2097152,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"createSource","kfsOperationId":"createKnowledgeSpaceSource","method":"POST","path":"/knowledge-spaces/{id}/sources","action":"sources.create","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":262144,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getSource","kfsOperationId":"getKnowledgeSpaceSource","method":"GET","path":"/knowledge-spaces/{id}/sources/{sourceId}","action":"sources.read","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"updateSource","kfsOperationId":"updateKnowledgeSpaceSource","method":"PATCH","path":"/knowledge-spaces/{id}/sources/{sourceId}","action":"sources.update","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":262144,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"deleteSource","kfsOperationId":"requestSourceDeletion","method":"DELETE","path":"/knowledge-spaces/{id}/sources/{sourceId}","action":"sources.delete","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":32768,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"testSource","kfsOperationId":"testKnowledgeSpaceSource","method":"POST","path":"/knowledge-spaces/{id}/sources/{sourceId}/test","action":"sources.test","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":262144,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"crawlSource","kfsOperationId":"crawlKnowledgeSpaceSource","method":"POST","path":"/knowledge-spaces/{id}/sources/{sourceId}/crawl","action":"sources.crawl","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":8388608,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listSourcePages","kfsOperationId":"listKnowledgeSpaceSourcePages","method":"GET","path":"/knowledge-spaces/{id}/sources/{sourceId}/pages","action":"sources.pages.list","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"importSourcePages","kfsOperationId":"importKnowledgeSpaceSourcePages","method":"POST","path":"/knowledge-spaces/{id}/sources/{sourceId}/import","action":"sources.pages.import","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":1048576,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listSourceFiles","kfsOperationId":"listKnowledgeSpaceSourceFiles","method":"GET","path":"/knowledge-spaces/{id}/sources/{sourceId}/files","action":"sources.files.list","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":32768,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"importSourceFiles","kfsOperationId":"importKnowledgeSpaceSourceFiles","method":"POST","path":"/knowledge-spaces/{id}/sources/{sourceId}/import-files","action":"sources.files.import","resource":"source","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":1048576,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"createQuery","kfsOperationId":"createQuery","method":"POST","path":"/queries","action":"queries.create","resource":"knowledge_space","transport":"direct","stream":{"productKind":"sse","kfsResponseKind":"stream"},"limits":{"productMaxRequestBytes":65536,"productMaxResponseBytes":0,"kfsMaxResponseBytes":67108864}},
+    {"productOperationId":"listResearchTasks","kfsOperationId":"listKnowledgeSpaceResearchTasks","method":"GET","path":"/knowledge-spaces/{id}/research-tasks","action":"research_tasks.list","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":2097152,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"createResearchTask","kfsOperationId":"createResearchTask","method":"POST","path":"/research-tasks","action":"research_tasks.create","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":65536,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"planResearchTask","kfsOperationId":"planResearchTask","method":"POST","path":"/research-tasks/plan","action":"research_tasks.plan","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":65536,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getResearchTask","kfsOperationId":"getResearchTask","method":"GET","path":"/research-tasks/{id}","action":"research_tasks.read","resource":"research_task","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listResearchTaskPartials","kfsOperationId":"listResearchTaskPartials","method":"GET","path":"/research-tasks/{id}/partials","action":"research_tasks.partials.list","resource":"research_task","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":8388608,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"cancelResearchTask","kfsOperationId":"cancelResearchTask","method":"DELETE","path":"/research-tasks/{id}","action":"research_tasks.cancel","resource":"research_task","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":524288,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listTraces","kfsOperationId":"listKnowledgeSpaceQualityTraces","method":"GET","path":"/knowledge-spaces/{id}/quality/traces","action":"quality.traces.list","resource":"knowledge_space","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"getTrace","kfsOperationId":"getAnswerTrace","method":"GET","path":"/queries/{traceId}","action":"queries.read","resource":"query","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":2097152,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listTraceEvidence","kfsOperationId":"listQueryEvidence","method":"GET","path":"/queries/{traceId}/evidence","action":"queries.evidence.list","resource":"query","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listTraceConflicts","kfsOperationId":"listQueryConflicts","method":"GET","path":"/queries/{traceId}/conflicts","action":"queries.conflicts.list","resource":"query","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"listTraceMissing","kfsOperationId":"listQueryMissing","method":"GET","path":"/queries/{traceId}/missing","action":"queries.missing.list","resource":"query","transport":"json","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":16384,"productMaxResponseBytes":4194304,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"createUploadSession","kfsOperationId":"createUploadSession","method":"POST","path":"/knowledge-spaces/{id}/upload-sessions","action":"upload_sessions.create","resource":"knowledge_space","transport":"direct","stream":{"productKind":"direct-upload","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":65536,"productMaxResponseBytes":65536,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"presignUploadSessionPart","kfsOperationId":"presignUploadSessionPart","method":"POST","path":"/upload-sessions/{id}/parts/{partNumber}/presign","action":"upload_sessions.write","resource":"upload_session","transport":"direct","stream":{"productKind":"direct-upload","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":32768,"productMaxResponseBytes":65536,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"uploadSmallFile","kfsOperationId":"uploadSmallFile","method":"POST","path":"/upload-sessions/{id}/small-file","action":"upload_sessions.write","resource":"upload_session","transport":"binary","stream":{"productKind":"json","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":8388608,"productMaxResponseBytes":131072,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"completeUploadSession","kfsOperationId":"completeUploadSession","method":"POST","path":"/upload-sessions/{id}/complete","action":"upload_sessions.complete","resource":"upload_session","transport":"direct","stream":{"productKind":"direct-upload","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":131072,"productMaxResponseBytes":131072,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"abortUploadSession","kfsOperationId":"abortUploadSession","method":"POST","path":"/upload-sessions/{id}/abort","action":"upload_sessions.abort","resource":"upload_session","transport":"direct","stream":{"productKind":"direct-upload","kfsResponseKind":"buffered"},"limits":{"productMaxRequestBytes":32768,"productMaxResponseBytes":65536,"kfsMaxResponseBytes":1048576}},
+    {"productOperationId":"streamResearchTask","kfsOperationId":"streamResearchTaskProgress","method":"GET","path":"/research-tasks/{id}/events","action":"research_tasks.stream","resource":"research_task","transport":"direct","stream":{"productKind":"sse","kfsResponseKind":"stream"},"limits":{"productMaxRequestBytes":0,"productMaxResponseBytes":0,"kfsMaxResponseBytes":67108864}}
+  ]
+}

api/libs/archive_storage.py

@@ -16,12 +16,19 @@ from urllib.parse import quote
 import boto3
 import orjson
 from botocore.client import Config
-from botocore.exceptions import ClientError
+from botocore.exceptions import BotoCoreError, ClientError
 
 from configs import dify_config
 
 logger = logging.getLogger(__name__)
 
+_OBJECT_NOT_FOUND_ERROR_CODES = frozenset({"404", "NoSuchKey", "NotFound"})
+
+
+def _is_object_not_found_error(error: ClientError) -> bool:
+    error_code = str(error.response.get("Error", {}).get("Code", ""))
+    return error_code in _OBJECT_NOT_FOUND_ERROR_CODES
+
 
 class ArchiveStorageError(Exception):
     """Base exception for archive storage operations."""
@@ -138,10 +145,11 @@ class ArchiveStorage:
             response = self.client.get_object(Bucket=self.bucket, Key=key)
             return response["Body"].read()
         except ClientError as e:
-            error_code = e.response.get("Error", {}).get("Code")
-            if error_code == "NoSuchKey":
-                raise FileNotFoundError(f"Archive object not found: {key}")
-            raise ArchiveStorageError(f"Failed to download object '{key}': {e}")
+            if _is_object_not_found_error(e):
+                raise FileNotFoundError(f"Archive object not found: {key}") from e
+            raise ArchiveStorageError(f"Failed to download object '{key}': {e}") from e
+        except BotoCoreError as e:
+            raise ArchiveStorageError(f"Failed to download object '{key}': {e}") from e
 
     def get_object_stream(self, key: str) -> Generator[bytes, None, None]:
         """
@@ -161,10 +169,11 @@ class ArchiveStorage:
             response = self.client.get_object(Bucket=self.bucket, Key=key)
             yield from response["Body"].iter_chunks()
         except ClientError as e:
-            error_code = e.response.get("Error", {}).get("Code")
-            if error_code == "NoSuchKey":
-                raise FileNotFoundError(f"Archive object not found: {key}")
-            raise ArchiveStorageError(f"Failed to stream object '{key}': {e}")
+            if _is_object_not_found_error(e):
+                raise FileNotFoundError(f"Archive object not found: {key}") from e
+            raise ArchiveStorageError(f"Failed to stream object '{key}': {e}") from e
+        except BotoCoreError as e:
+            raise ArchiveStorageError(f"Failed to stream object '{key}': {e}") from e
 
     def object_exists(self, key: str) -> bool:
         """
@@ -175,12 +184,19 @@ class ArchiveStorage:
 
         Returns:
             True if object exists, False otherwise
+
+        Raises:
+            ArchiveStorageError: If storage cannot authoritatively determine object existence
         """
         try:
             self.client.head_object(Bucket=self.bucket, Key=key)
             return True
-        except ClientError:
-            return False
+        except ClientError as e:
+            if _is_object_not_found_error(e):
+                return False
+            raise ArchiveStorageError(f"Failed to check archive object '{key}': {e}") from e
+        except BotoCoreError as e:
+            raise ArchiveStorageError(f"Failed to check archive object '{key}': {e}") from e
 
     def delete_object(self, key: str) -> None:
         """
@@ -196,7 +212,12 @@ class ArchiveStorage:
             self.client.delete_object(Bucket=self.bucket, Key=key)
             logger.debug("Deleted object: %s", key)
         except ClientError as e:
-            raise ArchiveStorageError(f"Failed to delete object '{key}': {e}")
+            if _is_object_not_found_error(e):
+                logger.debug("Archive object was already absent: %s", key)
+                return
+            raise ArchiveStorageError(f"Failed to delete object '{key}': {e}") from e
+        except BotoCoreError as e:
+            raise ArchiveStorageError(f"Failed to delete object '{key}': {e}") from e
 
     def generate_presigned_url(
         self,

api/migrations/versions/2026_07_15_1600-3c9f8e2a1d7b_add_workflow_run_archive_bundle_cursor_indexes.py

@@ -0,0 +1,45 @@
+"""add workflow-run archive bundle cursor indexes
+
+Revision ID: 3c9f8e2a1d7b
+Revises: 7a1c2d9e4b60
+Create Date: 2026-07-15 16:00:00.000000
+
+"""
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "3c9f8e2a1d7b"
+down_revision = "7a1c2d9e4b60"
+branch_labels = None
+depends_on = None
+
+_TABLE_NAME = "workflow_run_archive_bundles"
+_INDEXES = (
+    ("workflow_run_archive_bundle_month_id_idx", ("year", "month", "id")),
+    ("workflow_run_archive_bundle_month_shard_id_idx", ("year", "month", "shard", "id")),
+)
+
+
+def _is_postgresql() -> bool:
+    return op.get_bind().dialect.name == "postgresql"
+
+
+def upgrade() -> None:
+    if _is_postgresql():
+        with op.get_context().autocommit_block():
+            for index_name, columns in _INDEXES:
+                op.create_index(index_name, _TABLE_NAME, columns, postgresql_concurrently=True)
+        return
+    for index_name, columns in _INDEXES:
+        op.create_index(index_name, _TABLE_NAME, columns)
+
+
+def downgrade() -> None:
+    if _is_postgresql():
+        with op.get_context().autocommit_block():
+            for index_name, _ in reversed(_INDEXES):
+                op.drop_index(index_name, table_name=_TABLE_NAME, postgresql_concurrently=True)
+        return
+    for index_name, _ in reversed(_INDEXES):
+        op.drop_index(index_name, table_name=_TABLE_NAME)

api/migrations/versions/2026_07_21_1200-a4e7c2f91b30_add_knowledge_fs_control_plane.py

@@ -0,0 +1,387 @@
+"""add independent KnowledgeFS control-plane tables
+
+Revision ID: a4e7c2f91b30
+Revises: 3c9f8e2a1d7b
+Create Date: 2026-07-21 12:00:00.000000
+
+"""
+
+import sqlalchemy as sa
+from alembic import op
+
+import models
+
+# revision identifiers, used by Alembic.
+revision = "a4e7c2f91b30"
+down_revision = "3c9f8e2a1d7b"
+branch_labels = None
+depends_on = None
+
+
+def _uuid_column(name: str, **kwargs):
+    if op.get_bind().dialect.name == "postgresql":
+        kwargs.setdefault("server_default", sa.text("uuidv7()"))
+    return sa.Column(name, models.types.StringUUID(), **kwargs)
+
+
+def _timestamps() -> tuple[sa.Column, sa.Column]:
+    return (
+        sa.Column("created_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.Column("updated_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+    )
+
+
+def upgrade() -> None:
+    op.create_table(
+        "knowledge_fs_control_spaces",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("knowledge_space_id", models.types.StringUUID(), nullable=True),
+        sa.Column("knowledge_space_revision", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("owner_account_id", models.types.StringUUID(), nullable=False),
+        sa.Column("visibility", sa.String(length=32), server_default=sa.text("'only_me'"), nullable=False),
+        sa.Column("provisioning_key", sa.String(length=255), nullable=False),
+        sa.Column("lifecycle_operation_id", models.types.StringUUID(), nullable=True),
+        sa.Column("state", sa.String(length=32), server_default=sa.text("'provisioning'"), nullable=False),
+        sa.Column("resource_version", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("attempt_count", sa.Integer(), server_default=sa.text("0"), nullable=False),
+        sa.Column("last_attempt_at", sa.DateTime(), nullable=True),
+        sa.Column("last_error_code", sa.String(length=255), nullable=True),
+        sa.Column("last_error_message", models.types.LongText(), nullable=True),
+        sa.Column("last_synced_at", sa.DateTime(), nullable=True),
+        sa.Column("deletion_irreversible_at", sa.DateTime(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_control_space_pkey"),
+        sa.UniqueConstraint("tenant_id", "id", name="kfs_control_space_tenant_id_uq"),
+        sa.UniqueConstraint("provisioning_key", name="kfs_control_space_provisioning_key_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id"],
+            ["tenants.id"],
+            name="kfs_control_space_workspace_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("resource_version >= 0", name="kfs_control_space_resource_version_ck"),
+        sa.CheckConstraint("attempt_count >= 0", name="kfs_control_space_attempt_count_ck"),
+        sa.CheckConstraint(
+            "state != 'active' OR knowledge_space_id IS NOT NULL",
+            name="kfs_control_space_active_registration_ck",
+        ),
+        sa.CheckConstraint(
+            "deletion_irreversible_at IS NULL OR state IN ('deleting', 'deleted', 'error')",
+            name="kfs_control_space_irreversible_state_ck",
+        ),
+        sa.CheckConstraint(
+            "knowledge_space_revision >= 0",
+            name="kfs_control_space_remote_revision_ck",
+        ),
+    )
+    if op.get_bind().dialect.name == "postgresql":
+        op.create_index(
+            "kfs_control_space_tenant_space_uq",
+            "knowledge_fs_control_spaces",
+            ["tenant_id", "knowledge_space_id"],
+            unique=True,
+            postgresql_where=sa.text("knowledge_space_id IS NOT NULL"),
+        )
+    else:
+        op.create_index(
+            "kfs_control_space_tenant_space_uq",
+            "knowledge_fs_control_spaces",
+            ["tenant_id", "knowledge_space_id"],
+            unique=True,
+        )
+    op.create_index(
+        "kfs_control_space_state_updated_idx",
+        "knowledge_fs_control_spaces",
+        ["state", "updated_at"],
+    )
+    op.create_index(
+        "kfs_control_space_tenant_state_updated_idx",
+        "knowledge_fs_control_spaces",
+        ["tenant_id", "state", "updated_at"],
+    )
+    op.create_index(
+        "kfs_control_space_tenant_owner_state_idx",
+        "knowledge_fs_control_spaces",
+        ["tenant_id", "owner_account_id", "state"],
+    )
+
+    op.create_table(
+        "knowledge_fs_control_space_permissions",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("account_id", models.types.StringUUID(), nullable=False),
+        sa.Column("role", sa.String(length=32), nullable=False),
+        sa.Column("status", sa.String(length=32), server_default=sa.text("'active'"), nullable=False),
+        sa.Column("revision", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("granted_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("revoked_at", sa.DateTime(), nullable=True),
+        sa.Column("revoked_by_account_id", models.types.StringUUID(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_control_space_permission_pkey"),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "control_space_id",
+            "account_id",
+            name="kfs_control_space_permission_identity_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_control_space_permission_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("revision >= 0", name="kfs_control_space_permission_revision_ck"),
+    )
+    op.create_index(
+        "kfs_control_space_permission_account_idx",
+        "knowledge_fs_control_space_permissions",
+        ["tenant_id", "account_id", "status"],
+    )
+
+    op.create_table(
+        "knowledge_fs_external_access_policies",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("service_api_enabled", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("agent_enabled", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("workflow_enabled", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("mcp_enabled", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("revision", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("updated_by_account_id", models.types.StringUUID(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_external_access_policy_pkey"),
+        sa.UniqueConstraint("tenant_id", "control_space_id", name="kfs_external_access_policy_space_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_external_access_policy_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("revision >= 0", name="kfs_external_access_policy_revision_ck"),
+    )
+
+    op.create_table(
+        "knowledge_fs_api_credentials",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("credential_hash", sa.String(length=255), nullable=False),
+        sa.Column("credential_prefix", sa.String(length=32), nullable=False),
+        sa.Column("credential_last4", sa.String(length=4), nullable=False),
+        sa.Column("principal", sa.String(length=255), nullable=False),
+        sa.Column("allowed_actions", sa.JSON(), nullable=False),
+        sa.Column("status", sa.String(length=32), server_default=sa.text("'active'"), nullable=False),
+        sa.Column("revision", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("expires_at", sa.DateTime(), nullable=True),
+        sa.Column("last_used_at", sa.DateTime(), nullable=True),
+        sa.Column("created_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("revoked_at", sa.DateTime(), nullable=True),
+        sa.Column("revoked_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("revoke_reason", sa.String(length=255), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_api_credential_pkey"),
+        sa.UniqueConstraint("credential_hash", name="kfs_api_credential_hash_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_api_credential_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("revision >= 0", name="kfs_api_credential_revision_ck"),
+    )
+    op.create_index(
+        "kfs_api_credential_tenant_space_status_idx",
+        "knowledge_fs_api_credentials",
+        ["tenant_id", "control_space_id", "status"],
+    )
+    op.create_index(
+        "kfs_api_credential_tenant_prefix_idx",
+        "knowledge_fs_api_credentials",
+        ["tenant_id", "credential_prefix"],
+    )
+
+    op.create_table(
+        "app_knowledge_fs_space_joins",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("app_id", models.types.StringUUID(), nullable=False),
+        sa.Column("join_type", sa.String(length=32), nullable=False),
+        sa.Column("status", sa.String(length=32), server_default=sa.text("'active'"), nullable=False),
+        sa.Column("revision", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("created_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("revoked_at", sa.DateTime(), nullable=True),
+        sa.Column("revoked_by_account_id", models.types.StringUUID(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="app_kfs_space_join_pkey"),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "app_id",
+            "control_space_id",
+            "join_type",
+            name="app_kfs_space_join_identity_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="app_kfs_space_join_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("revision >= 0", name="app_kfs_space_join_revision_ck"),
+    )
+    op.create_index(
+        "app_kfs_space_join_app_status_idx",
+        "app_knowledge_fs_space_joins",
+        ["tenant_id", "app_id", "status"],
+    )
+
+    op.create_table(
+        "knowledge_fs_authorization_revisions",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("membership_epoch", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("space_acl_epoch", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("external_access_epoch", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("content_policy_revision", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("revoke_sequence", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_authorization_revision_pkey"),
+        sa.UniqueConstraint("tenant_id", "control_space_id", name="kfs_authorization_revision_space_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_authorization_revision_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("membership_epoch >= 0", name="kfs_authorization_membership_epoch_ck"),
+        sa.CheckConstraint("space_acl_epoch >= 0", name="kfs_authorization_space_acl_epoch_ck"),
+        sa.CheckConstraint("external_access_epoch >= 0", name="kfs_authorization_external_access_epoch_ck"),
+        sa.CheckConstraint("content_policy_revision >= 0", name="kfs_authorization_content_policy_revision_ck"),
+        sa.CheckConstraint("revoke_sequence >= 0", name="kfs_authorization_revoke_sequence_ck"),
+    )
+
+    op.create_table(
+        "knowledge_fs_capability_issuance_audits",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("trace_id", sa.String(length=128), nullable=False),
+        sa.Column("jti_hash", sa.String(length=80), nullable=False),
+        sa.Column("claims_summary", sa.JSON(), nullable=False),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_capability_issuance_audit_pkey"),
+        sa.UniqueConstraint("jti_hash", name="kfs_capability_issuance_audit_jti_hash_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_capability_issuance_audit_space_fk",
+            ondelete="RESTRICT",
+        ),
+    )
+    op.create_index(
+        "kfs_capability_issuance_audit_space_created_idx",
+        "knowledge_fs_capability_issuance_audits",
+        ["tenant_id", "control_space_id", "created_at"],
+    )
+    op.create_index(
+        "kfs_capability_issuance_audit_trace_idx",
+        "knowledge_fs_capability_issuance_audits",
+        ["tenant_id", "trace_id"],
+    )
+
+    op.create_table(
+        "knowledge_fs_lifecycle_outbox",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("operation_id", models.types.StringUUID(), nullable=False),
+        sa.Column("idempotency_key", sa.String(length=255), nullable=False),
+        sa.Column("operation", sa.String(length=32), nullable=False),
+        sa.Column("command_payload", sa.JSON(), nullable=False),
+        sa.Column("expected_control_space_version", sa.BigInteger(), nullable=False),
+        sa.Column("expected_knowledge_space_revision", sa.BigInteger(), nullable=True),
+        sa.Column("command_schema_version", sa.Integer(), server_default=sa.text("1"), nullable=False),
+        sa.Column("status", sa.String(length=32), server_default=sa.text("'pending'"), nullable=False),
+        sa.Column("attempt_count", sa.Integer(), server_default=sa.text("0"), nullable=False),
+        sa.Column("next_attempt_at", sa.DateTime(), nullable=True),
+        sa.Column("last_attempt_at", sa.DateTime(), nullable=True),
+        sa.Column("lease_owner", sa.String(length=255), nullable=True),
+        sa.Column("lease_expires_at", sa.DateTime(), nullable=True),
+        sa.Column("last_error_code", sa.String(length=255), nullable=True),
+        sa.Column("last_error_message", models.types.LongText(), nullable=True),
+        sa.Column("completed_at", sa.DateTime(), nullable=True),
+        sa.Column("retain_until", sa.DateTime(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_lifecycle_outbox_pkey"),
+        sa.UniqueConstraint("tenant_id", "operation_id", name="kfs_lifecycle_outbox_operation_uq"),
+        sa.UniqueConstraint("tenant_id", "idempotency_key", name="kfs_lifecycle_outbox_idempotency_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_lifecycle_outbox_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("command_schema_version >= 1", name="kfs_lifecycle_outbox_schema_version_ck"),
+        sa.CheckConstraint("expected_control_space_version >= 0", name="kfs_lifecycle_outbox_expected_version_ck"),
+        sa.CheckConstraint("attempt_count >= 0", name="kfs_lifecycle_outbox_attempt_count_ck"),
+        sa.CheckConstraint(
+            "status IN ('pending', 'processing', 'succeeded', 'retry', 'dead_letter')",
+            name="kfs_lifecycle_outbox_status_ck",
+        ),
+        sa.CheckConstraint(
+            "(status = 'processing' AND lease_owner IS NOT NULL AND lease_expires_at IS NOT NULL) "
+            "OR (status != 'processing' AND lease_owner IS NULL AND lease_expires_at IS NULL)",
+            name="kfs_lifecycle_outbox_lease_state_ck",
+        ),
+        sa.CheckConstraint(
+            "(status IN ('succeeded', 'dead_letter') AND completed_at IS NOT NULL) "
+            "OR (status NOT IN ('succeeded', 'dead_letter') AND completed_at IS NULL)",
+            name="kfs_lifecycle_outbox_terminal_state_ck",
+        ),
+    )
+    op.create_index(
+        "kfs_lifecycle_outbox_dispatch_idx",
+        "knowledge_fs_lifecycle_outbox",
+        ["status", "next_attempt_at", "id"],
+    )
+    op.create_index(
+        "kfs_lifecycle_outbox_space_created_idx",
+        "knowledge_fs_lifecycle_outbox",
+        ["tenant_id", "control_space_id", "created_at"],
+    )
+
+
+def downgrade() -> None:
+    op.drop_index("kfs_lifecycle_outbox_space_created_idx", table_name="knowledge_fs_lifecycle_outbox")
+    op.drop_index("kfs_lifecycle_outbox_dispatch_idx", table_name="knowledge_fs_lifecycle_outbox")
+    op.drop_table("knowledge_fs_lifecycle_outbox")
+    op.drop_index(
+        "kfs_capability_issuance_audit_trace_idx",
+        table_name="knowledge_fs_capability_issuance_audits",
+    )
+    op.drop_index(
+        "kfs_capability_issuance_audit_space_created_idx",
+        table_name="knowledge_fs_capability_issuance_audits",
+    )
+    op.drop_table("knowledge_fs_capability_issuance_audits")
+    op.drop_table("knowledge_fs_authorization_revisions")
+    op.drop_index("app_kfs_space_join_app_status_idx", table_name="app_knowledge_fs_space_joins")
+    op.drop_table("app_knowledge_fs_space_joins")
+    op.drop_index("kfs_api_credential_tenant_prefix_idx", table_name="knowledge_fs_api_credentials")
+    op.drop_index("kfs_api_credential_tenant_space_status_idx", table_name="knowledge_fs_api_credentials")
+    op.drop_table("knowledge_fs_api_credentials")
+    op.drop_table("knowledge_fs_external_access_policies")
+    op.drop_index(
+        "kfs_control_space_permission_account_idx",
+        table_name="knowledge_fs_control_space_permissions",
+    )
+    op.drop_table("knowledge_fs_control_space_permissions")
+    op.drop_index("kfs_control_space_tenant_owner_state_idx", table_name="knowledge_fs_control_spaces")
+    op.drop_index("kfs_control_space_tenant_state_updated_idx", table_name="knowledge_fs_control_spaces")
+    op.drop_index("kfs_control_space_state_updated_idx", table_name="knowledge_fs_control_spaces")
+    op.drop_index("kfs_control_space_tenant_space_uq", table_name="knowledge_fs_control_spaces")
+    op.drop_table("knowledge_fs_control_spaces")

