if session unauthorized, rejoin

- Introduced AmplitudeProvider component for initializing Amplitude analytics.
- Integrated user tracking in AppContextProvider to report user and workspace information.
- Added event tracking for workflow log filter selections.
- Updated package.json and pnpm-lock.yaml to include @amplitude/unified dependency.

.devcontainer/post_create_command.sh

@@ -6,7 +6,7 @@ cd web && pnpm install
 pipx install uv
 
 echo "alias start-api=\"cd $WORKSPACE_ROOT/api && uv run python -m flask run --host 0.0.0.0 --port=5001 --debug\"" >> ~/.bashrc
-echo "alias start-worker=\"cd $WORKSPACE_ROOT/api && uv run python -m celery -A app.celery worker -P gevent -c 1 --loglevel INFO -Q dataset,generation,mail,ops_trace,app_deletion,plugin,workflow_storage\"" >> ~/.bashrc
+echo "alias start-worker=\"cd $WORKSPACE_ROOT/api && uv run python -m celery -A app.celery worker -P threads -c 1 --loglevel INFO -Q dataset,generation,mail,ops_trace,app_deletion,plugin,workflow_storage\"" >> ~/.bashrc
 echo "alias start-web=\"cd $WORKSPACE_ROOT/web && pnpm dev\"" >> ~/.bashrc
 echo "alias start-web-prod=\"cd $WORKSPACE_ROOT/web && pnpm build && pnpm start\"" >> ~/.bashrc
 echo "alias start-containers=\"cd $WORKSPACE_ROOT/docker && docker-compose -f docker-compose.middleware.yaml -p dify --env-file middleware.env up -d\"" >> ~/.bashrc

.vscode/launch.json.template

@@ -8,8 +8,7 @@
             "module": "flask",
             "env": {
                 "FLASK_APP": "app.py",
-                "FLASK_ENV": "development",
-                "GEVENT_SUPPORT": "True"
+                "FLASK_ENV": "development"
             },
             "args": [
                 "run",
@@ -28,9 +27,7 @@
             "type": "debugpy",
             "request": "launch",
             "module": "celery",
-            "env": {
-                "GEVENT_SUPPORT": "True"
-            },
+            "env": {},
             "args": [
                 "-A",
                 "app.celery",

api/.env.example

@@ -30,6 +30,9 @@ INTERNAL_FILES_URL=http://127.0.0.1:5001
 # The time in seconds after the signature is rejected
 FILES_ACCESS_TIMEOUT=300
 
+# Collaboration mode toggle
+ENABLE_COLLABORATION_MODE=false
+
 # Access token expiration time in minutes
 ACCESS_TOKEN_EXPIRE_MINUTES=60
 
@@ -371,6 +374,12 @@ UPLOAD_IMAGE_FILE_SIZE_LIMIT=10
 UPLOAD_VIDEO_FILE_SIZE_LIMIT=100
 UPLOAD_AUDIO_FILE_SIZE_LIMIT=50
 
+# Comma-separated list of file extensions blocked from upload for security reasons.
+# Extensions should be lowercase without dots (e.g., exe,bat,sh,dll).
+# Empty by default to allow all file types.
+# Recommended: exe,bat,cmd,com,scr,vbs,ps1,msi,dll
+UPLOAD_FILE_EXTENSION_BLACKLIST=
+
 # Model configuration
 MULTIMODAL_SEND_FORMAT=base64
 PROMPT_GENERATION_MAX_TOKENS=512
@@ -608,3 +617,6 @@ SWAGGER_UI_PATH=/swagger-ui.html
 # Whether to encrypt dataset IDs when exporting DSL files (default: true)
 # Set to false to export dataset IDs as plain text for easier cross-environment import
 DSL_EXPORT_ENCRYPT_DATASET_ID=true
+
+# Maximum number of segments for dataset segments API (0 for unlimited)
+DATASET_MAX_SEGMENTS_PER_REQUEST=0

api/Dockerfile

@@ -15,7 +15,11 @@ FROM base AS packages
 # RUN sed -i 's@deb.debian.org@mirrors.aliyun.com@g' /etc/apt/sources.list.d/debian.sources
 