api/migrations/versions/2026_07_21_1300-b7f2a9d41c60_add_knowledge_fs_cutover.py

@@ -0,0 +1,205 @@
+"""add KnowledgeFS workspace cutover ledger and migration evidence
+
+Revision ID: b7f2a9d41c60
+Revises: a4e7c2f91b30
+Create Date: 2026-07-21 13:00:00.000000
+
+"""
+
+import sqlalchemy as sa
+from alembic import op
+
+import models
+
+# revision identifiers, used by Alembic.
+revision = "b7f2a9d41c60"
+down_revision = "a4e7c2f91b30"
+branch_labels = None
+depends_on = None
+
+
+def _uuid_column(name: str, **kwargs):
+    if op.get_bind().dialect.name == "postgresql":
+        kwargs.setdefault("server_default", sa.text("uuidv7()"))
+    return sa.Column(name, models.types.StringUUID(), **kwargs)
+
+
+def _timestamps() -> tuple[sa.Column, sa.Column]:
+    return (
+        sa.Column("created_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.Column("updated_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+    )
+
+
+def upgrade() -> None:
+    op.create_table(
+        "knowledge_fs_workspace_cutover_ledgers",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("phase", sa.String(length=32), server_default=sa.text("'inventory'"), nullable=False),
+        sa.Column("source_revision_watermark", sa.JSON(), nullable=False),
+        sa.Column("final_revision_watermark", sa.JSON(), nullable=True),
+        sa.Column("applied_revision_watermark", sa.JSON(), nullable=False),
+        sa.Column("source_task_watermark", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("final_task_watermark", sa.BigInteger(), nullable=True),
+        sa.Column("applied_task_watermark", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("freeze_at", sa.DateTime(), nullable=True),
+        sa.Column("cutover_at", sa.DateTime(), nullable=True),
+        sa.Column("rolled_back_at", sa.DateTime(), nullable=True),
+        sa.Column("rollback_cutoff_at", sa.DateTime(), nullable=True),
+        sa.Column("observation_started_at", sa.DateTime(), nullable=True),
+        sa.Column("observation_window_ends_at", sa.DateTime(), nullable=True),
+        sa.Column("observation_completed_at", sa.DateTime(), nullable=True),
+        sa.Column("maximum_task_expires_at", sa.DateTime(), nullable=True),
+        sa.Column("irreversible_cleanup_at", sa.DateTime(), nullable=True),
+        sa.Column("product_routes_enabled", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("capability_v2_enabled", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("integrated_mode_enabled", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("legacy_acl_read_only", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("smoke_results", sa.JSON(), nullable=True),
+        sa.Column("legacy_dependency_report", sa.JSON(), nullable=True),
+        sa.Column("legacy_dependency_checked_at", sa.DateTime(), nullable=True),
+        sa.Column("legacy_dependency_ready", sa.Boolean(), server_default=sa.text("false"), nullable=False),
+        sa.Column("cas_version", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_workspace_cutover_ledger_pkey"),
+        sa.UniqueConstraint("tenant_id", name="kfs_workspace_cutover_ledger_tenant_uq"),
+        sa.UniqueConstraint("tenant_id", "id", name="kfs_workspace_cutover_ledger_tenant_id_uq"),
+        sa.CheckConstraint("cas_version >= 0", name="kfs_workspace_cutover_cas_version_ck"),
+        sa.CheckConstraint(
+            "source_task_watermark >= 0 AND applied_task_watermark >= 0 "
+            "AND (final_task_watermark IS NULL OR final_task_watermark >= 0)",
+            name="kfs_workspace_cutover_task_watermark_ck",
+        ),
+        sa.CheckConstraint(
+            "cutover_at IS NULL OR freeze_at IS NOT NULL",
+            name="kfs_workspace_cutover_freeze_before_cutover_ck",
+        ),
+        sa.CheckConstraint(
+            "rollback_cutoff_at IS NULL OR cutover_at IS NOT NULL",
+            name="kfs_workspace_cutover_rollback_cutoff_ck",
+        ),
+    )
+    op.create_index(
+        "kfs_workspace_cutover_ledger_phase_updated_idx",
+        "knowledge_fs_workspace_cutover_ledgers",
+        ["phase", "updated_at"],
+    )
+
+    op.create_table(
+        "knowledge_fs_migration_issues",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("ledger_id", models.types.StringUUID(), nullable=False),
+        sa.Column("issue_key", sa.String(length=255), nullable=False),
+        sa.Column("kind", sa.String(length=48), nullable=False),
+        sa.Column("status", sa.String(length=32), server_default=sa.text("'open'"), nullable=False),
+        sa.Column("resource_type", sa.String(length=64), nullable=True),
+        sa.Column("resource_id", sa.String(length=255), nullable=True),
+        sa.Column("details", sa.JSON(), nullable=False),
+        sa.Column("approved_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("approved_at", sa.DateTime(), nullable=True),
+        sa.Column("resolved_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("resolved_at", sa.DateTime(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_migration_issue_pkey"),
+        sa.UniqueConstraint("tenant_id", "ledger_id", "issue_key", name="kfs_migration_issue_key_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_migration_issue_ledger_fk",
+            ondelete="CASCADE",
+        ),
+    )
+    op.create_index(
+        "kfs_migration_issue_gate_idx",
+        "knowledge_fs_migration_issues",
+        ["tenant_id", "ledger_id", "status", "kind"],
+    )
+
+    op.create_table(
+        "knowledge_fs_migration_quarantine",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("ledger_id", models.types.StringUUID(), nullable=False),
+        sa.Column("source_kind", sa.String(length=32), nullable=False),
+        sa.Column("source_id", sa.String(length=255), nullable=False),
+        sa.Column("reason_code", sa.String(length=64), nullable=False),
+        sa.Column("disposition", sa.String(length=32), nullable=False),
+        sa.Column("details", sa.JSON(), nullable=False),
+        sa.Column("resolved_at", sa.DateTime(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_migration_quarantine_pkey"),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "ledger_id",
+            "source_kind",
+            "source_id",
+            name="kfs_migration_quarantine_source_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_migration_quarantine_ledger_fk",
+            ondelete="CASCADE",
+        ),
+    )
+    op.create_index(
+        "kfs_migration_quarantine_disposition_idx",
+        "knowledge_fs_migration_quarantine",
+        ["tenant_id", "ledger_id", "disposition"],
+    )
+
+    op.create_table(
+        "knowledge_fs_shadow_authorization_diffs",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("ledger_id", models.types.StringUUID(), nullable=False),
+        sa.Column("diff_key", sa.String(length=255), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=True),
+        sa.Column("principal", sa.String(length=255), nullable=False),
+        sa.Column("legacy_allowed", sa.Boolean(), nullable=True),
+        sa.Column("dify_allowed", sa.Boolean(), nullable=False),
+        sa.Column("decision", sa.String(length=32), nullable=False),
+        sa.Column("reason", models.types.LongText(), nullable=False),
+        sa.Column("observed_revision", sa.JSON(), nullable=False),
+        sa.Column("status", sa.String(length=32), nullable=False),
+        sa.Column("approved_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("approved_at", sa.DateTime(), nullable=True),
+        sa.Column("resolved_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("resolved_at", sa.DateTime(), nullable=True),
+        *_timestamps(),
+        sa.PrimaryKeyConstraint("id", name="kfs_shadow_authorization_diff_pkey"),
+        sa.UniqueConstraint("tenant_id", "ledger_id", "diff_key", name="kfs_shadow_authorization_diff_key_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_shadow_authorization_diff_ledger_fk",
+            ondelete="CASCADE",
+        ),
+    )
+    op.create_index(
+        "kfs_shadow_authorization_diff_gate_idx",
+        "knowledge_fs_shadow_authorization_diffs",
+        ["tenant_id", "ledger_id", "status", "decision"],
+    )
+
+
+def downgrade() -> None:
+    op.drop_index(
+        "kfs_shadow_authorization_diff_gate_idx",
+        table_name="knowledge_fs_shadow_authorization_diffs",
+    )
+    op.drop_table("knowledge_fs_shadow_authorization_diffs")
+    op.drop_index(
+        "kfs_migration_quarantine_disposition_idx",
+        table_name="knowledge_fs_migration_quarantine",
+    )
+    op.drop_table("knowledge_fs_migration_quarantine")
+    op.drop_index("kfs_migration_issue_gate_idx", table_name="knowledge_fs_migration_issues")
+    op.drop_table("knowledge_fs_migration_issues")
+    op.drop_index(
+        "kfs_workspace_cutover_ledger_phase_updated_idx",
+        table_name="knowledge_fs_workspace_cutover_ledgers",
+    )
+    op.drop_table("knowledge_fs_workspace_cutover_ledgers")

api/migrations/versions/2026_07_21_1400-c8e31b7d52a4_add_knowledge_fs_cleanup_authorization.py