 RUN apt-get update \
-    && apt-get install -y --no-install-recommends gcc g++ libc-dev libffi-dev libgmp-dev libmpfr-dev libmpc-dev
+    && apt-get install -y --no-install-recommends \
+          # basic environment
+          g++ \
+          # for building gmpy2
+          libmpfr-dev libmpc-dev
 
 # Install Python dependencies
 COPY pyproject.toml uv.lock ./
@@ -49,7 +53,9 @@ RUN \
     # Install dependencies
     && apt-get install -y --no-install-recommends \
         # basic environment
-        curl nodejs libgmp-dev libmpfr-dev libmpc-dev \
+        curl nodejs \
+        # for gmpy2 \
+        libgmp-dev libmpfr-dev libmpc-dev \
         # For Security
         expat libldap-2.5-0 perl libsqlite3-0 zlib1g \
         # install fonts to support the use of tools like pypdfium2

api/README.md

@@ -80,7 +80,7 @@
 1. If you need to handle and debug the async tasks (e.g. dataset importing and documents indexing), please start the worker service.
 
 ```bash
-uv run celery -A app.celery worker -P gevent -c 2 --loglevel INFO -Q dataset,mail,ops_trace,app_deletion,plugin,workflow_storage,conversation,priority_pipeline,pipeline
+uv run celery -A app.celery worker -P threads -c 2 --loglevel INFO -Q dataset,mail,ops_trace,app_deletion,plugin,workflow_storage,conversation,priority_pipeline,pipeline
 ```
 
 Additionally, if you want to debug the celery scheduled tasks, you can run the following command in another terminal to start the beat service:

api/app.py

@@ -1,3 +1,4 @@
+import os
 import sys
 
 
@@ -8,33 +9,35 @@ def is_db_command():
 
 
 # create app
+celery = None
+flask_app = None
+socketio_app = None
+
 if is_db_command():
     from app_factory import create_migrations_app
 
     app = create_migrations_app()
+    socketio_app = app
+    flask_app = app
 else:
-    # It seems that JetBrains Python debugger does not work well with gevent,
-    # so we need to disable gevent in debug mode.
-    # If you are using debugpy and set GEVENT_SUPPORT=True, you can debug with gevent.
-    # if (flask_debug := os.environ.get("FLASK_DEBUG", "0")) and flask_debug.lower() in {"false", "0", "no"}:
-    # from gevent import monkey
+    # Gunicorn and Celery handle monkey patching automatically in production by
+    # specifying the `gevent` worker class. Manual monkey patching is not required here.
     #
-    # # gevent
-    # monkey.patch_all()
+    # See `api/docker/entrypoint.sh` (lines 33 and 47) for details.
     #
-    # from grpc.experimental import gevent as grpc_gevent  # type: ignore
-    #
-    # # grpc gevent
-    # grpc_gevent.init_gevent()
-
-    # import psycogreen.gevent  # type: ignore
-    #
-    # psycogreen.gevent.patch_psycopg()
+    # For third-party library patching, refer to `gunicorn.conf.py` and `celery_entrypoint.py`.
 
     from app_factory import create_app
 
-    app = create_app()
-    celery = app.extensions["celery"]
+    socketio_app, flask_app = create_app()
+    app = flask_app
+    celery = flask_app.extensions["celery"]
 
 if __name__ == "__main__":
-    app.run(host="0.0.0.0", port=5001)
+    from gevent import pywsgi
+    from geventwebsocket.handler import WebSocketHandler
+
+    host = os.environ.get("HOST", "0.0.0.0")
+    port = int(os.environ.get("PORT", 5001))
+    server = pywsgi.WSGIServer((host, port), socketio_app, handler_class=WebSocketHandler)
+    server.serve_forever()

api/app_factory.py

@@ -31,14 +31,22 @@ def create_flask_app_with_configs() -> DifyApp:
     return dify_app
 
 
-def create_app() -> DifyApp:
+def create_app() -> tuple[any, DifyApp]:
     start_time = time.perf_counter()
     app = create_flask_app_with_configs()
     initialize_extensions(app)
+
+    import socketio
+
+    from extensions.ext_socketio import sio
+
+    sio.app = app
+    socketio_app = socketio.WSGIApp(sio, app)
+
     end_time = time.perf_counter()
     if dify_config.DEBUG:
         logger.info("Finished create_app (%s ms)", round((end_time - start_time) * 1000, 2))
-    return app
+    return socketio_app, app
 
 
 def initialize_extensions(app: DifyApp):

api/commands.py

@@ -1601,7 +1601,7 @@ def transform_datasource_credentials():
                             "integration_secret": api_key,
                         }
                         datasource_provider = DatasourceProvider(
-                            provider="jina",
+                            provider="jinareader",
                             tenant_id=tenant_id,
                             plugin_id=jina_plugin_id,
                             auth_type=api_key_credential_type.value,

api/configs/feature/__init__.py

@@ -331,6 +331,31 @@ class FileUploadConfig(BaseSettings):
         default=10,
     )
 
+    inner_UPLOAD_FILE_EXTENSION_BLACKLIST: str = Field(
+        description=(
+            "Comma-separated list of file extensions that are blocked from upload. "
+            "Extensions should be lowercase without dots (e.g., 'exe,bat,sh,dll'). "
+            "Empty by default to allow all file types."
+        ),
+        validation_alias=AliasChoices("UPLOAD_FILE_EXTENSION_BLACKLIST"),
+        default="",
+    )
+
+    @computed_field  # type: ignore[misc]
+    @property
+    def UPLOAD_FILE_EXTENSION_BLACKLIST(self) -> set[str]:
+        """
+        Parse and return the blacklist as a set of lowercase extensions.
+        Returns an empty set if no blacklist is configured.
+        """
+        if not self.inner_UPLOAD_FILE_EXTENSION_BLACKLIST:
+            return set()
+        return {
+            ext.strip().lower().strip(".")
+            for ext in self.inner_UPLOAD_FILE_EXTENSION_BLACKLIST.split(",")
+            if ext.strip()
+        }
+
 
 class HttpConfig(BaseSettings):
     """
@@ -846,6 +871,16 @@ class MailConfig(BaseSettings):
         default=None,
     )
 
+    ENABLE_TRIAL_APP: bool = Field(
+        description="Enable trial app",
+        default=False,
+    )
+
+    ENABLE_EXPLORE_BANNER: bool = Field(
+        description="Enable explore banner",
+        default=False,
+    )
+
 
 class RagEtlConfig(BaseSettings):
     """
@@ -920,6 +955,11 @@ class DataSetConfig(BaseSettings):
         default=True,
     )
 
+    DATASET_MAX_SEGMENTS_PER_REQUEST: NonNegativeInt = Field(
+        description="Maximum number of segments for dataset segments API (0 for unlimited)",
+        default=0,
+    )
+
 
 class WorkspaceConfig(BaseSettings):
     """
@@ -1053,6 +1093,13 @@ class PositionConfig(BaseSettings):
         return {item.strip() for item in self.POSITION_TOOL_EXCLUDES.split(",") if item.strip() != ""}
 
 
+class CollaborationConfig(BaseSettings):
+    ENABLE_COLLABORATION_MODE: bool = Field(
+        description="Whether to enable collaboration mode features across the workspace",
+        default=False,
+    )
+
+
 class LoginConfig(BaseSettings):
     ENABLE_EMAIL_CODE_LOGIN: bool = Field(
         description="whether to enable email code login",
@@ -1112,6 +1159,13 @@ class SwaggerUIConfig(BaseSettings):
     )
 