@@ -0,0 +1,365 @@
+"""add guarded cleanup, cutover evidence, and capability issuance reservations
+
+Revision ID: c8e31b7d52a4
+Revises: b7f2a9d41c60
+Create Date: 2026-07-21 14:00:00.000000
+
+"""
+
+import sqlalchemy as sa
+from alembic import op
+
+import models
+
+# revision identifiers, used by Alembic.
+revision = "c8e31b7d52a4"
+down_revision = "b7f2a9d41c60"
+branch_labels = None
+depends_on = None
+
+
+def _uuid_column(name: str, **kwargs):
+    if op.get_bind().dialect.name == "postgresql":
+        kwargs.setdefault("server_default", sa.text("uuidv7()"))
+    return sa.Column(name, models.types.StringUUID(), **kwargs)
+
+
+def upgrade() -> None:
+    op.create_table(
+        "knowledge_fs_cleanup_authorizations",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("ledger_id", models.types.StringUUID(), nullable=False),
+        sa.Column("request_id", models.types.StringUUID(), nullable=False),
+        sa.Column("plan_schema_version", sa.String(length=32), nullable=False),
+        sa.Column("plan_digest", sa.String(length=71), nullable=False),
+        sa.Column("targets", sa.JSON(), nullable=False),
+        sa.Column("readiness_evidence", sa.JSON(), nullable=False),
+        sa.Column("requested_by_account_id", models.types.StringUUID(), nullable=False),
+        sa.Column("requested_at", sa.DateTime(), nullable=False),
+        sa.Column("readiness_ledger_cas_version", sa.BigInteger(), nullable=False),
+        sa.Column("status", sa.String(length=16), server_default=sa.text("'requested'"), nullable=False),
+        sa.Column("approved_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("approved_at", sa.DateTime(), nullable=True),
+        sa.Column("approval_expires_at", sa.DateTime(), nullable=True),
+        sa.Column("approved_ledger_cas_version", sa.BigInteger(), nullable=True),
+        sa.Column("started_by_account_id", models.types.StringUUID(), nullable=True),
+        sa.Column("started_at", sa.DateTime(), nullable=True),
+        sa.Column("started_ledger_cas_version", sa.BigInteger(), nullable=True),
+        sa.Column("row_version", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("created_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.Column("updated_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.PrimaryKeyConstraint("id", name="kfs_cleanup_authorization_pkey"),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "ledger_id",
+            "request_id",
+            name="kfs_cleanup_authorization_request_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_cleanup_authorization_ledger_fk",
+        ),
+        sa.CheckConstraint("row_version >= 0", name="kfs_cleanup_authorization_version_ck"),
+        sa.CheckConstraint(
+            "readiness_ledger_cas_version >= 0 "
+            "AND (approved_ledger_cas_version IS NULL OR approved_ledger_cas_version >= 0) "
+            "AND (started_ledger_cas_version IS NULL OR started_ledger_cas_version >= 0)",
+            name="kfs_cleanup_authorization_ledger_versions_ck",
+        ),
+        sa.CheckConstraint(
+            "(status = 'requested' "
+            "AND approved_by_account_id IS NULL AND approved_at IS NULL "
+            "AND approval_expires_at IS NULL AND approved_ledger_cas_version IS NULL "
+            "AND started_by_account_id IS NULL AND started_at IS NULL "
+            "AND started_ledger_cas_version IS NULL) OR "
+            "(status = 'approved' "
+            "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+            "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+            "AND started_by_account_id IS NULL AND started_at IS NULL "
+            "AND started_ledger_cas_version IS NULL) OR "
+            "(status = 'started' "
+            "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+            "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+            "AND started_by_account_id IS NOT NULL AND started_at IS NOT NULL "
+            "AND started_ledger_cas_version IS NOT NULL)",
+            name="kfs_cleanup_authorization_status_fields_ck",
+        ),
+        sa.CheckConstraint(
+            "approval_expires_at IS NULL OR approval_expires_at > approved_at",
+            name="kfs_cleanup_authorization_approval_window_ck",
+        ),
+    )
+    op.create_index(
+        "kfs_cleanup_authorization_status_idx",
+        "knowledge_fs_cleanup_authorizations",
+        ["tenant_id", "status", "updated_at"],
+    )
+    op.create_table(
+        "knowledge_fs_capability_issuance_reservations",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=False),
+        sa.Column("grant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("trace_id", sa.String(length=128), nullable=False),
+        sa.Column("subject", sa.String(length=255), nullable=False),
+        sa.Column("caller_kind", sa.String(length=32), nullable=False),
+        sa.Column("request_summary", sa.JSON(), nullable=False),
+        sa.Column("status", sa.String(length=16), server_default=sa.text("'reserved'"), nullable=False),
+        sa.Column("issued_at", sa.DateTime(), nullable=True),
+        sa.Column("token_expires_at", sa.DateTime(), nullable=True),
+        sa.Column("failed_at", sa.DateTime(), nullable=True),
+        sa.Column("failure_code", sa.String(length=128), nullable=True),
+        sa.Column("cleanup_after", sa.DateTime(), nullable=True),
+        sa.Column("row_version", sa.BigInteger(), server_default=sa.text("0"), nullable=False),
+        sa.Column("created_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.Column("updated_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.PrimaryKeyConstraint("id", name="kfs_capability_issuance_reservation_pkey"),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "grant_id",
+            name="kfs_capability_issuance_reservation_grant_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_capability_issuance_reservation_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint(
+            "row_version >= 0",
+            name="kfs_capability_issuance_reservation_version_ck",
+        ),
+        sa.CheckConstraint(
+            "(status = 'reserved' AND issued_at IS NULL AND token_expires_at IS NULL "
+            "AND failed_at IS NULL AND failure_code IS NULL AND cleanup_after IS NULL) OR "
+            "(status = 'issued' AND issued_at IS NOT NULL AND token_expires_at IS NOT NULL "
+            "AND failed_at IS NULL AND failure_code IS NULL AND cleanup_after IS NOT NULL) OR "
+            "(status = 'failed' AND issued_at IS NULL AND token_expires_at IS NULL "
+            "AND failed_at IS NOT NULL AND failure_code IS NOT NULL AND cleanup_after IS NOT NULL)",
+            name="kfs_capability_issuance_reservation_status_fields_ck",
+        ),
+        sa.CheckConstraint(
+            "token_expires_at IS NULL OR cleanup_after >= token_expires_at",
+            name="kfs_capability_issuance_reservation_cleanup_window_ck",
+        ),
+    )
+    op.create_index(
+        "kfs_capability_issuance_reservation_subject_idx",
+        "knowledge_fs_capability_issuance_reservations",
+        ["tenant_id", "control_space_id", "subject", "caller_kind"],
+    )
+    op.create_index(
+        "kfs_capability_issuance_reservation_trace_idx",
+        "knowledge_fs_capability_issuance_reservations",
+        ["tenant_id", "trace_id"],
+    )
+    with op.batch_alter_table("knowledge_fs_migration_quarantine", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("resolved_by_operator", sa.String(length=255), nullable=True))
+        batch_op.add_column(sa.Column("resolved_by_account_id", models.types.StringUUID(), nullable=True))
+        batch_op.add_column(sa.Column("evidence", sa.JSON(none_as_null=True), nullable=True))
+        batch_op.add_column(sa.Column("row_version", sa.BigInteger(), server_default=sa.text("0"), nullable=False))
+        batch_op.create_check_constraint(
+            "kfs_migration_quarantine_version_ck",
+            "row_version >= 0",
+        )
+        batch_op.create_check_constraint(
+            "kfs_migration_quarantine_resolution_fields_ck",
+            "(disposition = 'resolved' "
+            "AND resolved_by_operator IS NOT NULL AND resolved_by_account_id IS NOT NULL "
+            "AND evidence IS NOT NULL AND resolved_at IS NOT NULL) OR "
+            "(disposition <> 'resolved' "
+            "AND resolved_by_operator IS NULL AND resolved_by_account_id IS NULL "
+            "AND evidence IS NULL AND resolved_at IS NULL)",
+        )
+    with op.batch_alter_table("knowledge_fs_workspace_cutover_ledgers", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("shadow_started_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("shadow_completed_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("shadow_evidence_digest", sa.String(length=71), nullable=True))
+        batch_op.add_column(
+            sa.Column("shadow_observation_count", sa.BigInteger(), server_default=sa.text("0"), nullable=False)
+        )
+        batch_op.add_column(sa.Column("shadow_window_started_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("shadow_window_ended_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(
+            sa.Column("shadow_traffic_zero", sa.Boolean(), server_default=sa.text("false"), nullable=False)
+        )
+        batch_op.add_column(sa.Column("shadow_traffic_zero_evidence", sa.JSON(none_as_null=True), nullable=True))
+        batch_op.add_column(sa.Column("shadow_latest_observed_revision", sa.JSON(none_as_null=True), nullable=True))
+        batch_op.add_column(sa.Column("shadow_producer", sa.String(length=255), nullable=True))
+        batch_op.add_column(sa.Column("shadow_completed_by_operator", sa.String(length=255), nullable=True))
+        batch_op.add_column(sa.Column("shadow_completed_by_account_id", models.types.StringUUID(), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_id", sa.String(length=255), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_revision", sa.BigInteger(), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_digest", sa.String(length=71), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_control_space_id", models.types.StringUUID(), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_activated_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_updated_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_acknowledged_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_applied", sa.Boolean(), nullable=True))
+        batch_op.add_column(sa.Column("remote_activation_replayed", sa.Boolean(), nullable=True))
+        batch_op.create_check_constraint(
+            "kfs_workspace_cutover_shadow_count_ck",
+            "shadow_observation_count >= 0",
+        )
+        batch_op.create_check_constraint(
+            "kfs_workspace_cutover_shadow_window_ck",
+            "shadow_window_ended_at IS NULL OR shadow_window_started_at IS NOT NULL",
+        )
+        batch_op.create_check_constraint(
+            "kfs_workspace_cutover_shadow_completion_fields_ck",
+            "(shadow_completed_at IS NULL "
+            "AND shadow_evidence_digest IS NULL AND shadow_producer IS NULL "
+            "AND shadow_completed_by_operator IS NULL AND shadow_completed_by_account_id IS NULL "
+            "AND shadow_window_started_at IS NULL AND shadow_window_ended_at IS NULL "
+            "AND shadow_traffic_zero = false AND shadow_traffic_zero_evidence IS NULL "
+            "AND shadow_latest_observed_revision IS NULL) OR "
+            "(shadow_completed_at IS NOT NULL AND shadow_started_at IS NOT NULL "
+            "AND shadow_evidence_digest IS NOT NULL AND shadow_producer IS NOT NULL "
+            "AND shadow_completed_by_operator IS NOT NULL AND shadow_completed_by_account_id IS NOT NULL "
+            "AND ((shadow_traffic_zero = true AND shadow_observation_count = 0 "
+            "AND shadow_traffic_zero_evidence IS NOT NULL AND shadow_window_started_at IS NULL "
+            "AND shadow_window_ended_at IS NULL AND shadow_latest_observed_revision IS NULL) OR "
+            "(shadow_traffic_zero = false AND shadow_observation_count > 0 "
+            "AND shadow_traffic_zero_evidence IS NULL AND shadow_window_started_at IS NOT NULL "
+            "AND shadow_window_ended_at IS NOT NULL AND shadow_latest_observed_revision IS NOT NULL)))",
+        )
+        batch_op.create_check_constraint(
+            "kfs_workspace_cutover_remote_activation_fields_ck",
+            "(remote_activation_id IS NULL AND remote_activation_revision IS NULL "
+            "AND remote_activation_digest IS NULL AND remote_activation_control_space_id IS NULL "
+            "AND remote_activation_activated_at IS NULL AND remote_activation_updated_at IS NULL "
+            "AND remote_activation_acknowledged_at IS NULL AND remote_activation_applied IS NULL "
+            "AND remote_activation_replayed IS NULL) OR "
+            "(remote_activation_id IS NOT NULL AND remote_activation_revision BETWEEN 1 AND 9007199254740991 "
+            "AND remote_activation_digest IS NOT NULL AND remote_activation_control_space_id IS NOT NULL "
+            "AND remote_activation_activated_at IS NOT NULL AND remote_activation_updated_at IS NOT NULL "
+            "AND remote_activation_acknowledged_at IS NOT NULL "
+            "AND ((remote_activation_applied = true AND remote_activation_replayed = false) "
+            "OR (remote_activation_applied = false AND remote_activation_replayed = true)))",
+        )
+        batch_op.create_check_constraint(
+            "kfs_workspace_cutover_remote_activation_time_ck",
+            "remote_activation_updated_at IS NULL OR remote_activation_updated_at >= remote_activation_activated_at",
+        )
+    with op.batch_alter_table("knowledge_fs_shadow_authorization_diffs", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("current_evidence_digest", sa.String(length=71), nullable=True))
+        batch_op.add_column(sa.Column("last_observed_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("row_version", sa.BigInteger(), server_default=sa.text("0"), nullable=False))
+        batch_op.create_check_constraint(
+            "kfs_shadow_authorization_diff_version_ck",
+            "row_version >= 0",
+        )
+    op.execute(
+        sa.text(
+            "UPDATE knowledge_fs_shadow_authorization_diffs "
+            "SET current_evidence_digest = 'sha256:legacy-unversioned', last_observed_at = created_at "
+            "WHERE current_evidence_digest IS NULL OR last_observed_at IS NULL"
+        )
+    )
+    with op.batch_alter_table("knowledge_fs_shadow_authorization_diffs", schema=None) as batch_op:
+        batch_op.alter_column("current_evidence_digest", existing_type=sa.String(length=71), nullable=False)
+        batch_op.alter_column("last_observed_at", existing_type=sa.DateTime(), nullable=False)
+    op.create_table(
+        "knowledge_fs_shadow_authorization_observations",
+        _uuid_column("id", nullable=False),
+        sa.Column("tenant_id", models.types.StringUUID(), nullable=False),
+        sa.Column("ledger_id", models.types.StringUUID(), nullable=False),
+        sa.Column("diff_key", sa.String(length=255), nullable=False),
+        sa.Column("producer", sa.String(length=255), nullable=False),
+        sa.Column("control_space_id", models.types.StringUUID(), nullable=True),
+        sa.Column("principal", sa.String(length=255), nullable=False),
+        sa.Column("legacy_allowed", sa.Boolean(), nullable=True),
+        sa.Column("dify_allowed", sa.Boolean(), nullable=False),
+        sa.Column("decision", sa.String(length=32), nullable=False),
+        sa.Column("reason", models.types.LongText(), nullable=False),
+        sa.Column("observed_revision", sa.JSON(), nullable=False),
+        sa.Column("observed_at", sa.DateTime(), nullable=False),
+        sa.Column("evidence_digest", sa.String(length=71), nullable=False),
+        sa.Column("created_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.Column("updated_at", sa.DateTime(), server_default=sa.text("CURRENT_TIMESTAMP"), nullable=False),
+        sa.PrimaryKeyConstraint("id", name="kfs_shadow_authorization_observation_pkey"),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "ledger_id",
+            "diff_key",
+            "evidence_digest",
+            name="kfs_shadow_authorization_observation_evidence_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_shadow_authorization_observation_ledger_fk",
+            ondelete="CASCADE",
+        ),
+    )
+    op.create_index(
+        "kfs_shadow_authorization_observation_window_idx",
+        "knowledge_fs_shadow_authorization_observations",
+        ["tenant_id", "ledger_id", "observed_at"],
+    )
+
+
+def downgrade() -> None:
+    op.drop_index(
+        "kfs_shadow_authorization_observation_window_idx",
+        table_name="knowledge_fs_shadow_authorization_observations",
+    )
+    op.drop_table("knowledge_fs_shadow_authorization_observations")
+    with op.batch_alter_table("knowledge_fs_shadow_authorization_diffs", schema=None) as batch_op:
+        batch_op.drop_constraint("kfs_shadow_authorization_diff_version_ck", type_="check")
+        batch_op.drop_column("row_version")
+        batch_op.drop_column("last_observed_at")
+        batch_op.drop_column("current_evidence_digest")
+    with op.batch_alter_table("knowledge_fs_workspace_cutover_ledgers", schema=None) as batch_op:
+        batch_op.drop_constraint("kfs_workspace_cutover_remote_activation_time_ck", type_="check")
+        batch_op.drop_constraint("kfs_workspace_cutover_remote_activation_fields_ck", type_="check")
+        batch_op.drop_constraint("kfs_workspace_cutover_shadow_completion_fields_ck", type_="check")
+        batch_op.drop_constraint("kfs_workspace_cutover_shadow_window_ck", type_="check")
+        batch_op.drop_constraint("kfs_workspace_cutover_shadow_count_ck", type_="check")
+        batch_op.drop_column("remote_activation_replayed")
+        batch_op.drop_column("remote_activation_applied")
+        batch_op.drop_column("remote_activation_acknowledged_at")
+        batch_op.drop_column("remote_activation_updated_at")
+        batch_op.drop_column("remote_activation_activated_at")
+        batch_op.drop_column("remote_activation_control_space_id")
+        batch_op.drop_column("remote_activation_digest")
+        batch_op.drop_column("remote_activation_revision")
+        batch_op.drop_column("remote_activation_id")
+        batch_op.drop_column("shadow_completed_by_account_id")
+        batch_op.drop_column("shadow_completed_by_operator")
+        batch_op.drop_column("shadow_producer")
+        batch_op.drop_column("shadow_latest_observed_revision")
+        batch_op.drop_column("shadow_traffic_zero_evidence")
+        batch_op.drop_column("shadow_traffic_zero")
+        batch_op.drop_column("shadow_window_ended_at")
+        batch_op.drop_column("shadow_window_started_at")
+        batch_op.drop_column("shadow_observation_count")
+        batch_op.drop_column("shadow_evidence_digest")
+        batch_op.drop_column("shadow_completed_at")
+        batch_op.drop_column("shadow_started_at")
+    with op.batch_alter_table("knowledge_fs_migration_quarantine", schema=None) as batch_op:
+        batch_op.drop_constraint(
+            "kfs_migration_quarantine_resolution_fields_ck",
+            type_="check",
+        )
+        batch_op.drop_constraint("kfs_migration_quarantine_version_ck", type_="check")
+        batch_op.drop_column("row_version")
+        batch_op.drop_column("evidence")
+        batch_op.drop_column("resolved_by_account_id")
+        batch_op.drop_column("resolved_by_operator")
+    op.drop_index(
+        "kfs_capability_issuance_reservation_trace_idx",
+        table_name="knowledge_fs_capability_issuance_reservations",
+    )
+    op.drop_index(
+        "kfs_capability_issuance_reservation_subject_idx",
+        table_name="knowledge_fs_capability_issuance_reservations",
+    )
+    op.drop_table("knowledge_fs_capability_issuance_reservations")
+    op.drop_index(
+        "kfs_cleanup_authorization_status_idx",
+        table_name="knowledge_fs_cleanup_authorizations",
+    )
+    op.drop_table("knowledge_fs_cleanup_authorizations")

api/migrations/versions/2026_07_21_1500-d4f6e8a1c305_add_knowledge_fs_remote_freeze_evidence.py

@@ -0,0 +1,66 @@
+"""add durable KnowledgeFS remote freeze evidence
+
+Revision ID: d4f6e8a1c305
+Revises: c8e31b7d52a4
+Create Date: 2026-07-21 15:00:00.000000
+
+"""
+
+import sqlalchemy as sa
+from alembic import op
+
+import models
+
+# revision identifiers, used by Alembic.
+revision = "d4f6e8a1c305"
+down_revision = "c8e31b7d52a4"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("knowledge_fs_workspace_cutover_ledgers", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("remote_freeze_id", sa.String(length=255), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_revision", sa.BigInteger(), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_digest", sa.String(length=71), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_task_watermark", sa.BigInteger(), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_control_space_id", models.types.StringUUID(), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_frozen_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_updated_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_acknowledged_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_applied", sa.Boolean(), nullable=True))
+        batch_op.add_column(sa.Column("remote_freeze_replayed", sa.Boolean(), nullable=True))
+        batch_op.create_check_constraint(
+            "kfs_workspace_cutover_remote_freeze_fields_ck",
+            "(remote_freeze_id IS NULL AND remote_freeze_revision IS NULL "
+            "AND remote_freeze_digest IS NULL AND remote_freeze_task_watermark IS NULL "
+            "AND remote_freeze_control_space_id IS NULL AND remote_freeze_frozen_at IS NULL "
+            "AND remote_freeze_updated_at IS NULL AND remote_freeze_acknowledged_at IS NULL "
+            "AND remote_freeze_applied IS NULL AND remote_freeze_replayed IS NULL) OR "
+            "(remote_freeze_id IS NOT NULL AND remote_freeze_revision BETWEEN 1 AND 9007199254740991 "
+            "AND remote_freeze_digest IS NOT NULL AND remote_freeze_task_watermark >= 0 "
+            "AND remote_freeze_control_space_id IS NOT NULL AND remote_freeze_frozen_at IS NOT NULL "
+            "AND remote_freeze_updated_at IS NOT NULL AND remote_freeze_acknowledged_at IS NOT NULL "
+            "AND ((remote_freeze_applied = true AND remote_freeze_replayed = false) "
+            "OR (remote_freeze_applied = false AND remote_freeze_replayed = true)))",
+        )
+        batch_op.create_check_constraint(
+            "kfs_workspace_cutover_remote_freeze_time_ck",
+            "remote_freeze_updated_at IS NULL OR remote_freeze_updated_at >= remote_freeze_frozen_at",
+        )
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("knowledge_fs_workspace_cutover_ledgers", schema=None) as batch_op:
+        batch_op.drop_constraint("kfs_workspace_cutover_remote_freeze_time_ck", type_="check")
+        batch_op.drop_constraint("kfs_workspace_cutover_remote_freeze_fields_ck", type_="check")
+        batch_op.drop_column("remote_freeze_replayed")
+        batch_op.drop_column("remote_freeze_applied")
+        batch_op.drop_column("remote_freeze_acknowledged_at")
+        batch_op.drop_column("remote_freeze_updated_at")
+        batch_op.drop_column("remote_freeze_frozen_at")
+        batch_op.drop_column("remote_freeze_control_space_id")
+        batch_op.drop_column("remote_freeze_task_watermark")
+        batch_op.drop_column("remote_freeze_digest")
+        batch_op.drop_column("remote_freeze_revision")
+        batch_op.drop_column("remote_freeze_id")

api/migrations/versions/2026_07_21_1600-e5a7c9b2d416_add_knowledge_fs_cleanup_completion.py

@@ -0,0 +1,113 @@
+"""add global KnowledgeFS cleanup completion evidence
+
+Revision ID: e5a7c9b2d416
+Revises: d4f6e8a1c305
+Create Date: 2026-07-21 16:00:00.000000
+
+"""
+
+import sqlalchemy as sa
+from alembic import op
+
+import models
+
+# revision identifiers, used by Alembic.
+revision = "e5a7c9b2d416"
+down_revision = "d4f6e8a1c305"
+branch_labels = None
+depends_on = None
+
+
+_LEDGER_VERSIONS_CHECK = (
+    "readiness_ledger_cas_version >= 0 "
+    "AND (approved_ledger_cas_version IS NULL OR approved_ledger_cas_version >= 0) "
+    "AND (started_ledger_cas_version IS NULL OR started_ledger_cas_version >= 0) "
+    "AND (completed_ledger_cas_version IS NULL OR completed_ledger_cas_version >= 0)"
+)
+_STATUS_FIELDS_CHECK = (
+    "(status = 'requested' "
+    "AND approved_by_account_id IS NULL AND approved_at IS NULL "
+    "AND approval_expires_at IS NULL AND approved_ledger_cas_version IS NULL "
+    "AND started_by_account_id IS NULL AND started_at IS NULL "
+    "AND started_ledger_cas_version IS NULL "
+    "AND completed_by_account_id IS NULL AND completed_at IS NULL "
+    "AND completion_evidence IS NULL AND completed_ledger_cas_version IS NULL) OR "
+    "(status = 'approved' "
+    "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+    "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+    "AND started_by_account_id IS NULL AND started_at IS NULL "
+    "AND started_ledger_cas_version IS NULL "
+    "AND completed_by_account_id IS NULL AND completed_at IS NULL "
+    "AND completion_evidence IS NULL AND completed_ledger_cas_version IS NULL) OR "
+    "(status = 'started' "
+    "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+    "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+    "AND started_by_account_id IS NOT NULL AND started_at IS NOT NULL "
+    "AND started_ledger_cas_version IS NOT NULL "
+    "AND completed_by_account_id IS NULL AND completed_at IS NULL "
+    "AND completion_evidence IS NULL AND completed_ledger_cas_version IS NULL) OR "
+    "(status = 'completed' "
+    "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+    "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+    "AND started_by_account_id IS NOT NULL AND started_at IS NOT NULL "
+    "AND started_ledger_cas_version IS NOT NULL "
+    "AND completed_by_account_id IS NOT NULL AND completed_at IS NOT NULL "
+    "AND completion_evidence IS NOT NULL AND completed_ledger_cas_version IS NOT NULL)"
+)
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("knowledge_fs_cleanup_authorizations", schema=None) as batch_op:
+        batch_op.drop_constraint("kfs_cleanup_authorization_status_fields_ck", type_="check")
+        batch_op.drop_constraint("kfs_cleanup_authorization_ledger_versions_ck", type_="check")
+        batch_op.add_column(sa.Column("completed_by_account_id", models.types.StringUUID(), nullable=True))
+        batch_op.add_column(sa.Column("completed_at", sa.DateTime(), nullable=True))
+        batch_op.add_column(sa.Column("completion_evidence", sa.JSON(none_as_null=True), nullable=True))
+        batch_op.add_column(sa.Column("completed_ledger_cas_version", sa.BigInteger(), nullable=True))
+        batch_op.create_check_constraint(
+            "kfs_cleanup_authorization_ledger_versions_ck",
+            _LEDGER_VERSIONS_CHECK,
+        )
+        batch_op.create_check_constraint(
+            "kfs_cleanup_authorization_status_fields_ck",
+            _STATUS_FIELDS_CHECK,
+        )
+        batch_op.create_check_constraint(
+            "kfs_cleanup_authorization_completion_time_ck",
+            "completed_at IS NULL OR completed_at >= started_at",
+        )
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("knowledge_fs_cleanup_authorizations", schema=None) as batch_op:
+        batch_op.drop_constraint("kfs_cleanup_authorization_completion_time_ck", type_="check")
+        batch_op.drop_constraint("kfs_cleanup_authorization_status_fields_ck", type_="check")
+        batch_op.drop_constraint("kfs_cleanup_authorization_ledger_versions_ck", type_="check")
+        batch_op.create_check_constraint(
+            "kfs_cleanup_authorization_ledger_versions_ck",
+            "readiness_ledger_cas_version >= 0 "
+            "AND (approved_ledger_cas_version IS NULL OR approved_ledger_cas_version >= 0) "
+            "AND (started_ledger_cas_version IS NULL OR started_ledger_cas_version >= 0)",
+        )
+        batch_op.create_check_constraint(
+            "kfs_cleanup_authorization_status_fields_ck",
+            "(status = 'requested' "
+            "AND approved_by_account_id IS NULL AND approved_at IS NULL "
+            "AND approval_expires_at IS NULL AND approved_ledger_cas_version IS NULL "
+            "AND started_by_account_id IS NULL AND started_at IS NULL "
+            "AND started_ledger_cas_version IS NULL) OR "
+            "(status = 'approved' "
+            "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+            "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+            "AND started_by_account_id IS NULL AND started_at IS NULL "
+            "AND started_ledger_cas_version IS NULL) OR "
+            "(status = 'started' "
+            "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+            "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+            "AND started_by_account_id IS NOT NULL AND started_at IS NOT NULL "
+            "AND started_ledger_cas_version IS NOT NULL)",
+        )
+        batch_op.drop_column("completed_ledger_cas_version")
+        batch_op.drop_column("completion_evidence")
+        batch_op.drop_column("completed_at")
+        batch_op.drop_column("completed_by_account_id")

api/models/__init__.py

@@ -66,6 +66,45 @@ from .enums import (
 )
 from .execution_extra_content import ExecutionExtraContent, HumanInputContent
 from .human_input import HumanInputForm, HumanInputFormUploadFile, HumanInputFormUploadToken
+from .knowledge_fs import (
+    AppKnowledgeFSSpaceJoin,
+    KnowledgeFSApiCredential,
+    KnowledgeFSApiCredentialStatus,
+    KnowledgeFSAppSpaceJoinStatus,
+    KnowledgeFSAppSpaceJoinType,
+    KnowledgeFSAuthorizationRevision,
+    KnowledgeFSCapabilityIssuanceAudit,
+    KnowledgeFSCapabilityIssuanceReservation,
+    KnowledgeFSCapabilityIssuanceReservationStatus,
+    KnowledgeFSControlSpace,
+    KnowledgeFSControlSpacePermission,
+    KnowledgeFSControlSpacePermissionRole,
+    KnowledgeFSControlSpacePermissionStatus,
+    KnowledgeFSControlSpaceState,
+    KnowledgeFSControlSpaceVisibility,
+    KnowledgeFSExternalAccessPolicy,
+    KnowledgeFSLifecycleOperation,
+    KnowledgeFSLifecycleOutbox,
+    KnowledgeFSLifecycleOutboxStatus,
+)
+from .knowledge_fs_cleanup import (
+    KnowledgeFSCleanupAuthorization,
+    KnowledgeFSCleanupAuthorizationStatus,
+    KnowledgeFSCleanupTarget,
+)
+from .knowledge_fs_cutover import (
+    KnowledgeFSMigrationIssue,
+    KnowledgeFSMigrationIssueKind,
+    KnowledgeFSMigrationIssueStatus,
+    KnowledgeFSMigrationQuarantine,
+    KnowledgeFSMigrationQuarantineDisposition,
+    KnowledgeFSMigrationQuarantineKind,
+    KnowledgeFSShadowAuthorizationDecision,
+    KnowledgeFSShadowAuthorizationDiff,
+    KnowledgeFSShadowAuthorizationObservation,
+    KnowledgeFSWorkspaceCutoverLedger,
+    KnowledgeFSWorkspaceCutoverPhase,
+)
 from .model import (
     AccountTrialAppRecord,
     ApiRequest,
@@ -180,6 +219,7 @@ __all__ = [
     "AppAnnotationHitHistory",
     "AppAnnotationSetting",
     "AppDatasetJoin",
+    "AppKnowledgeFSSpaceJoin",
     "AppMCPServer",
     "AppMode",
     "AppModelConfig",
@@ -224,6 +264,38 @@ __all__ = [
     "IconType",
     "InstalledApp",
     "InvitationCode",
+    "KnowledgeFSApiCredential",
+    "KnowledgeFSApiCredentialStatus",
+    "KnowledgeFSAppSpaceJoinStatus",
+    "KnowledgeFSAppSpaceJoinType",
+    "KnowledgeFSAuthorizationRevision",
+    "KnowledgeFSCapabilityIssuanceAudit",
+    "KnowledgeFSCapabilityIssuanceReservation",
+    "KnowledgeFSCapabilityIssuanceReservationStatus",
+    "KnowledgeFSCleanupAuthorization",
+    "KnowledgeFSCleanupAuthorizationStatus",
+    "KnowledgeFSCleanupTarget",
+    "KnowledgeFSControlSpace",
+    "KnowledgeFSControlSpacePermission",
+    "KnowledgeFSControlSpacePermissionRole",
+    "KnowledgeFSControlSpacePermissionStatus",
+    "KnowledgeFSControlSpaceState",
+    "KnowledgeFSControlSpaceVisibility",
+    "KnowledgeFSExternalAccessPolicy",
+    "KnowledgeFSLifecycleOperation",
+    "KnowledgeFSLifecycleOutbox",
+    "KnowledgeFSLifecycleOutboxStatus",
+    "KnowledgeFSMigrationIssue",
+    "KnowledgeFSMigrationIssueKind",
+    "KnowledgeFSMigrationIssueStatus",
+    "KnowledgeFSMigrationQuarantine",
+    "KnowledgeFSMigrationQuarantineDisposition",
+    "KnowledgeFSMigrationQuarantineKind",
+    "KnowledgeFSShadowAuthorizationDecision",
+    "KnowledgeFSShadowAuthorizationDiff",
+    "KnowledgeFSShadowAuthorizationObservation",
+    "KnowledgeFSWorkspaceCutoverLedger",
+    "KnowledgeFSWorkspaceCutoverPhase",
     "LoadBalancingModelConfig",
     "Message",
     "MessageAgentThought",

api/models/knowledge_fs.py

@@ -0,0 +1,713 @@
+"""Dify-owned control-plane persistence for the KnowledgeFS product.
+
+These models intentionally have no dependency on Dify Dataset or Document
+models. KnowledgeFS data-plane objects remain owned by KnowledgeFS; these rows
+only describe Dify authorization, registration, and durable lifecycle intent.
+"""
+
+from __future__ import annotations
+
+from datetime import datetime
+from enum import StrEnum
+from typing import NotRequired, TypedDict
+
+import sqlalchemy as sa
+from sqlalchemy import DateTime, Index, String, UniqueConstraint
+from sqlalchemy.orm import Mapped, mapped_column
+
+from .base import DefaultFieldsDCMixin, TypeBase
+from .types import EnumText, LongText, StringUUID
+
+KnowledgeFSAllowedActions = list[str]
+
+
+class KnowledgeFSCommandEnvelope(TypedDict):
+    schema_version: int
+    idempotency_key: str
+    expected_revision: int
+
+
+class KnowledgeFSModelSelectionIntentPayload(TypedDict):
+    pluginId: str
+    provider: str
+    model: str
+
+
+class KnowledgeFSRerankIntentPayload(TypedDict):
+    enabled: bool
+    model: NotRequired[KnowledgeFSModelSelectionIntentPayload]
+
+
+class KnowledgeFSScoreThresholdIntentPayload(TypedDict):
+    enabled: bool
+    stage: str
+    value: NotRequired[float]
+
+
+class KnowledgeFSRetrievalProfileIntentPayload(TypedDict):
+    defaultMode: str
+    reasoningModel: KnowledgeFSModelSelectionIntentPayload
+    rerank: KnowledgeFSRerankIntentPayload
+    scoreThreshold: KnowledgeFSScoreThresholdIntentPayload
+    topK: int
+
+
+class KnowledgeFSProvisionCommandPayload(KnowledgeFSCommandEnvelope):
+    provisioning_key: str
+    name: str
+    icon: str | None
+    description: str | None
+    slug: str
+    model_intent: KnowledgeFSModelSelectionIntentPayload
+    profile_intent: KnowledgeFSRetrievalProfileIntentPayload
+
+
+class KnowledgeFSMetadataUpdateCommandPayload(KnowledgeFSCommandEnvelope):
+    metadata: dict[str, object]
+
+
+class KnowledgeFSDeleteCommandPayload(KnowledgeFSCommandEnvelope):
+    knowledge_space_id: str | None
+    provisioning_key: str
+
+
+class KnowledgeFSRevokeCommandPayload(KnowledgeFSCommandEnvelope):
+    event_id: str
+    grant_id: str
+    knowledge_space_id: str
+    principal: str
+    reason_code: str
+    revoke_sequence: int
+
+
+class KnowledgeFSRepairCommandPayload(KnowledgeFSCommandEnvelope):
+    repair_reason: str
+
+
+KnowledgeFSCommandPayload = (
+    KnowledgeFSProvisionCommandPayload
+    | KnowledgeFSMetadataUpdateCommandPayload
+    | KnowledgeFSDeleteCommandPayload
+    | KnowledgeFSRevokeCommandPayload
+    | KnowledgeFSRepairCommandPayload
+)
+
+
+class KnowledgeFSCapabilityAuthzRevisionSummary(TypedDict):
+    membership_epoch: int
+    space_acl_epoch: int
+    external_access_epoch: int
+    credential_revision: int | None
+
+
+class KnowledgeFSCapabilityClaimsSummary(TypedDict):
+    action: str
+    actor: str
+    authz_revision: KnowledgeFSCapabilityAuthzRevisionSummary
+    caller_kind: str
+    content_policy_revision: int
+    content_scope_ids: list[str]
+    control_space_id: str
+    expires_at: str
+    grant_id: str
+    issued_at: str
+    namespace_id: str
+    operation_id: str
+    resource_id: str
+    resource_parent_id: str | None
+    resource_type: str
+    subject: str
+
+
+class KnowledgeFSCapabilityReservationSummary(TypedDict):
+    """Sanitized, exact input bound before a capability can be signed."""
+
+    action: str
+    actor: str
+    authz_revision: KnowledgeFSCapabilityAuthzRevisionSummary
+    caller_kind: str
+    content_policy_revision: int
+    content_scope_ids: list[str]
+    control_space_id: str
+    grant_id: str
+    namespace_id: str
+    operation_id: str
+    resource_id: str
+    resource_parent_id: str | None
+    resource_type: str
+    subject: str
+    trace_id: str
+
+
+class KnowledgeFSControlSpaceVisibility(StrEnum):
+    ONLY_ME = "only_me"
+    ALL_TEAM_MEMBERS = "all_team_members"
+    PARTIAL_MEMBERS = "partial_members"
+
+
+class KnowledgeFSControlSpaceState(StrEnum):
+    PROVISIONING = "provisioning"
+    ACTIVE = "active"
+    DELETING = "deleting"
+    DELETED = "deleted"
+    ERROR = "error"
+
+
+class KnowledgeFSControlSpacePermissionRole(StrEnum):
+    OWNER = "owner"
+    EDITOR = "editor"
+    VIEWER = "viewer"
+
+
+class KnowledgeFSControlSpacePermissionStatus(StrEnum):
+    ACTIVE = "active"
+    REVOKED = "revoked"
+
+
+class KnowledgeFSApiCredentialStatus(StrEnum):
+    ACTIVE = "active"
+    REVOKED = "revoked"
+    EXPIRED = "expired"
+
+
+class KnowledgeFSAppSpaceJoinType(StrEnum):
+    AGENT = "agent"
+    WORKFLOW = "workflow"
+
+
+class KnowledgeFSAppSpaceJoinStatus(StrEnum):
+    ACTIVE = "active"
+    REVOKED = "revoked"
+
+
+class KnowledgeFSCapabilityIssuanceReservationStatus(StrEnum):
+    RESERVED = "reserved"
+    ISSUED = "issued"
+    FAILED = "failed"
+
+
+class KnowledgeFSLifecycleOperation(StrEnum):
+    PROVISION = "provision"
+    METADATA_UPDATE = "metadata_update"
+    DELETE = "delete"
+    REVOKE = "revoke"
+    REPAIR = "repair"
+
+
+class KnowledgeFSLifecycleOutboxStatus(StrEnum):
+    PENDING = "pending"
+    PROCESSING = "processing"
+    SUCCEEDED = "succeeded"
+    RETRY = "retry"
+    DEAD_LETTER = "dead_letter"
+
+
+class KnowledgeFSControlSpace(DefaultFieldsDCMixin, TypeBase):
+    """Dify product resource registered to at most one KnowledgeFS Space."""
+
+    __tablename__ = "knowledge_fs_control_spaces"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_control_space_pkey"),
+        UniqueConstraint("tenant_id", "id", name="kfs_control_space_tenant_id_uq"),
+        UniqueConstraint("provisioning_key", name="kfs_control_space_provisioning_key_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id"],
+            ["tenants.id"],
+            name="kfs_control_space_workspace_fk",
+            ondelete="RESTRICT",
+        ),
+        Index(
+            "kfs_control_space_tenant_space_uq",
+            "tenant_id",
+            "knowledge_space_id",
+            unique=True,
+            postgresql_where=sa.text("knowledge_space_id IS NOT NULL"),
+        ),
+        Index("kfs_control_space_state_updated_idx", "state", "updated_at"),
+        Index("kfs_control_space_tenant_state_updated_idx", "tenant_id", "state", "updated_at"),
+        Index("kfs_control_space_tenant_owner_state_idx", "tenant_id", "owner_account_id", "state"),
+        sa.CheckConstraint(
+            "resource_version >= 0",
+            name=sa.schema.conv("kfs_control_space_resource_version_ck"),
+        ),
+        sa.CheckConstraint("attempt_count >= 0", name=sa.schema.conv("kfs_control_space_attempt_count_ck")),
+        sa.CheckConstraint(
+            "state != 'active' OR knowledge_space_id IS NOT NULL",
+            name=sa.schema.conv("kfs_control_space_active_registration_ck"),
+        ),
+        sa.CheckConstraint(
+            "deletion_irreversible_at IS NULL OR state IN ('deleting', 'deleted', 'error')",
+            name=sa.schema.conv("kfs_control_space_irreversible_state_ck"),
+        ),
+        sa.CheckConstraint(
+            "knowledge_space_revision >= 0",
+            name=sa.schema.conv("kfs_control_space_remote_revision_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    owner_account_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    provisioning_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    knowledge_space_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    knowledge_space_revision: Mapped[int] = mapped_column(
+        sa.BigInteger,
+        nullable=False,
+        server_default=sa.text("0"),
+        default=0,
+    )
+    visibility: Mapped[KnowledgeFSControlSpaceVisibility] = mapped_column(
+        EnumText(KnowledgeFSControlSpaceVisibility, length=32),
+        nullable=False,
+        server_default=sa.text("'only_me'"),
+        default=KnowledgeFSControlSpaceVisibility.ONLY_ME,
+    )
+    lifecycle_operation_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    state: Mapped[KnowledgeFSControlSpaceState] = mapped_column(
+        EnumText(KnowledgeFSControlSpaceState, length=32),
+        nullable=False,
+        server_default=sa.text("'provisioning'"),
+        default=KnowledgeFSControlSpaceState.PROVISIONING,
+    )
+    resource_version: Mapped[int] = mapped_column(
+        sa.BigInteger,
+        nullable=False,
+        server_default=sa.text("0"),
+        default=0,
+    )
+    attempt_count: Mapped[int] = mapped_column(
+        sa.Integer,
+        nullable=False,
+        server_default=sa.text("0"),
+        default=0,
+    )
+    last_attempt_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    last_error_code: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    last_error_message: Mapped[str | None] = mapped_column(LongText, nullable=True, default=None)
+    last_synced_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    deletion_irreversible_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+
+
+class KnowledgeFSControlSpacePermission(DefaultFieldsDCMixin, TypeBase):
+    """Account-level product authorization for one control-space."""
+
+    __tablename__ = "knowledge_fs_control_space_permissions"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_control_space_permission_pkey"),
+        UniqueConstraint(
+            "tenant_id",
+            "control_space_id",
+            "account_id",
+            name="kfs_control_space_permission_identity_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_control_space_permission_space_fk",
+            ondelete="RESTRICT",
+        ),
+        Index("kfs_control_space_permission_account_idx", "tenant_id", "account_id", "status"),
+        sa.CheckConstraint(
+            "revision >= 0",
+            name=sa.schema.conv("kfs_control_space_permission_revision_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    account_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    role: Mapped[KnowledgeFSControlSpacePermissionRole] = mapped_column(
+        EnumText(KnowledgeFSControlSpacePermissionRole, length=32), nullable=False
+    )
+    status: Mapped[KnowledgeFSControlSpacePermissionStatus] = mapped_column(
+        EnumText(KnowledgeFSControlSpacePermissionStatus, length=32),
+        nullable=False,
+        server_default=sa.text("'active'"),
+        default=KnowledgeFSControlSpacePermissionStatus.ACTIVE,
+    )
+    revision: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+    granted_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    revoked_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    revoked_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+
+
+class KnowledgeFSExternalAccessPolicy(DefaultFieldsDCMixin, TypeBase):
+    """Fail-closed caller-channel policy for a control-space."""
+
+    __tablename__ = "knowledge_fs_external_access_policies"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_external_access_policy_pkey"),
+        UniqueConstraint("tenant_id", "control_space_id", name="kfs_external_access_policy_space_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_external_access_policy_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint("revision >= 0", name=sa.schema.conv("kfs_external_access_policy_revision_ck")),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    service_api_enabled: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    agent_enabled: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    workflow_enabled: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    mcp_enabled: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    revision: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+    updated_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+
+
+class KnowledgeFSApiCredential(DefaultFieldsDCMixin, TypeBase):
+    """Hashed, resource-bound KnowledgeFS API credential metadata."""
+
+    __tablename__ = "knowledge_fs_api_credentials"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_api_credential_pkey"),
+        UniqueConstraint("credential_hash", name="kfs_api_credential_hash_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_api_credential_space_fk",
+            ondelete="RESTRICT",
+        ),
+        Index("kfs_api_credential_tenant_space_status_idx", "tenant_id", "control_space_id", "status"),
+        Index("kfs_api_credential_tenant_prefix_idx", "tenant_id", "credential_prefix"),
+        sa.CheckConstraint("revision >= 0", name=sa.schema.conv("kfs_api_credential_revision_ck")),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    credential_hash: Mapped[str] = mapped_column(String(255), nullable=False)
+    credential_prefix: Mapped[str] = mapped_column(String(32), nullable=False)
+    credential_last4: Mapped[str] = mapped_column(String(4), nullable=False)
+    principal: Mapped[str] = mapped_column(String(255), nullable=False)
+    allowed_actions: Mapped[KnowledgeFSAllowedActions] = mapped_column(sa.JSON, nullable=False)
+    status: Mapped[KnowledgeFSApiCredentialStatus] = mapped_column(
+        EnumText(KnowledgeFSApiCredentialStatus, length=32),
+        nullable=False,
+        server_default=sa.text("'active'"),
+        default=KnowledgeFSApiCredentialStatus.ACTIVE,
+    )
+    revision: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+    expires_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    last_used_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    created_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    revoked_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    revoked_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    revoke_reason: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+
+
+class AppKnowledgeFSSpaceJoin(DefaultFieldsDCMixin, TypeBase):
+    """Explicit Agent or Workflow app authorization for a control-space."""
+
+    __tablename__ = "app_knowledge_fs_space_joins"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="app_kfs_space_join_pkey"),
+        UniqueConstraint(
+            "tenant_id",
+            "app_id",
+            "control_space_id",
+            "join_type",
+            name="app_kfs_space_join_identity_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="app_kfs_space_join_space_fk",
+            ondelete="RESTRICT",
+        ),
+        Index("app_kfs_space_join_app_status_idx", "tenant_id", "app_id", "status"),
+        sa.CheckConstraint("revision >= 0", name=sa.schema.conv("app_kfs_space_join_revision_ck")),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    app_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    join_type: Mapped[KnowledgeFSAppSpaceJoinType] = mapped_column(
+        EnumText(KnowledgeFSAppSpaceJoinType, length=32), nullable=False
+    )
+    status: Mapped[KnowledgeFSAppSpaceJoinStatus] = mapped_column(
+        EnumText(KnowledgeFSAppSpaceJoinStatus, length=32),
+        nullable=False,
+        server_default=sa.text("'active'"),
+        default=KnowledgeFSAppSpaceJoinStatus.ACTIVE,
+    )
+    revision: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+    created_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    revoked_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    revoked_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+
+
+class KnowledgeFSAuthorizationRevision(DefaultFieldsDCMixin, TypeBase):
+    """Monotonic authorization epochs included in future Capability claims."""
+
+    __tablename__ = "knowledge_fs_authorization_revisions"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_authorization_revision_pkey"),
+        UniqueConstraint("tenant_id", "control_space_id", name="kfs_authorization_revision_space_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_authorization_revision_space_fk",
+            ondelete="RESTRICT",
+        ),
+        sa.CheckConstraint(
+            "membership_epoch >= 0",
+            name=sa.schema.conv("kfs_authorization_membership_epoch_ck"),
+        ),
+        sa.CheckConstraint(
+            "space_acl_epoch >= 0",
+            name=sa.schema.conv("kfs_authorization_space_acl_epoch_ck"),
+        ),
+        sa.CheckConstraint(
+            "external_access_epoch >= 0",
+            name=sa.schema.conv("kfs_authorization_external_access_epoch_ck"),
+        ),
+        sa.CheckConstraint(
+            "content_policy_revision >= 0",
+            name=sa.schema.conv("kfs_authorization_content_policy_revision_ck"),
+        ),
+        sa.CheckConstraint(
+            "revoke_sequence >= 0",
+            name=sa.schema.conv("kfs_authorization_revoke_sequence_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    membership_epoch: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+    space_acl_epoch: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+    external_access_epoch: Mapped[int] = mapped_column(
+        sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0
+    )
+    content_policy_revision: Mapped[int] = mapped_column(
+        sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0
+    )
+    revoke_sequence: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+
+
+class KnowledgeFSCapabilityIssuanceAudit(DefaultFieldsDCMixin, TypeBase):
+    """Sanitized durable evidence for a Capability v2 issuance."""
+
+    __tablename__ = "knowledge_fs_capability_issuance_audits"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_capability_issuance_audit_pkey"),
+        UniqueConstraint("jti_hash", name="kfs_capability_issuance_audit_jti_hash_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_capability_issuance_audit_space_fk",
+            ondelete="RESTRICT",
+        ),
+        Index(
+            "kfs_capability_issuance_audit_space_created_idx",
+            "tenant_id",
+            "control_space_id",
+            "created_at",
+        ),
+        Index("kfs_capability_issuance_audit_trace_idx", "tenant_id", "trace_id"),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    trace_id: Mapped[str] = mapped_column(String(128), nullable=False)
+    jti_hash: Mapped[str] = mapped_column(String(80), nullable=False)
+    claims_summary: Mapped[KnowledgeFSCapabilityClaimsSummary] = mapped_column(sa.JSON, nullable=False)
+
+
+class KnowledgeFSCapabilityIssuanceReservation(DefaultFieldsDCMixin, TypeBase):
+    """Durable authorization fence written before signing starts.
+
+    The summary intentionally excludes bearer tokens, signing material, and raw JTI values.
+    Revocation producers scan these rows as well as completed issuance audits, closing the
+    authorization-to-signing race without holding a database transaction across signing.
+    """
+
+    __tablename__ = "knowledge_fs_capability_issuance_reservations"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_capability_issuance_reservation_pkey"),
+        UniqueConstraint(
+            "tenant_id",
+            "grant_id",
+            name="kfs_capability_issuance_reservation_grant_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_capability_issuance_reservation_space_fk",
+            ondelete="RESTRICT",
+        ),
+        Index(
+            "kfs_capability_issuance_reservation_subject_idx",
+            "tenant_id",
+            "control_space_id",
+            "subject",
+            "caller_kind",
+        ),
+        Index(
+            "kfs_capability_issuance_reservation_trace_idx",
+            "tenant_id",
+            "trace_id",
+        ),
+        sa.CheckConstraint(
+            "row_version >= 0",
+            name=sa.schema.conv("kfs_capability_issuance_reservation_version_ck"),
+        ),
+        sa.CheckConstraint(
+            "(status = 'reserved' AND issued_at IS NULL AND token_expires_at IS NULL "
+            "AND failed_at IS NULL AND failure_code IS NULL AND cleanup_after IS NULL) OR "
+            "(status = 'issued' AND issued_at IS NOT NULL AND token_expires_at IS NOT NULL "
+            "AND failed_at IS NULL AND failure_code IS NULL AND cleanup_after IS NOT NULL) OR "
+            "(status = 'failed' AND issued_at IS NULL AND token_expires_at IS NULL "
+            "AND failed_at IS NOT NULL AND failure_code IS NOT NULL AND cleanup_after IS NOT NULL)",
+            name=sa.schema.conv("kfs_capability_issuance_reservation_status_fields_ck"),
+        ),
+        sa.CheckConstraint(
+            "token_expires_at IS NULL OR cleanup_after >= token_expires_at",
+            name=sa.schema.conv("kfs_capability_issuance_reservation_cleanup_window_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    grant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    trace_id: Mapped[str] = mapped_column(String(128), nullable=False)
+    subject: Mapped[str] = mapped_column(String(255), nullable=False)
+    caller_kind: Mapped[str] = mapped_column(String(32), nullable=False)
+    request_summary: Mapped[KnowledgeFSCapabilityReservationSummary] = mapped_column(sa.JSON, nullable=False)
+    status: Mapped[KnowledgeFSCapabilityIssuanceReservationStatus] = mapped_column(
+        EnumText(KnowledgeFSCapabilityIssuanceReservationStatus, length=16),
+        nullable=False,
+        server_default=sa.text("'reserved'"),
+        default=KnowledgeFSCapabilityIssuanceReservationStatus.RESERVED,
+    )
+    issued_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    token_expires_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    failed_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    failure_code: Mapped[str | None] = mapped_column(String(128), nullable=True, default=None)
+    cleanup_after: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    row_version: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+
+
+class KnowledgeFSLifecycleOutbox(DefaultFieldsDCMixin, TypeBase):
+    """Durable command snapshot; it is never a product read-model source.
+
+    Provision payload writers must snapshot name, icon, description, slug,
+    model/profile intent, schema version, idempotency key, and expected
+    revision before enabling the later P1B dispatcher.
+    """
+
+    __tablename__ = "knowledge_fs_lifecycle_outbox"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_lifecycle_outbox_pkey"),
+        UniqueConstraint("tenant_id", "operation_id", name="kfs_lifecycle_outbox_operation_uq"),
+        UniqueConstraint("tenant_id", "idempotency_key", name="kfs_lifecycle_outbox_idempotency_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "control_space_id"],
+            ["knowledge_fs_control_spaces.tenant_id", "knowledge_fs_control_spaces.id"],
+            name="kfs_lifecycle_outbox_space_fk",
+            ondelete="RESTRICT",
+        ),
+        Index("kfs_lifecycle_outbox_dispatch_idx", "status", "next_attempt_at", "id"),
+        Index("kfs_lifecycle_outbox_space_created_idx", "tenant_id", "control_space_id", "created_at"),
+        sa.CheckConstraint(
+            "command_schema_version >= 1",
+            name=sa.schema.conv("kfs_lifecycle_outbox_schema_version_ck"),
+        ),
+        sa.CheckConstraint(
+            "expected_control_space_version >= 0",
+            name=sa.schema.conv("kfs_lifecycle_outbox_expected_version_ck"),
+        ),
+        sa.CheckConstraint(
+            "attempt_count >= 0",
+            name=sa.schema.conv("kfs_lifecycle_outbox_attempt_count_ck"),
+        ),
+        sa.CheckConstraint(
+            "status IN ('pending', 'processing', 'succeeded', 'retry', 'dead_letter')",
+            name=sa.schema.conv("kfs_lifecycle_outbox_status_ck"),
+        ),
+        sa.CheckConstraint(
+            "(status = 'processing' AND lease_owner IS NOT NULL AND lease_expires_at IS NOT NULL) "
+            "OR (status != 'processing' AND lease_owner IS NULL AND lease_expires_at IS NULL)",
+            name=sa.schema.conv("kfs_lifecycle_outbox_lease_state_ck"),
+        ),
+        sa.CheckConstraint(
+            "(status IN ('succeeded', 'dead_letter') AND completed_at IS NOT NULL) "
+            "OR (status NOT IN ('succeeded', 'dead_letter') AND completed_at IS NULL)",
+            name=sa.schema.conv("kfs_lifecycle_outbox_terminal_state_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    control_space_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    operation_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    idempotency_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    operation: Mapped[KnowledgeFSLifecycleOperation] = mapped_column(
+        EnumText(KnowledgeFSLifecycleOperation, length=32), nullable=False
+    )
+    command_payload: Mapped[KnowledgeFSCommandPayload] = mapped_column(sa.JSON, nullable=False)
+    expected_control_space_version: Mapped[int] = mapped_column(sa.BigInteger, nullable=False)
+    expected_knowledge_space_revision: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    command_schema_version: Mapped[int] = mapped_column(
+        sa.Integer, nullable=False, server_default=sa.text("1"), default=1
+    )
+    status: Mapped[KnowledgeFSLifecycleOutboxStatus] = mapped_column(
+        EnumText(KnowledgeFSLifecycleOutboxStatus, length=32),
+        nullable=False,
+        server_default=sa.text("'pending'"),
+        default=KnowledgeFSLifecycleOutboxStatus.PENDING,
+    )
+    attempt_count: Mapped[int] = mapped_column(sa.Integer, nullable=False, server_default=sa.text("0"), default=0)
+    next_attempt_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    last_attempt_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    lease_owner: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    lease_expires_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    last_error_code: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    last_error_message: Mapped[str | None] = mapped_column(LongText, nullable=True, default=None)
+    completed_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    retain_until: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+
+
+__all__ = [
+    "AppKnowledgeFSSpaceJoin",
+    "KnowledgeFSAllowedActions",
+    "KnowledgeFSApiCredential",
+    "KnowledgeFSApiCredentialStatus",
+    "KnowledgeFSAppSpaceJoinStatus",
+    "KnowledgeFSAppSpaceJoinType",
+    "KnowledgeFSAuthorizationRevision",
+    "KnowledgeFSCapabilityAuthzRevisionSummary",
+    "KnowledgeFSCapabilityClaimsSummary",
+    "KnowledgeFSCapabilityIssuanceAudit",
+    "KnowledgeFSCapabilityIssuanceReservation",
+    "KnowledgeFSCapabilityIssuanceReservationStatus",
+    "KnowledgeFSCapabilityReservationSummary",
+    "KnowledgeFSCommandEnvelope",
+    "KnowledgeFSCommandPayload",
+    "KnowledgeFSControlSpace",
+    "KnowledgeFSControlSpacePermission",
+    "KnowledgeFSControlSpacePermissionRole",
+    "KnowledgeFSControlSpacePermissionStatus",
+    "KnowledgeFSControlSpaceState",
+    "KnowledgeFSControlSpaceVisibility",
+    "KnowledgeFSDeleteCommandPayload",
+    "KnowledgeFSExternalAccessPolicy",
+    "KnowledgeFSLifecycleOperation",
+    "KnowledgeFSLifecycleOutbox",
+    "KnowledgeFSLifecycleOutboxStatus",
+    "KnowledgeFSMetadataUpdateCommandPayload",
+    "KnowledgeFSModelSelectionIntentPayload",
+    "KnowledgeFSProvisionCommandPayload",
+    "KnowledgeFSRepairCommandPayload",
+    "KnowledgeFSRerankIntentPayload",
+    "KnowledgeFSRetrievalProfileIntentPayload",
+    "KnowledgeFSRevokeCommandPayload",
+    "KnowledgeFSScoreThresholdIntentPayload",
+]

api/models/knowledge_fs_cleanup.py

@@ -0,0 +1,130 @@
+"""Persistent authorization for the irreversible P9 KnowledgeFS cleanup window."""
+
+from __future__ import annotations
+
+from datetime import datetime
+from enum import StrEnum
+
+import sqlalchemy as sa
+from sqlalchemy import DateTime, Index, String, UniqueConstraint
+from sqlalchemy.orm import Mapped, mapped_column
+
+from .base import DefaultFieldsDCMixin, TypeBase
+from .types import EnumText, StringUUID
+
+
+class KnowledgeFSCleanupAuthorizationStatus(StrEnum):
+    REQUESTED = "requested"
+    APPROVED = "approved"
+    STARTED = "started"
+    COMPLETED = "completed"
+
+
+class KnowledgeFSCleanupTarget(StrEnum):
+    LEGACY_SNAPSHOT_FOREIGN_KEYS = "legacy_snapshot_foreign_keys"
+    LEGACY_ACL_ROUTES = "legacy_acl_routes"
+    LEGACY_ACL_SCHEMA = "legacy_acl_schema"
+    LEGACY_API_KEY_SCHEMA = "legacy_api_key_schema"
+    LEGACY_V1_AUTH = "legacy_v1_auth"
+    RAW_LIST_CREATE_PROXY = "raw_list_create_proxy"
+
+
+class KnowledgeFSCleanupAuthorization(DefaultFieldsDCMixin, TypeBase):
+    """Four-eyes approval and immutable evidence for one cleanup start fence."""
+
+    __tablename__ = "knowledge_fs_cleanup_authorizations"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_cleanup_authorization_pkey"),
+        UniqueConstraint("tenant_id", "ledger_id", "request_id", name="kfs_cleanup_authorization_request_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_cleanup_authorization_ledger_fk",
+        ),
+        Index("kfs_cleanup_authorization_status_idx", "tenant_id", "status", "updated_at"),
+        sa.CheckConstraint("row_version >= 0", name=sa.schema.conv("kfs_cleanup_authorization_version_ck")),
+        sa.CheckConstraint(
+            "readiness_ledger_cas_version >= 0 "
+            "AND (approved_ledger_cas_version IS NULL OR approved_ledger_cas_version >= 0) "
+            "AND (started_ledger_cas_version IS NULL OR started_ledger_cas_version >= 0) "
+            "AND (completed_ledger_cas_version IS NULL OR completed_ledger_cas_version >= 0)",
+            name=sa.schema.conv("kfs_cleanup_authorization_ledger_versions_ck"),
+        ),
+        sa.CheckConstraint(
+            "(status = 'requested' "
+            "AND approved_by_account_id IS NULL AND approved_at IS NULL "
+            "AND approval_expires_at IS NULL AND approved_ledger_cas_version IS NULL "
+            "AND started_by_account_id IS NULL AND started_at IS NULL "
+            "AND started_ledger_cas_version IS NULL "
+            "AND completed_by_account_id IS NULL AND completed_at IS NULL "
+            "AND completion_evidence IS NULL AND completed_ledger_cas_version IS NULL) OR "
+            "(status = 'approved' "
+            "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+            "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+            "AND started_by_account_id IS NULL AND started_at IS NULL "
+            "AND started_ledger_cas_version IS NULL "
+            "AND completed_by_account_id IS NULL AND completed_at IS NULL "
+            "AND completion_evidence IS NULL AND completed_ledger_cas_version IS NULL) OR "
+            "(status = 'started' "
+            "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+            "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+            "AND started_by_account_id IS NOT NULL AND started_at IS NOT NULL "
+            "AND started_ledger_cas_version IS NOT NULL "
+            "AND completed_by_account_id IS NULL AND completed_at IS NULL "
+            "AND completion_evidence IS NULL AND completed_ledger_cas_version IS NULL) OR "
+            "(status = 'completed' "
+            "AND approved_by_account_id IS NOT NULL AND approved_at IS NOT NULL "
+            "AND approval_expires_at IS NOT NULL AND approved_ledger_cas_version IS NOT NULL "
+            "AND started_by_account_id IS NOT NULL AND started_at IS NOT NULL "
+            "AND started_ledger_cas_version IS NOT NULL "
+            "AND completed_by_account_id IS NOT NULL AND completed_at IS NOT NULL "
+            "AND completion_evidence IS NOT NULL AND completed_ledger_cas_version IS NOT NULL)",
+            name=sa.schema.conv("kfs_cleanup_authorization_status_fields_ck"),
+        ),
+        sa.CheckConstraint(
+            "approval_expires_at IS NULL OR approval_expires_at > approved_at",
+            name=sa.schema.conv("kfs_cleanup_authorization_approval_window_ck"),
+        ),
+        sa.CheckConstraint(
+            "completed_at IS NULL OR completed_at >= started_at",
+            name=sa.schema.conv("kfs_cleanup_authorization_completion_time_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    ledger_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    request_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    plan_schema_version: Mapped[str] = mapped_column(String(32), nullable=False)
+    plan_digest: Mapped[str] = mapped_column(String(71), nullable=False)
+    targets: Mapped[list[str]] = mapped_column(sa.JSON, nullable=False)
+    readiness_evidence: Mapped[dict[str, object]] = mapped_column(sa.JSON, nullable=False)
+    requested_by_account_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    requested_at: Mapped[datetime] = mapped_column(DateTime, nullable=False)
+    readiness_ledger_cas_version: Mapped[int] = mapped_column(sa.BigInteger, nullable=False)
+    status: Mapped[KnowledgeFSCleanupAuthorizationStatus] = mapped_column(
+        EnumText(KnowledgeFSCleanupAuthorizationStatus, length=16),
+        nullable=False,
+        server_default=sa.text("'requested'"),
+        default=KnowledgeFSCleanupAuthorizationStatus.REQUESTED,
+    )
+    approved_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    approved_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    approval_expires_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    approved_ledger_cas_version: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    started_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    started_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    started_ledger_cas_version: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    completed_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    completed_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    completion_evidence: Mapped[dict[str, object] | None] = mapped_column(
+        sa.JSON(none_as_null=True), nullable=True, default=None
+    )
+    completed_ledger_cas_version: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    row_version: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+
+
+__all__ = [
+    "KnowledgeFSCleanupAuthorization",
+    "KnowledgeFSCleanupAuthorizationStatus",
+    "KnowledgeFSCleanupTarget",
+]

api/models/knowledge_fs_cutover.py

@@ -0,0 +1,507 @@
+"""Workspace-scoped KnowledgeFS migration and cutover persistence.
+
+The ledger is Dify control-plane state only. It deliberately stores opaque
+KnowledgeFS identifiers and authorization watermarks without foreign keys or
+queries to legacy product tables. Cutover feature switches live on one
+row so the repository can change them with a single CAS statement. Quarantine
+remediation is separately CAS-versioned and retains immutable operator evidence.
+"""
+
+from __future__ import annotations
+
+from datetime import datetime
+from enum import StrEnum
+from typing import TypedDict
+
+import sqlalchemy as sa
+from sqlalchemy import DateTime, Index, String, UniqueConstraint
+from sqlalchemy.orm import Mapped, mapped_column
+
+from .base import DefaultFieldsDCMixin, TypeBase
+from .types import EnumText, LongText, StringUUID
+
+
+class KnowledgeFSCutoverRevisionWatermark(TypedDict):
+    membership_epoch: int
+    space_acl_epoch: int
+    external_access_epoch: int
+    content_policy_revision: int
+
+
+class KnowledgeFSCutoverSmokeChecks(TypedDict):
+    authorization: bool
+    list_spaces: bool
+    create_space: bool
+    query: bool
+    upload: bool
+    stream: bool
+    deletion: bool
+
+
+class KnowledgeFSCutoverSmokeEvidenceReferences(TypedDict):
+    authorization: str
+    list_spaces: str
+    create_space: str
+    query: str
+    upload: str
+    stream: str
+    deletion: str
+
+
+class KnowledgeFSCutoverSmokeResults(TypedDict):
+    schema_version: str
+    tenant_id: str
+    environment: str
+    operator: str
+    operator_account_id: str
+    observed_at: str
+    checks: KnowledgeFSCutoverSmokeChecks
+    evidence_references: KnowledgeFSCutoverSmokeEvidenceReferences
+
+
+def knowledge_fs_cutover_smoke_results_passed(value: KnowledgeFSCutoverSmokeResults | None) -> bool:
+    """Reject incomplete or pre-production JSON evidence instead of treating metadata as booleans."""
+
+    if not isinstance(value, dict):
+        return False
+    checks = value.get("checks")
+    references = value.get("evidence_references")
+    expected = {
+        "authorization",
+        "list_spaces",
+        "create_space",
+        "query",
+        "upload",
+        "stream",
+        "deletion",
+    }
+    return bool(
+        value.get("schema_version") == "knowledge-fs-p8-cutover-smoke/v1"
+        and value.get("environment") == "production"
+        and isinstance(checks, dict)
+        and set(checks) == expected
+        and all(checks.get(name) is True for name in expected)
+        and isinstance(references, dict)
+        and set(references) == expected
+        and all(isinstance(reference := references.get(name), str) and bool(reference) for name in expected)
+    )
+
+
+class KnowledgeFSWorkspaceCutoverPhase(StrEnum):
+    INVENTORY = "inventory"
+    BACKFILL = "backfill"
+    SHADOW = "shadow"
+    FROZEN = "frozen"
+    CUTOVER = "cutover"
+    OBSERVING = "observing"
+    READY_FOR_CLEANUP = "ready_for_cleanup"
+
+
+class KnowledgeFSMigrationIssueStatus(StrEnum):
+    OPEN = "open"
+    APPROVED_FAIL_CLOSED = "approved_fail_closed"
+    RESOLVED = "resolved"
+
+
+class KnowledgeFSMigrationIssueKind(StrEnum):
+    REGISTRATION_CONFLICT = "registration_conflict"
+    UNRESOLVED_SUBJECT = "unresolved_subject"
+    UNKNOWN_EXTERNAL_ACCESS = "unknown_external_access"
+    REVISION_DRIFT = "revision_drift"
+    TASK_QUARANTINE = "task_quarantine"
+    LEGACY_SNAPSHOT_DEPENDENCY = "legacy_snapshot_dependency"
+    LEGACY_FOREIGN_KEY_DEPENDENCY = "legacy_foreign_key_dependency"
+
+
+class KnowledgeFSMigrationQuarantineKind(StrEnum):
+    CONTROL_SPACE = "control_space"
+    SUBJECT = "subject"
+    TASK = "task"
+    LEGACY_API_KEY = "legacy_api_key"
+    ORPHAN_RESOURCE = "orphan_resource"
+
+
+class KnowledgeFSMigrationQuarantineDisposition(StrEnum):
+    PENDING = "pending"
+    MIGRATABLE = "migratable"
+    WAIT_FOR_COMPLETION = "wait_for_completion"
+    CANCEL = "cancel"
+    ISOLATE = "isolate"
+    ROTATE_CREDENTIAL = "rotate_credential"
+    RESOLVED = "resolved"
+
+
+class KnowledgeFSShadowAuthorizationDecision(StrEnum):
+    MATCH = "match"
+    TIGHTENED = "tightened"
+    EXPANDED = "expanded"
+    UNKNOWN = "unknown"
+
+
+class KnowledgeFSWorkspaceCutoverLedger(DefaultFieldsDCMixin, TypeBase):
+    """Single-Workspace migration state, watermarks, switches, and time fences."""
+
+    __tablename__ = "knowledge_fs_workspace_cutover_ledgers"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_workspace_cutover_ledger_pkey"),
+        UniqueConstraint("tenant_id", name="kfs_workspace_cutover_ledger_tenant_uq"),
+        UniqueConstraint("tenant_id", "id", name="kfs_workspace_cutover_ledger_tenant_id_uq"),
+        Index("kfs_workspace_cutover_ledger_phase_updated_idx", "phase", "updated_at"),
+        sa.CheckConstraint("cas_version >= 0", name=sa.schema.conv("kfs_workspace_cutover_cas_version_ck")),
+        sa.CheckConstraint(
+            "source_task_watermark >= 0 AND applied_task_watermark >= 0 "
+            "AND (final_task_watermark IS NULL OR final_task_watermark >= 0)",
+            name=sa.schema.conv("kfs_workspace_cutover_task_watermark_ck"),
+        ),
+        sa.CheckConstraint(
+            "cutover_at IS NULL OR freeze_at IS NOT NULL",
+            name=sa.schema.conv("kfs_workspace_cutover_freeze_before_cutover_ck"),
+        ),
+        sa.CheckConstraint(
+            "rollback_cutoff_at IS NULL OR cutover_at IS NOT NULL",
+            name=sa.schema.conv("kfs_workspace_cutover_rollback_cutoff_ck"),
+        ),
+        sa.CheckConstraint(
+            "shadow_observation_count >= 0",
+            name=sa.schema.conv("kfs_workspace_cutover_shadow_count_ck"),
+        ),
+        sa.CheckConstraint(
+            "shadow_window_ended_at IS NULL OR shadow_window_started_at IS NOT NULL",
+            name=sa.schema.conv("kfs_workspace_cutover_shadow_window_ck"),
+        ),
+        sa.CheckConstraint(
+            "(shadow_completed_at IS NULL "
+            "AND shadow_evidence_digest IS NULL AND shadow_producer IS NULL "
+            "AND shadow_completed_by_operator IS NULL AND shadow_completed_by_account_id IS NULL "
+            "AND shadow_window_started_at IS NULL AND shadow_window_ended_at IS NULL "
+            "AND shadow_traffic_zero = false AND shadow_traffic_zero_evidence IS NULL "
+            "AND shadow_latest_observed_revision IS NULL) OR "
+            "(shadow_completed_at IS NOT NULL AND shadow_started_at IS NOT NULL "
+            "AND shadow_evidence_digest IS NOT NULL AND shadow_producer IS NOT NULL "
+            "AND shadow_completed_by_operator IS NOT NULL AND shadow_completed_by_account_id IS NOT NULL "
+            "AND ((shadow_traffic_zero = true AND shadow_observation_count = 0 "
+            "AND shadow_traffic_zero_evidence IS NOT NULL AND shadow_window_started_at IS NULL "
+            "AND shadow_window_ended_at IS NULL AND shadow_latest_observed_revision IS NULL) OR "
+            "(shadow_traffic_zero = false AND shadow_observation_count > 0 "
+            "AND shadow_traffic_zero_evidence IS NULL AND shadow_window_started_at IS NOT NULL "
+            "AND shadow_window_ended_at IS NOT NULL AND shadow_latest_observed_revision IS NOT NULL)))",
+            name=sa.schema.conv("kfs_workspace_cutover_shadow_completion_fields_ck"),
+        ),
+        sa.CheckConstraint(
+            "(remote_freeze_id IS NULL AND remote_freeze_revision IS NULL "
+            "AND remote_freeze_digest IS NULL AND remote_freeze_task_watermark IS NULL "
+            "AND remote_freeze_control_space_id IS NULL AND remote_freeze_frozen_at IS NULL "
+            "AND remote_freeze_updated_at IS NULL AND remote_freeze_acknowledged_at IS NULL "
+            "AND remote_freeze_applied IS NULL AND remote_freeze_replayed IS NULL) OR "
+            "(remote_freeze_id IS NOT NULL AND remote_freeze_revision BETWEEN 1 AND 9007199254740991 "
+            "AND remote_freeze_digest IS NOT NULL AND remote_freeze_task_watermark >= 0 "
+            "AND remote_freeze_control_space_id IS NOT NULL AND remote_freeze_frozen_at IS NOT NULL "
+            "AND remote_freeze_updated_at IS NOT NULL AND remote_freeze_acknowledged_at IS NOT NULL "
+            "AND ((remote_freeze_applied = true AND remote_freeze_replayed = false) "
+            "OR (remote_freeze_applied = false AND remote_freeze_replayed = true)))",
+            name=sa.schema.conv("kfs_workspace_cutover_remote_freeze_fields_ck"),
+        ),
+        sa.CheckConstraint(
+            "remote_freeze_updated_at IS NULL OR remote_freeze_updated_at >= remote_freeze_frozen_at",
+            name=sa.schema.conv("kfs_workspace_cutover_remote_freeze_time_ck"),
+        ),
+        sa.CheckConstraint(
+            "(remote_activation_id IS NULL AND remote_activation_revision IS NULL "
+            "AND remote_activation_digest IS NULL AND remote_activation_control_space_id IS NULL "
+            "AND remote_activation_activated_at IS NULL AND remote_activation_updated_at IS NULL "
+            "AND remote_activation_acknowledged_at IS NULL AND remote_activation_applied IS NULL "
+            "AND remote_activation_replayed IS NULL) OR "
+            "(remote_activation_id IS NOT NULL AND remote_activation_revision BETWEEN 1 AND 9007199254740991 "
+            "AND remote_activation_digest IS NOT NULL AND remote_activation_control_space_id IS NOT NULL "
+            "AND remote_activation_activated_at IS NOT NULL AND remote_activation_updated_at IS NOT NULL "
+            "AND remote_activation_acknowledged_at IS NOT NULL "
+            "AND ((remote_activation_applied = true AND remote_activation_replayed = false) "
+            "OR (remote_activation_applied = false AND remote_activation_replayed = true)))",
+            name=sa.schema.conv("kfs_workspace_cutover_remote_activation_fields_ck"),
+        ),
+        sa.CheckConstraint(
+            "remote_activation_updated_at IS NULL OR remote_activation_updated_at >= remote_activation_activated_at",
+            name=sa.schema.conv("kfs_workspace_cutover_remote_activation_time_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    source_revision_watermark: Mapped[KnowledgeFSCutoverRevisionWatermark] = mapped_column(sa.JSON, nullable=False)
+    applied_revision_watermark: Mapped[KnowledgeFSCutoverRevisionWatermark] = mapped_column(sa.JSON, nullable=False)
+    phase: Mapped[KnowledgeFSWorkspaceCutoverPhase] = mapped_column(
+        EnumText(KnowledgeFSWorkspaceCutoverPhase, length=32),
+        nullable=False,
+        server_default=sa.text("'inventory'"),
+        default=KnowledgeFSWorkspaceCutoverPhase.INVENTORY,
+    )
+    final_revision_watermark: Mapped[KnowledgeFSCutoverRevisionWatermark | None] = mapped_column(
+        sa.JSON, nullable=True, default=None
+    )
+    source_task_watermark: Mapped[int] = mapped_column(
+        sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0
+    )
+    final_task_watermark: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    applied_task_watermark: Mapped[int] = mapped_column(
+        sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0
+    )
+    shadow_started_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    shadow_completed_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    shadow_evidence_digest: Mapped[str | None] = mapped_column(String(71), nullable=True, default=None)
+    shadow_observation_count: Mapped[int] = mapped_column(
+        sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0
+    )
+    shadow_window_started_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    shadow_window_ended_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    shadow_traffic_zero: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    shadow_traffic_zero_evidence: Mapped[dict[str, object] | None] = mapped_column(
+        sa.JSON(none_as_null=True), nullable=True, default=None
+    )
+    shadow_latest_observed_revision: Mapped[KnowledgeFSCutoverRevisionWatermark | None] = mapped_column(
+        sa.JSON(none_as_null=True), nullable=True, default=None
+    )
+    shadow_producer: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    shadow_completed_by_operator: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    shadow_completed_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    remote_freeze_id: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    remote_freeze_revision: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    remote_freeze_digest: Mapped[str | None] = mapped_column(String(71), nullable=True, default=None)
+    remote_freeze_task_watermark: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    remote_freeze_control_space_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    remote_freeze_frozen_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    remote_freeze_updated_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    remote_freeze_acknowledged_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    remote_freeze_applied: Mapped[bool | None] = mapped_column(sa.Boolean, nullable=True, default=None)
+    remote_freeze_replayed: Mapped[bool | None] = mapped_column(sa.Boolean, nullable=True, default=None)
+    remote_activation_id: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    remote_activation_revision: Mapped[int | None] = mapped_column(sa.BigInteger, nullable=True, default=None)
+    remote_activation_digest: Mapped[str | None] = mapped_column(String(71), nullable=True, default=None)
+    remote_activation_control_space_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    remote_activation_activated_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    remote_activation_updated_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    remote_activation_acknowledged_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    remote_activation_applied: Mapped[bool | None] = mapped_column(sa.Boolean, nullable=True, default=None)
+    remote_activation_replayed: Mapped[bool | None] = mapped_column(sa.Boolean, nullable=True, default=None)
+    freeze_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    cutover_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    rolled_back_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    rollback_cutoff_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    observation_started_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    observation_window_ends_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    observation_completed_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    maximum_task_expires_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    irreversible_cleanup_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    product_routes_enabled: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    capability_v2_enabled: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    integrated_mode_enabled: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    legacy_acl_read_only: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    smoke_results: Mapped[KnowledgeFSCutoverSmokeResults | None] = mapped_column(sa.JSON, nullable=True, default=None)
+    legacy_dependency_report: Mapped[list[dict[str, object]] | None] = mapped_column(
+        sa.JSON, nullable=True, default=None
+    )
+    legacy_dependency_checked_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    legacy_dependency_ready: Mapped[bool] = mapped_column(
+        sa.Boolean, nullable=False, server_default=sa.text("false"), default=False
+    )
+    cas_version: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+
+
+class KnowledgeFSMigrationIssue(DefaultFieldsDCMixin, TypeBase):
+    """Auditable blocker; unknown access can only be approved as fail-closed."""
+
+    __tablename__ = "knowledge_fs_migration_issues"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_migration_issue_pkey"),
+        UniqueConstraint("tenant_id", "ledger_id", "issue_key", name="kfs_migration_issue_key_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_migration_issue_ledger_fk",
+            ondelete="CASCADE",
+        ),
+        Index("kfs_migration_issue_gate_idx", "tenant_id", "ledger_id", "status", "kind"),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    ledger_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    issue_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    kind: Mapped[KnowledgeFSMigrationIssueKind] = mapped_column(
+        EnumText(KnowledgeFSMigrationIssueKind, length=48), nullable=False
+    )
+    status: Mapped[KnowledgeFSMigrationIssueStatus] = mapped_column(
+        EnumText(KnowledgeFSMigrationIssueStatus, length=32),
+        nullable=False,
+        server_default=sa.text("'open'"),
+        default=KnowledgeFSMigrationIssueStatus.OPEN,
+    )
+    resource_type: Mapped[str | None] = mapped_column(String(64), nullable=True, default=None)
+    resource_id: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    details: Mapped[dict[str, object]] = mapped_column(sa.JSON, nullable=False, default_factory=dict)
+    approved_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    approved_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    resolved_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    resolved_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+
+
+class KnowledgeFSMigrationQuarantine(DefaultFieldsDCMixin, TypeBase):
+    """Fail-closed inventory item whose operator resolution is immutable and CAS-versioned."""
+
+    __tablename__ = "knowledge_fs_migration_quarantine"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_migration_quarantine_pkey"),
+        UniqueConstraint(
+            "tenant_id",
+            "ledger_id",
+            "source_kind",
+            "source_id",
+            name="kfs_migration_quarantine_source_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_migration_quarantine_ledger_fk",
+            ondelete="CASCADE",
+        ),
+        Index("kfs_migration_quarantine_disposition_idx", "tenant_id", "ledger_id", "disposition"),
+        sa.CheckConstraint("row_version >= 0", name=sa.schema.conv("kfs_migration_quarantine_version_ck")),
+        sa.CheckConstraint(
+            "(disposition = 'resolved' "
+            "AND resolved_by_operator IS NOT NULL AND resolved_by_account_id IS NOT NULL "
+            "AND evidence IS NOT NULL AND resolved_at IS NOT NULL) OR "
+            "(disposition <> 'resolved' "
+            "AND resolved_by_operator IS NULL AND resolved_by_account_id IS NULL "
+            "AND evidence IS NULL AND resolved_at IS NULL)",
+            name=sa.schema.conv("kfs_migration_quarantine_resolution_fields_ck"),
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    ledger_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    source_kind: Mapped[KnowledgeFSMigrationQuarantineKind] = mapped_column(
+        EnumText(KnowledgeFSMigrationQuarantineKind, length=32), nullable=False
+    )
+    source_id: Mapped[str] = mapped_column(String(255), nullable=False)
+    reason_code: Mapped[str] = mapped_column(String(64), nullable=False)
+    disposition: Mapped[KnowledgeFSMigrationQuarantineDisposition] = mapped_column(
+        EnumText(KnowledgeFSMigrationQuarantineDisposition, length=32), nullable=False
+    )
+    details: Mapped[dict[str, object]] = mapped_column(sa.JSON, nullable=False, default_factory=dict)
+    resolved_by_operator: Mapped[str | None] = mapped_column(String(255), nullable=True, default=None)
+    resolved_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    evidence: Mapped[dict[str, object] | None] = mapped_column(sa.JSON(none_as_null=True), nullable=True, default=None)
+    resolved_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    row_version: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+
+
+class KnowledgeFSShadowAuthorizationDiff(DefaultFieldsDCMixin, TypeBase):
+    """Current disposition for a diff key; append-only observations retain its history."""
+
+    __tablename__ = "knowledge_fs_shadow_authorization_diffs"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_shadow_authorization_diff_pkey"),
+        UniqueConstraint("tenant_id", "ledger_id", "diff_key", name="kfs_shadow_authorization_diff_key_uq"),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_shadow_authorization_diff_ledger_fk",
+            ondelete="CASCADE",
+        ),
+        Index("kfs_shadow_authorization_diff_gate_idx", "tenant_id", "ledger_id", "status", "decision"),
+        sa.CheckConstraint("row_version >= 0", name=sa.schema.conv("kfs_shadow_authorization_diff_version_ck")),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    ledger_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    diff_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    principal: Mapped[str] = mapped_column(String(255), nullable=False)
+    dify_allowed: Mapped[bool] = mapped_column(sa.Boolean, nullable=False)
+    decision: Mapped[KnowledgeFSShadowAuthorizationDecision] = mapped_column(
+        EnumText(KnowledgeFSShadowAuthorizationDecision, length=32), nullable=False
+    )
+    reason: Mapped[str] = mapped_column(LongText, nullable=False)
+    observed_revision: Mapped[KnowledgeFSCutoverRevisionWatermark] = mapped_column(sa.JSON, nullable=False)
+    status: Mapped[KnowledgeFSMigrationIssueStatus] = mapped_column(
+        EnumText(KnowledgeFSMigrationIssueStatus, length=32), nullable=False
+    )
+    current_evidence_digest: Mapped[str] = mapped_column(String(71), nullable=False)
+    last_observed_at: Mapped[datetime] = mapped_column(DateTime, nullable=False)
+    control_space_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    legacy_allowed: Mapped[bool | None] = mapped_column(sa.Boolean, nullable=True, default=None)
+    approved_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    approved_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    resolved_by_account_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    resolved_at: Mapped[datetime | None] = mapped_column(DateTime, nullable=True, default=None)
+    row_version: Mapped[int] = mapped_column(sa.BigInteger, nullable=False, server_default=sa.text("0"), default=0)
+
+
+class KnowledgeFSShadowAuthorizationObservation(DefaultFieldsDCMixin, TypeBase):
+    """Append-only producer evidence used to derive a shadow diff disposition."""
+
+    __tablename__ = "knowledge_fs_shadow_authorization_observations"
+    __table_args__ = (
+        sa.PrimaryKeyConstraint("id", name="kfs_shadow_authorization_observation_pkey"),
+        UniqueConstraint(
+            "tenant_id",
+            "ledger_id",
+            "diff_key",
+            "evidence_digest",
+            name="kfs_shadow_authorization_observation_evidence_uq",
+        ),
+        sa.ForeignKeyConstraint(
+            ["tenant_id", "ledger_id"],
+            ["knowledge_fs_workspace_cutover_ledgers.tenant_id", "knowledge_fs_workspace_cutover_ledgers.id"],
+            name="kfs_shadow_authorization_observation_ledger_fk",
+            ondelete="CASCADE",
+        ),
+        Index(
+            "kfs_shadow_authorization_observation_window_idx",
+            "tenant_id",
+            "ledger_id",
+            "observed_at",
+        ),
+    )
+
+    tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    ledger_id: Mapped[str] = mapped_column(StringUUID, nullable=False)
+    diff_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    producer: Mapped[str] = mapped_column(String(255), nullable=False)
+    principal: Mapped[str] = mapped_column(String(255), nullable=False)
+    dify_allowed: Mapped[bool] = mapped_column(sa.Boolean, nullable=False)
+    decision: Mapped[KnowledgeFSShadowAuthorizationDecision] = mapped_column(
+        EnumText(KnowledgeFSShadowAuthorizationDecision, length=32), nullable=False
+    )
+    reason: Mapped[str] = mapped_column(LongText, nullable=False)
+    observed_revision: Mapped[KnowledgeFSCutoverRevisionWatermark] = mapped_column(sa.JSON, nullable=False)
+    observed_at: Mapped[datetime] = mapped_column(DateTime, nullable=False)
+    evidence_digest: Mapped[str] = mapped_column(String(71), nullable=False)
+    control_space_id: Mapped[str | None] = mapped_column(StringUUID, nullable=True, default=None)
+    legacy_allowed: Mapped[bool | None] = mapped_column(sa.Boolean, nullable=True, default=None)
+
+
+__all__ = [
+    "KnowledgeFSCutoverRevisionWatermark",
+    "KnowledgeFSCutoverSmokeChecks",
+    "KnowledgeFSCutoverSmokeEvidenceReferences",
+    "KnowledgeFSCutoverSmokeResults",
+    "KnowledgeFSMigrationIssue",
+    "KnowledgeFSMigrationIssueKind",
+    "KnowledgeFSMigrationIssueStatus",
+    "KnowledgeFSMigrationQuarantine",
+    "KnowledgeFSMigrationQuarantineDisposition",
+    "KnowledgeFSMigrationQuarantineKind",
+    "KnowledgeFSShadowAuthorizationDecision",
+    "KnowledgeFSShadowAuthorizationDiff",
+    "KnowledgeFSShadowAuthorizationObservation",
+    "KnowledgeFSWorkspaceCutoverLedger",
+    "KnowledgeFSWorkspaceCutoverPhase",
+    "knowledge_fs_cutover_smoke_results_passed",
+]

api/models/workflow.py

@@ -1476,6 +1476,8 @@ class WorkflowRunArchiveBundle(DefaultFieldsDCMixin, TypeBase):
             name="workflow_run_archive_bundle_identity_uq",
         ),
         sa.Index("workflow_run_archive_bundle_tenant_month_idx", "tenant_id", "year", "month"),
+        sa.Index("workflow_run_archive_bundle_month_id_idx", "year", "month", "id"),
+        sa.Index("workflow_run_archive_bundle_month_shard_id_idx", "year", "month", "shard", "id"),
     )
 
     tenant_id: Mapped[str] = mapped_column(StringUUID, nullable=False)

api/openapi/markdown/console-openapi.md

@@ -9639,6 +9639,27 @@ Bedrock retrieval test (internal use only)
 | ---- | ----------- | ------ |
 | 200 | Success | **application/json**: [TrialDatasetListResponse](#trialdatasetlistresponse)<br> |
 
+### [POST] /trial-apps/{app_id}/files/upload
+**Upload a file into the tenant that owns the trial app**
+
+#### Parameters
+
+| Name | Located in | Description | Required | Schema |
+| ---- | ---------- | ----------- | -------- | ------ |
+| app_id | path |  | Yes | string (uuid) |
+
+#### Request Body
+
+| Required | Schema |
+| -------- | ------ |
+|  Yes | **multipart/form-data**: { **"file"**: binary, **"source"**: string, <br>**Available values:** "datasets" }<br> |
+
+#### Responses
+
+| Code | Description | Schema |
+| ---- | ----------- | ------ |
+| 201 | File uploaded successfully | **application/json**: [FileResponse](#fileresponse)<br> |
+
 ### [GET] /trial-apps/{app_id}/messages/{message_id}/suggested-questions
 #### Parameters
 
@@ -9668,6 +9689,27 @@ Bedrock retrieval test (internal use only)
 | ---- | ----------- | ------ |
 | 200 | Success | **application/json**: [Parameters](#parameters)<br> |
 
+### [POST] /trial-apps/{app_id}/remote-files/upload
+**Upload a remote file into the tenant that owns the trial app**
+
+#### Parameters
+
+| Name | Located in | Description | Required | Schema |
+| ---- | ---------- | ----------- | -------- | ------ |
+| app_id | path |  | Yes | string (uuid) |
+
+#### Request Body
+
+| Required | Schema |
+| -------- | ------ |
+|  Yes | **application/json**: [RemoteFileUploadPayload](#remotefileuploadpayload)<br> |
+
+#### Responses
+
+| Code | Description | Schema |
+| ---- | ----------- | ------ |
+| 201 | File uploaded successfully | **application/json**: [FileWithSignedUrl](#filewithsignedurl)<br> |
+
 ### [GET] /trial-apps/{app_id}/site
 **Retrieve app site info**
 
@@ -21859,6 +21901,7 @@ The subscription constructor of the trigger provider
 
 | Name | Type | Description | Required |
 | ---- | ---- | ----------- | -------- |
+| _is_collaborative | boolean |  | No |
 | conversation_variables | [ object ] |  | No |
 | environment_variables | [ object ] |  | No |
 | features | object |  | Yes |

api/repositories/knowledge_fs_cleanup_repository.py

@@ -0,0 +1,53 @@
+"""Persistence contract for irreversible KnowledgeFS cleanup authorization."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from datetime import datetime
+from typing import Protocol
+
+from models.knowledge_fs_cleanup import (
+    KnowledgeFSCleanupAuthorization,
+    KnowledgeFSCleanupAuthorizationStatus,
+)
+
+
+@dataclass(frozen=True, slots=True)
+class KnowledgeFSCleanupAuthorizationCASUpdate:
+    tenant_id: str
+    ledger_id: str
+    request_id: str
+    expected_status: KnowledgeFSCleanupAuthorizationStatus
+    expected_row_version: int
+    new_status: KnowledgeFSCleanupAuthorizationStatus
+    approved_by_account_id: str | None = None
+    approved_at: datetime | None = None
+    approval_expires_at: datetime | None = None
+    approved_ledger_cas_version: int | None = None
+    started_by_account_id: str | None = None
+    started_at: datetime | None = None
+    started_ledger_cas_version: int | None = None
+    completed_by_account_id: str | None = None
+    completed_at: datetime | None = None
+    completion_evidence: dict[str, object] | None = None
+    completed_ledger_cas_version: int | None = None
+
+
+class KnowledgeFSCleanupAuthorizationRepository(Protocol):
+    def add(self, authorization: KnowledgeFSCleanupAuthorization) -> KnowledgeFSCleanupAuthorization: ...
+
+    def get(
+        self,
+        *,
+        tenant_id: str,
+        ledger_id: str,
+        request_id: str,
+    ) -> KnowledgeFSCleanupAuthorization | None: ...
+
+    def compare_and_set(self, update_values: KnowledgeFSCleanupAuthorizationCASUpdate) -> bool: ...
+
+
+__all__ = [
+    "KnowledgeFSCleanupAuthorizationCASUpdate",
+    "KnowledgeFSCleanupAuthorizationRepository",
+]

api/repositories/knowledge_fs_control_space_repository.py

@@ -0,0 +1,60 @@
+"""Persistence boundary for tenant-scoped KnowledgeFS lifecycle operations."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from datetime import datetime
+from typing import Protocol
+
+from models.knowledge_fs import KnowledgeFSControlSpace, KnowledgeFSControlSpaceState
+
+
+@dataclass(frozen=True, slots=True)
+class KnowledgeFSControlSpaceCASUpdate:
+    """Values applied only when the stored state and resource version match."""
+
+    tenant_id: str
+    control_space_id: str
+    expected_resource_version: int
+    expected_state: KnowledgeFSControlSpaceState
+    new_state: KnowledgeFSControlSpaceState
+    lifecycle_operation_id: str
+    knowledge_space_id: str | None = None
+    knowledge_space_revision: int | None = None
+    attempted_at: datetime | None = None
+    last_error_code: str | None = None
+    last_error_message: str | None = None
+
+
+class KnowledgeFSControlSpaceRepository(Protocol):
+    """Repository operations that always include the owning Dify tenant."""
+
+    def add(self, control_space: KnowledgeFSControlSpace) -> KnowledgeFSControlSpace: ...
+
+    def get(self, *, tenant_id: str, control_space_id: str) -> KnowledgeFSControlSpace | None: ...
+
+    def find_by_provisioning_key(self, *, provisioning_key: str) -> KnowledgeFSControlSpace | None: ...
+
+    def find_by_knowledge_space_id(
+        self, *, tenant_id: str, knowledge_space_id: str
+    ) -> KnowledgeFSControlSpace | None: ...
+
+    def list_for_tenant(self, *, tenant_id: str) -> tuple[KnowledgeFSControlSpace, ...]: ...
+
+    def list_for_reconciliation(self, *, limit: int) -> tuple[KnowledgeFSControlSpace, ...]: ...
+
+    def compare_and_set_lifecycle(self, update_values: KnowledgeFSControlSpaceCASUpdate) -> bool: ...
+
+    def mark_deletion_irreversible(
+        self,
+        *,
+        tenant_id: str,
+        control_space_id: str,
+        lifecycle_operation_id: str,
+        irreversible_at: datetime,
+    ) -> bool: ...
+
+    def list_workspace_deletion_blockers(self, *, tenant_id: str) -> tuple[str, ...]: ...
+
+
+__all__ = ["KnowledgeFSControlSpaceCASUpdate", "KnowledgeFSControlSpaceRepository"]

api/repositories/knowledge_fs_cutover_repository.py

@@ -0,0 +1,229 @@
+"""Persistence contract for KnowledgeFS workspace migration and cutover state."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from datetime import datetime
+from typing import Protocol
+
+from models.knowledge_fs_cutover import (
+    KnowledgeFSCutoverRevisionWatermark,
+    KnowledgeFSCutoverSmokeResults,
+    KnowledgeFSMigrationIssue,
+    KnowledgeFSMigrationIssueStatus,
+    KnowledgeFSMigrationQuarantine,
+    KnowledgeFSMigrationQuarantineDisposition,
+    KnowledgeFSMigrationQuarantineKind,
+    KnowledgeFSShadowAuthorizationDecision,
+    KnowledgeFSShadowAuthorizationDiff,
+    KnowledgeFSShadowAuthorizationObservation,
+    KnowledgeFSWorkspaceCutoverLedger,
+    KnowledgeFSWorkspaceCutoverPhase,
+)
+
+
+@dataclass(frozen=True, slots=True)
+class KnowledgeFSCutoverCASUpdate:
+    """Optional values changed with the phase and CAS version in one SQL statement."""
+
+    tenant_id: str
+    expected_phase: KnowledgeFSWorkspaceCutoverPhase
+    expected_cas_version: int
+    new_phase: KnowledgeFSWorkspaceCutoverPhase
+    source_revision_watermark: KnowledgeFSCutoverRevisionWatermark | None = None
+    final_revision_watermark: KnowledgeFSCutoverRevisionWatermark | None = None
+    applied_revision_watermark: KnowledgeFSCutoverRevisionWatermark | None = None
+    source_task_watermark: int | None = None
+    final_task_watermark: int | None = None
+    applied_task_watermark: int | None = None
+    shadow_started_at: datetime | None = None
+    shadow_completed_at: datetime | None = None
+    shadow_evidence_digest: str | None = None
+    shadow_observation_count: int | None = None
+    shadow_window_started_at: datetime | None = None
+    shadow_window_ended_at: datetime | None = None
+    shadow_traffic_zero: bool | None = None
+    shadow_traffic_zero_evidence: dict[str, object] | None = None
+    shadow_latest_observed_revision: KnowledgeFSCutoverRevisionWatermark | None = None
+    shadow_producer: str | None = None
+    shadow_completed_by_operator: str | None = None
+    shadow_completed_by_account_id: str | None = None
+    remote_freeze_id: str | None = None
+    remote_freeze_revision: int | None = None
+    remote_freeze_digest: str | None = None
+    remote_freeze_task_watermark: int | None = None
+    remote_freeze_control_space_id: str | None = None
+    remote_freeze_frozen_at: datetime | None = None
+    remote_freeze_updated_at: datetime | None = None
+    remote_freeze_acknowledged_at: datetime | None = None
+    remote_freeze_applied: bool | None = None
+    remote_freeze_replayed: bool | None = None
+    remote_activation_id: str | None = None
+    remote_activation_revision: int | None = None
+    remote_activation_digest: str | None = None
+    remote_activation_control_space_id: str | None = None
+    remote_activation_activated_at: datetime | None = None
+    remote_activation_updated_at: datetime | None = None
+    remote_activation_acknowledged_at: datetime | None = None
+    remote_activation_applied: bool | None = None
+    remote_activation_replayed: bool | None = None
+    clear_smoke_results: bool = False
+    freeze_at: datetime | None = None
+    cutover_at: datetime | None = None
+    rolled_back_at: datetime | None = None
+    rollback_cutoff_at: datetime | None = None
+    observation_started_at: datetime | None = None
+    observation_window_ends_at: datetime | None = None
+    observation_completed_at: datetime | None = None
+    maximum_task_expires_at: datetime | None = None
+    irreversible_cleanup_at: datetime | None = None
+    smoke_results: KnowledgeFSCutoverSmokeResults | None = None
+    legacy_dependency_report: list[dict[str, object]] | None = None
+    legacy_dependency_checked_at: datetime | None = None
+    legacy_dependency_ready: bool | None = None
+    product_routes_enabled: bool | None = None
+    capability_v2_enabled: bool | None = None
+    integrated_mode_enabled: bool | None = None
+    legacy_acl_read_only: bool | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class KnowledgeFSQuarantineCASUpdate:
+    """Tenant/item-scoped transition from an operator disposition to resolved."""
+
+    tenant_id: str
+    ledger_id: str
+    source_kind: KnowledgeFSMigrationQuarantineKind
+    source_id: str
+    expected_disposition: KnowledgeFSMigrationQuarantineDisposition
+    expected_row_version: int
+    resolved_by_operator: str
+    resolved_by_account_id: str
+    evidence: dict[str, object]
+    resolved_at: datetime
+
+
+@dataclass(frozen=True, slots=True)
+class KnowledgeFSShadowDiffCASUpdate:
+    """Replace one current diff disposition after append-only safe re-evaluation."""
+
+    tenant_id: str
+    ledger_id: str
+    diff_key: str
+    expected_row_version: int
+    control_space_id: str | None
+    principal: str
+    legacy_allowed: bool | None
+    dify_allowed: bool
+    decision: KnowledgeFSShadowAuthorizationDecision
+    reason: str
+    observed_revision: KnowledgeFSCutoverRevisionWatermark
+    status: KnowledgeFSMigrationIssueStatus
+    current_evidence_digest: str
+    last_observed_at: datetime
+
+
+class KnowledgeFSCutoverRepository(Protocol):
+    """Tenant-scoped storage; callers own the surrounding transaction."""
+
+    def add_ledger(self, ledger: KnowledgeFSWorkspaceCutoverLedger) -> KnowledgeFSWorkspaceCutoverLedger: ...
+
+    def get_ledger(self, *, tenant_id: str) -> KnowledgeFSWorkspaceCutoverLedger | None: ...
+
+    def list_ledgers(self) -> tuple[KnowledgeFSWorkspaceCutoverLedger, ...]: ...
+
+    def compare_and_set(self, update_values: KnowledgeFSCutoverCASUpdate) -> bool: ...
+
+    def add_issue(self, issue: KnowledgeFSMigrationIssue) -> KnowledgeFSMigrationIssue: ...
+
+    def get_issue(self, *, tenant_id: str, ledger_id: str, issue_key: str) -> KnowledgeFSMigrationIssue | None: ...
+
+    def set_issue_status(
+        self,
+        *,
+        tenant_id: str,
+        ledger_id: str,
+        issue_key: str,
+        expected_status: KnowledgeFSMigrationIssueStatus,
+        new_status: KnowledgeFSMigrationIssueStatus,
+        account_id: str,
+        changed_at: datetime,
+    ) -> bool: ...
+
+    def add_quarantine(self, item: KnowledgeFSMigrationQuarantine) -> KnowledgeFSMigrationQuarantine: ...
+
+    def get_quarantine(
+        self,
+        *,
+        tenant_id: str,
+        ledger_id: str,
+        source_kind: KnowledgeFSMigrationQuarantineKind,
+        source_id: str,
+    ) -> KnowledgeFSMigrationQuarantine | None: ...
+
+    def resolve_quarantine(self, update_values: KnowledgeFSQuarantineCASUpdate) -> bool: ...
+
+    def add_shadow_diff(self, diff: KnowledgeFSShadowAuthorizationDiff) -> KnowledgeFSShadowAuthorizationDiff: ...
+
+    def get_shadow_diff(
+        self, *, tenant_id: str, ledger_id: str, diff_key: str
+    ) -> KnowledgeFSShadowAuthorizationDiff | None: ...
+
+    def set_shadow_diff_status(
+        self,
+        *,
+        tenant_id: str,
+        ledger_id: str,
+        diff_key: str,
+        expected_status: KnowledgeFSMigrationIssueStatus,
+        new_status: KnowledgeFSMigrationIssueStatus,
+        account_id: str,
+        changed_at: datetime,
+    ) -> bool: ...
+
+    def reevaluate_shadow_diff(self, update_values: KnowledgeFSShadowDiffCASUpdate) -> bool: ...
+
+    def add_shadow_observation(
+        self, observation: KnowledgeFSShadowAuthorizationObservation
+    ) -> KnowledgeFSShadowAuthorizationObservation: ...
+
+    def get_shadow_observation(
+        self,
+        *,
+        tenant_id: str,
+        ledger_id: str,
+        diff_key: str,
+        evidence_digest: str,
+    ) -> KnowledgeFSShadowAuthorizationObservation | None: ...
+
+    def count_open_issues(self, *, tenant_id: str, ledger_id: str) -> int: ...
+
+    def count_unapproved_shadow_diffs(self, *, tenant_id: str, ledger_id: str) -> int: ...
+
+    def count_unresolved_quarantine(
+        self,
+        *,
+        tenant_id: str,
+        ledger_id: str,
+        source_kinds: tuple[KnowledgeFSMigrationQuarantineKind, ...],
+    ) -> int: ...
+
+    def list_issues(self, *, tenant_id: str, ledger_id: str) -> tuple[KnowledgeFSMigrationIssue, ...]: ...
+
+    def list_quarantine(self, *, tenant_id: str, ledger_id: str) -> tuple[KnowledgeFSMigrationQuarantine, ...]: ...
+
+    def list_shadow_diffs(
+        self, *, tenant_id: str, ledger_id: str
+    ) -> tuple[KnowledgeFSShadowAuthorizationDiff, ...]: ...
+
+    def list_shadow_observations(
+        self, *, tenant_id: str, ledger_id: str
+    ) -> tuple[KnowledgeFSShadowAuthorizationObservation, ...]: ...
+
+
+__all__ = [
+    "KnowledgeFSCutoverCASUpdate",
+    "KnowledgeFSCutoverRepository",
+    "KnowledgeFSQuarantineCASUpdate",
+    "KnowledgeFSShadowDiffCASUpdate",
+]

api/repositories/knowledge_fs_lifecycle_outbox_repository.py

@@ -0,0 +1,95 @@
+"""Persistence boundary for durable KnowledgeFS lifecycle commands."""
+
+from __future__ import annotations
+
+from datetime import datetime, timedelta
+from typing import Protocol
+
+from models.knowledge_fs import KnowledgeFSLifecycleOperation, KnowledgeFSLifecycleOutbox
+
+
+class KnowledgeFSLifecycleOutboxRepository(Protocol):
+    def add(self, command: KnowledgeFSLifecycleOutbox) -> KnowledgeFSLifecycleOutbox: ...
+
+    def get(self, *, outbox_id: str) -> KnowledgeFSLifecycleOutbox | None: ...
+
+    def get_by_operation_id(self, *, tenant_id: str, operation_id: str) -> KnowledgeFSLifecycleOutbox | None: ...
+
+    def find_open_for_control_space(
+        self,
+        *,
+        tenant_id: str,
+        control_space_id: str,
+        operation: KnowledgeFSLifecycleOperation,
+    ) -> KnowledgeFSLifecycleOutbox | None: ...
+
+    def find_latest_for_control_space(
+        self,
+        *,
+        tenant_id: str,
+        control_space_id: str,
+        operation: KnowledgeFSLifecycleOperation,
+    ) -> KnowledgeFSLifecycleOutbox | None: ...
+
+    def reactivate_dead_letter(self, *, outbox_id: str) -> bool: ...
+
+    def supersede_unattempted(
+        self,
+        *,
+        outbox_id: str,
+        completed_at: datetime,
+        error_code: str,
+        error_message: str,
+    ) -> bool: ...
+
+    def supersede_after_remote_absence(
+        self,
+        *,
+        outbox_id: str,
+        observed_at: datetime,
+        error_code: str,
+        error_message: str,
+    ) -> bool: ...
+
+    def claim_next(
+        self,
+        *,
+        lease_owner: str,
+        now: datetime,
+        lease_duration: timedelta,
+        allowed_operations: tuple[KnowledgeFSLifecycleOperation, ...],
+    ) -> KnowledgeFSLifecycleOutbox | None: ...
+
+    def acknowledge(
+        self,
+        *,
+        outbox_id: str,
+        lease_owner: str,
+        expected_lease_expires_at: datetime,
+        completed_at: datetime,
+    ) -> bool: ...
+
+    def schedule_retry(
+        self,
+        *,
+        outbox_id: str,
+        lease_owner: str,
+        expected_lease_expires_at: datetime,
+        next_attempt_at: datetime,
+        error_code: str,
+        error_message: str,
+    ) -> bool: ...
+
+    def mark_dead_letter(
+        self,
+        *,
+        outbox_id: str,
+        lease_owner: str,
+        expected_lease_expires_at: datetime,
+        completed_at: datetime,
+        error_code: str,
+        error_message: str,
+    ) -> bool: ...
+
+
+__all__ = ["KnowledgeFSLifecycleOutboxRepository"]

api/repositories/sqlalchemy_knowledge_fs_capability_issuance_auditor.py

@@ -0,0 +1,35 @@
+"""Durable SQLAlchemy sink for sanitized KnowledgeFS capability issuance audits."""
+
+from __future__ import annotations
+
+from typing import cast, override
+
+from sqlalchemy.orm import Session, sessionmaker
+
+from models.knowledge_fs import KnowledgeFSCapabilityClaimsSummary, KnowledgeFSCapabilityIssuanceAudit
+from services.knowledge_fs_capability import CapabilityIssuanceAuditEvent, CapabilityIssuanceAuditor
+
+
+class SQLAlchemyKnowledgeFSCapabilityIssuanceAuditor(CapabilityIssuanceAuditor):
+    """Commit the audit independently so token issuance fails closed on audit failure."""
+
+    def __init__(self, session_maker: sessionmaker[Session]):
+        self._session_maker = session_maker
+
+    @override
+    def record(self, event: CapabilityIssuanceAuditEvent) -> None:
+        serialized = event.model_dump(mode="json")
+        trace_id = cast(str, serialized.pop("trace_id"))
+        jti_hash = cast(str, serialized.pop("jti_hash"))
+        audit = KnowledgeFSCapabilityIssuanceAudit(
+            tenant_id=event.namespace_id,
+            control_space_id=event.control_space_id,
+            trace_id=trace_id,
+            jti_hash=jti_hash,
+            claims_summary=cast(KnowledgeFSCapabilityClaimsSummary, serialized),
+        )
+        with self._session_maker.begin() as session:
+            session.add(audit)
+
+
+__all__ = ["SQLAlchemyKnowledgeFSCapabilityIssuanceAuditor"]

api/repositories/sqlalchemy_knowledge_fs_capability_issuance_reservation_repository.py

@@ -0,0 +1,185 @@
+"""Transactional persistence for pre-signing KnowledgeFS capability reservations."""
+
+from __future__ import annotations
+
+from datetime import datetime, timedelta
+from typing import cast
+
+import sqlalchemy as sa
+from sqlalchemy.engine import CursorResult
+from sqlalchemy.orm import Session
+
+from models.knowledge_fs import (
+    KnowledgeFSCapabilityIssuanceReservation,
+    KnowledgeFSCapabilityIssuanceReservationStatus,
+    KnowledgeFSCapabilityReservationSummary,
+)
+from services.knowledge_fs_capability import (
+    CAPABILITY_ISSUANCE_PROFILES,
+    KNOWLEDGE_FS_CAPABILITY_OPERATIONS,
+    CapabilityIssueRequest,
+)
+
+_TERMINAL_RETENTION = timedelta(days=1)
+
+
+class KnowledgeFSCapabilityIssuanceReservationError(RuntimeError):
+    """A retry attempted to change or lose a durable issuance binding."""
+
+
+class SQLAlchemyKnowledgeFSCapabilityIssuanceReservationRepository:
+    def __init__(self, session: Session) -> None:
+        self._session = session
+
+    def reserve(self, request: CapabilityIssueRequest) -> KnowledgeFSCapabilityIssuanceReservation:
+        summary = _request_summary(request)
+        reservation = self._session.scalar(
+            sa.select(KnowledgeFSCapabilityIssuanceReservation)
+            .where(
+                KnowledgeFSCapabilityIssuanceReservation.tenant_id == request.namespace_id,
+                KnowledgeFSCapabilityIssuanceReservation.grant_id == request.grant_id,
+            )
+            .with_for_update()
+        )
+        if reservation is None:
+            reservation = KnowledgeFSCapabilityIssuanceReservation(
+                tenant_id=request.namespace_id,
+                control_space_id=request.control_space_id,
+                grant_id=request.grant_id,
+                trace_id=request.trace_id,
+                subject=summary["subject"],
+                caller_kind=request.caller_kind,
+                request_summary=summary,
+            )
+            self._session.add(reservation)
+            self._session.flush()
+            return reservation
+        if (
+            reservation.control_space_id != request.control_space_id
+            or reservation.trace_id != request.trace_id
+            or reservation.subject != summary["subject"]
+            or reservation.caller_kind != request.caller_kind
+            or reservation.request_summary != summary
+        ):
+            raise KnowledgeFSCapabilityIssuanceReservationError(
+                "Capability grant retry does not match its durable reservation"
+            )
+        if reservation.status is KnowledgeFSCapabilityIssuanceReservationStatus.FAILED:
+            reservation.status = KnowledgeFSCapabilityIssuanceReservationStatus.RESERVED
+            reservation.failed_at = None
+            reservation.failure_code = None
+            reservation.cleanup_after = None
+            reservation.row_version += 1
+            self._session.flush()
+        return reservation
+
+    def mark_issued(
+        self,
+        *,
+        tenant_id: str,
+        grant_id: str,
+        issued_at: datetime,
+        token_expires_at: datetime,
+    ) -> None:
+        reservation = self._locked(tenant_id=tenant_id, grant_id=grant_id)
+        reservation.status = KnowledgeFSCapabilityIssuanceReservationStatus.ISSUED
+        reservation.issued_at = issued_at
+        reservation.token_expires_at = token_expires_at
+        reservation.failed_at = None
+        reservation.failure_code = None
+        reservation.cleanup_after = token_expires_at + _TERMINAL_RETENTION
+        reservation.row_version += 1
+        self._session.flush()
+
+    def mark_failed(
+        self,
+        *,
+        tenant_id: str,
+        grant_id: str,
+        failed_at: datetime,
+        failure_code: str,
+    ) -> None:
+        reservation = self._locked(tenant_id=tenant_id, grant_id=grant_id)
+        if reservation.status is KnowledgeFSCapabilityIssuanceReservationStatus.ISSUED:
+            return
+        reservation.status = KnowledgeFSCapabilityIssuanceReservationStatus.FAILED
+        reservation.issued_at = None
+        reservation.token_expires_at = None
+        reservation.failed_at = failed_at
+        reservation.failure_code = failure_code[:128]
+        reservation.cleanup_after = failed_at + _TERMINAL_RETENTION
+        reservation.row_version += 1
+        self._session.flush()
+
+    def cleanup_terminal(self, *, before: datetime, limit: int = 1_000) -> int:
+        if limit <= 0:
+            raise ValueError("Capability reservation cleanup limit must be positive")
+        reservation_ids = tuple(
+            self._session.scalars(
+                sa.select(KnowledgeFSCapabilityIssuanceReservation.id)
+                .where(
+                    KnowledgeFSCapabilityIssuanceReservation.status
+                    != KnowledgeFSCapabilityIssuanceReservationStatus.RESERVED,
+                    KnowledgeFSCapabilityIssuanceReservation.cleanup_after <= before,
+                )
+                .order_by(
+                    KnowledgeFSCapabilityIssuanceReservation.cleanup_after,
+                    KnowledgeFSCapabilityIssuanceReservation.id,
+                )
+                .limit(limit)
+                .with_for_update()
+            )
+        )
+        if not reservation_ids:
+            return 0
+        result = self._session.execute(
+            sa.delete(KnowledgeFSCapabilityIssuanceReservation).where(
+                KnowledgeFSCapabilityIssuanceReservation.id.in_(reservation_ids)
+            )
+        )
+        return cast(CursorResult[tuple[object, ...]], result).rowcount or 0
+
+    def _locked(self, *, tenant_id: str, grant_id: str) -> KnowledgeFSCapabilityIssuanceReservation:
+        reservation = self._session.scalar(
+            sa.select(KnowledgeFSCapabilityIssuanceReservation)
+            .where(
+                KnowledgeFSCapabilityIssuanceReservation.tenant_id == tenant_id,
+                KnowledgeFSCapabilityIssuanceReservation.grant_id == grant_id,
+            )
+            .with_for_update()
+        )
+        if reservation is None:
+            raise KnowledgeFSCapabilityIssuanceReservationError("Capability issuance reservation is missing")
+        return reservation
+
+
+def _request_summary(request: CapabilityIssueRequest) -> KnowledgeFSCapabilityReservationSummary:
+    operation = KNOWLEDGE_FS_CAPABILITY_OPERATIONS[request.operation_id]
+    profile = CAPABILITY_ISSUANCE_PROFILES[request.caller_kind]
+    subject = f"{profile.subject_prefix}:{request.principal_id}"
+    return cast(
+        KnowledgeFSCapabilityReservationSummary,
+        {
+            "action": operation.action,
+            "actor": request.actor,
+            "authz_revision": request.authz_revision.model_dump(mode="json"),
+            "caller_kind": request.caller_kind,
+            "content_policy_revision": request.content_policy_revision,
+            "content_scope_ids": list(request.content_scope_ids),
+            "control_space_id": request.control_space_id,
+            "grant_id": request.grant_id,
+            "namespace_id": request.namespace_id,
+            "operation_id": request.operation_id,
+            "resource_id": request.resource.id,
+            "resource_parent_id": request.resource.parent_id,
+            "resource_type": request.resource.type,
+            "subject": subject,
+            "trace_id": request.trace_id,
+        },
+    )
+
+
+__all__ = [
+    "KnowledgeFSCapabilityIssuanceReservationError",
+    "SQLAlchemyKnowledgeFSCapabilityIssuanceReservationRepository",
+]

api/repositories/sqlalchemy_knowledge_fs_cleanup_repository.py

@@ -0,0 +1,87 @@
+"""SQLAlchemy cleanup-authorization repository with status and version CAS."""
+
+from __future__ import annotations
+
+from typing import cast, override
+
+from sqlalchemy import select, update
+from sqlalchemy.engine import CursorResult
+from sqlalchemy.orm import Session
+
+from models.knowledge_fs_cleanup import KnowledgeFSCleanupAuthorization
+from repositories.knowledge_fs_cleanup_repository import (
+    KnowledgeFSCleanupAuthorizationCASUpdate,
+    KnowledgeFSCleanupAuthorizationRepository,
+)
+
+
+class SQLAlchemyKnowledgeFSCleanupAuthorizationRepository(KnowledgeFSCleanupAuthorizationRepository):
+    _session: Session
+
+    def __init__(self, session: Session):
+        self._session = session
+
+    @override
+    def add(self, authorization: KnowledgeFSCleanupAuthorization) -> KnowledgeFSCleanupAuthorization:
+        self._session.add(authorization)
+        self._session.flush()
+        return authorization
+
+    @override
+    def get(
+        self,
+        *,
+        tenant_id: str,
+        ledger_id: str,
+        request_id: str,
+    ) -> KnowledgeFSCleanupAuthorization | None:
+        statement = (
+            select(KnowledgeFSCleanupAuthorization)
+            .where(
+                KnowledgeFSCleanupAuthorization.tenant_id == tenant_id,
+                KnowledgeFSCleanupAuthorization.ledger_id == ledger_id,
+                KnowledgeFSCleanupAuthorization.request_id == request_id,
+            )
+            .execution_options(populate_existing=True)
+        )
+        return self._session.scalar(statement)
+
+    @override
+    def compare_and_set(self, update_values: KnowledgeFSCleanupAuthorizationCASUpdate) -> bool:
+        values: dict[str, object] = {
+            "status": update_values.new_status,
+            "row_version": KnowledgeFSCleanupAuthorization.row_version + 1,
+        }
+        for field_name in (
+            "approved_by_account_id",
+            "approved_at",
+            "approval_expires_at",
+            "approved_ledger_cas_version",
+            "started_by_account_id",
+            "started_at",
+            "started_ledger_cas_version",
+            "completed_by_account_id",
+            "completed_at",
+            "completion_evidence",
+            "completed_ledger_cas_version",
+        ):
+            value = getattr(update_values, field_name)
+            if value is not None:
+                values[field_name] = value
+        statement = (
+            update(KnowledgeFSCleanupAuthorization)
+            .where(
+                KnowledgeFSCleanupAuthorization.tenant_id == update_values.tenant_id,
+                KnowledgeFSCleanupAuthorization.ledger_id == update_values.ledger_id,
+                KnowledgeFSCleanupAuthorization.request_id == update_values.request_id,
+                KnowledgeFSCleanupAuthorization.status == update_values.expected_status,
+                KnowledgeFSCleanupAuthorization.row_version == update_values.expected_row_version,
+            )
+            .values(**values)
+            .execution_options(synchronize_session=False)
+        )
+        result = self._session.execute(statement)
+        return (cast(CursorResult[tuple[object, ...]], result).rowcount or 0) == 1
+
+
+__all__ = ["SQLAlchemyKnowledgeFSCleanupAuthorizationRepository"]

api/repositories/sqlalchemy_knowledge_fs_control_space_repository.py

@@ -0,0 +1,187 @@
+"""SQLAlchemy implementation of the KnowledgeFS control-space repository."""
+
+from __future__ import annotations
+
+from datetime import datetime
+from typing import cast, override
+
+from sqlalchemy import case, select, update
+from sqlalchemy.engine import CursorResult
+from sqlalchemy.orm import Session
+
+from models.knowledge_fs import KnowledgeFSControlSpace, KnowledgeFSControlSpaceState
+from repositories.knowledge_fs_control_space_repository import (
+    KnowledgeFSControlSpaceCASUpdate,
+    KnowledgeFSControlSpaceRepository,
+)
+
+
+class SQLAlchemyKnowledgeFSControlSpaceRepository(KnowledgeFSControlSpaceRepository):
+    """Caller-transaction-owned repository with tenant and CAS guards."""
+
+    def __init__(self, session: Session):
+        self._session = session
+
+    @override
+    def add(self, control_space: KnowledgeFSControlSpace) -> KnowledgeFSControlSpace:
+        self._session.add(control_space)
+        self._session.flush()
+        return control_space
+
+    @override
+    def get(self, *, tenant_id: str, control_space_id: str) -> KnowledgeFSControlSpace | None:
+        statement = (
+            select(KnowledgeFSControlSpace)
+            .where(
+                KnowledgeFSControlSpace.tenant_id == tenant_id,
+                KnowledgeFSControlSpace.id == control_space_id,
+            )
+            .execution_options(populate_existing=True)
+        )
+        return self._session.scalar(statement)
+
+    @override
+    def find_by_provisioning_key(self, *, provisioning_key: str) -> KnowledgeFSControlSpace | None:
+        statement = (
+            select(KnowledgeFSControlSpace)
+            .where(KnowledgeFSControlSpace.provisioning_key == provisioning_key)
+            .execution_options(populate_existing=True)
+        )
+        return self._session.scalar(statement)
+
+    @override
+    def find_by_knowledge_space_id(self, *, tenant_id: str, knowledge_space_id: str) -> KnowledgeFSControlSpace | None:
+        statement = (
+            select(KnowledgeFSControlSpace)
+            .where(
+                KnowledgeFSControlSpace.tenant_id == tenant_id,
+                KnowledgeFSControlSpace.knowledge_space_id == knowledge_space_id,
+            )
+            .execution_options(populate_existing=True)
+        )
+        return self._session.scalar(statement)
+
+    @override
+    def list_for_tenant(self, *, tenant_id: str) -> tuple[KnowledgeFSControlSpace, ...]:
+        statement = (
+            select(KnowledgeFSControlSpace)
+            .where(KnowledgeFSControlSpace.tenant_id == tenant_id)
+            .order_by(KnowledgeFSControlSpace.created_at, KnowledgeFSControlSpace.id)
+            .execution_options(populate_existing=True)
+        )
+        return tuple(self._session.scalars(statement).all())
+
+    @override
+    def list_for_reconciliation(self, *, limit: int) -> tuple[KnowledgeFSControlSpace, ...]:
+        if limit <= 0:
+            return ()
+        statement = (
+            select(KnowledgeFSControlSpace)
+            .where(KnowledgeFSControlSpace.state != KnowledgeFSControlSpaceState.DELETED)
+            .order_by(
+                case(
+                    (
+                        KnowledgeFSControlSpace.state.in_(
+                            (
+                                KnowledgeFSControlSpaceState.PROVISIONING,
+                                KnowledgeFSControlSpaceState.DELETING,
+                                KnowledgeFSControlSpaceState.ERROR,
+                            )
+                        ),
+                        0,
+                    ),
+                    else_=1,
+                ),
+                KnowledgeFSControlSpace.updated_at,
+                KnowledgeFSControlSpace.id,
+            )
+            .limit(limit)
+            .execution_options(populate_existing=True)
+        )
+        return tuple(self._session.scalars(statement).all())
+
+    @override
+    def compare_and_set_lifecycle(self, update_values: KnowledgeFSControlSpaceCASUpdate) -> bool:
+        values: dict[str, object] = {
+            "state": update_values.new_state,
+            "resource_version": KnowledgeFSControlSpace.resource_version + 1,
+            "lifecycle_operation_id": update_values.lifecycle_operation_id,
+            "last_error_code": update_values.last_error_code,
+            "last_error_message": update_values.last_error_message,
+        }
+        if update_values.knowledge_space_id is not None:
+            values["knowledge_space_id"] = update_values.knowledge_space_id
+        if update_values.knowledge_space_revision is not None:
+            values["knowledge_space_revision"] = update_values.knowledge_space_revision
+        if update_values.attempted_at is not None:
+            values["attempt_count"] = KnowledgeFSControlSpace.attempt_count + 1
+            values["last_attempt_at"] = update_values.attempted_at
+
+        statement = (
+            update(KnowledgeFSControlSpace)
+            .where(
+                KnowledgeFSControlSpace.tenant_id == update_values.tenant_id,
+                KnowledgeFSControlSpace.id == update_values.control_space_id,
+                KnowledgeFSControlSpace.resource_version == update_values.expected_resource_version,
+                KnowledgeFSControlSpace.state == update_values.expected_state,
+            )
+            .values(**values)
+            .execution_options(synchronize_session=False)
+        )
+        if update_values.knowledge_space_id is not None:
+            statement = statement.where(
+                (KnowledgeFSControlSpace.knowledge_space_id.is_(None))
+                | (KnowledgeFSControlSpace.knowledge_space_id == update_values.knowledge_space_id)
+            )
+        if update_values.knowledge_space_revision is not None:
+            statement = statement.where(
+                KnowledgeFSControlSpace.knowledge_space_revision <= update_values.knowledge_space_revision
+            )
+        result = self._session.execute(statement)
+        return (cast(CursorResult[tuple[object, ...]], result).rowcount or 0) == 1
+
+    @override
+    def mark_deletion_irreversible(
+        self,
+        *,
+        tenant_id: str,
+        control_space_id: str,
+        lifecycle_operation_id: str,
+        irreversible_at: datetime,
+    ) -> bool:
+        statement = (
+            update(KnowledgeFSControlSpace)
+            .where(
+                KnowledgeFSControlSpace.tenant_id == tenant_id,
+                KnowledgeFSControlSpace.id == control_space_id,
+                KnowledgeFSControlSpace.lifecycle_operation_id == lifecycle_operation_id,
+                KnowledgeFSControlSpace.state.in_(
+                    (
+                        KnowledgeFSControlSpaceState.DELETING,
+                        KnowledgeFSControlSpaceState.ERROR,
+                        KnowledgeFSControlSpaceState.DELETED,
+                    )
+                ),
+                (KnowledgeFSControlSpace.deletion_irreversible_at.is_(None))
+                | (KnowledgeFSControlSpace.deletion_irreversible_at == irreversible_at),
+            )
+            .values(deletion_irreversible_at=irreversible_at)
+            .execution_options(synchronize_session=False)
+        )
+        result = self._session.execute(statement)
+        return (cast(CursorResult[tuple[object, ...]], result).rowcount or 0) == 1
+
+    @override
+    def list_workspace_deletion_blockers(self, *, tenant_id: str) -> tuple[str, ...]:
+        statement = (
+            select(KnowledgeFSControlSpace.id)
+            .where(
+                KnowledgeFSControlSpace.tenant_id == tenant_id,
+                KnowledgeFSControlSpace.state != KnowledgeFSControlSpaceState.DELETED,
+            )
+            .order_by(KnowledgeFSControlSpace.created_at, KnowledgeFSControlSpace.id)
+        )
+        return tuple(self._session.scalars(statement).all())
+
+
+__all__ = ["SQLAlchemyKnowledgeFSControlSpaceRepository"]