 
+class TenantSelfTaskQueueConfig(BaseSettings):
+    TENANT_SELF_TASK_QUEUE_PULL_SIZE: int = Field(
+        description="Default batch size for tenant self task queue pull operations",
+        default=1,
+    )
+
+
 class FeatureConfig(
     # place the configs in alphabet order
     AppExecutionConfig,
@@ -1136,11 +1190,13 @@ class FeatureConfig(
     RagEtlConfig,
     RepositoryConfig,
     SecurityConfig,
+    TenantSelfTaskQueueConfig,
     ToolConfig,
     UpdateConfig,
     WorkflowConfig,
     WorkflowNodeExecutionConfig,
     WorkspaceConfig,
+    CollaborationConfig,
     LoginConfig,
     AccountConfig,
     SwaggerUIConfig,

api/configs/feature/hosted_service/__init__.py

@@ -8,6 +8,11 @@ class HostedCreditConfig(BaseSettings):
         default="",
     )
 
+    HOSTED_POOL_CREDITS: int = Field(
+        description="Pool credits for hosted service",
+        default=200,
+    )
+
     def get_model_credits(self, model_name: str) -> int:
         """
         Get credit value for a specific model name.
@@ -60,19 +65,46 @@ class HostedOpenAiConfig(BaseSettings):
 
     HOSTED_OPENAI_TRIAL_MODELS: str = Field(
         description="Comma-separated list of available models for trial access",
-        default="gpt-3.5-turbo,"
-        "gpt-3.5-turbo-1106,"
-        "gpt-3.5-turbo-instruct,"
+        default="gpt-4,"
+        "gpt-4-turbo-preview,"
+        "gpt-4-turbo-2024-04-09,"
+        "gpt-4-1106-preview,"
+        "gpt-4-0125-preview,"
+        "gpt-4-turbo,"
+        "gpt-4.1,"
+        "gpt-4.1-2025-04-14,"
+        "gpt-4.1-mini,"
+        "gpt-4.1-mini-2025-04-14,"
+        "gpt-4.1-nano,"
+        "gpt-4.1-nano-2025-04-14,"
+        "gpt-3.5-turbo,"
         "gpt-3.5-turbo-16k,"
         "gpt-3.5-turbo-16k-0613,"
+        "gpt-3.5-turbo-1106,"
         "gpt-3.5-turbo-0613,"
         "gpt-3.5-turbo-0125,"
-        "text-davinci-003",
-    )
-
-    HOSTED_OPENAI_QUOTA_LIMIT: NonNegativeInt = Field(
-        description="Quota limit for hosted OpenAI service usage",
-        default=200,
+        "gpt-3.5-turbo-instruct,"
+        "text-davinci-003,"
+        "chatgpt-4o-latest,"
+        "gpt-4o,"
+        "gpt-4o-2024-05-13,"
+        "gpt-4o-2024-08-06,"
+        "gpt-4o-2024-11-20,"
+        "gpt-4o-audio-preview,"
+        "gpt-4o-audio-preview-2025-06-03,"
+        "gpt-4o-mini,"
+        "gpt-4o-mini-2024-07-18,"
+        "o3-mini,"
+        "o3-mini-2025-01-31,"
+        "gpt-5-mini-2025-08-07,"
+        "gpt-5-mini,"
+        "o4-mini,"
+        "o4-mini-2025-04-16,"
+        "gpt-5-chat-latest,"
+        "gpt-5,"
+        "gpt-5-2025-08-07,"
+        "gpt-5-nano,"
+        "gpt-5-nano-2025-08-07",
     )
 
     HOSTED_OPENAI_PAID_ENABLED: bool = Field(
@@ -87,6 +119,13 @@ class HostedOpenAiConfig(BaseSettings):
         "gpt-4-turbo-2024-04-09,"
         "gpt-4-1106-preview,"
         "gpt-4-0125-preview,"
+        "gpt-4-turbo,"
+        "gpt-4.1,"
+        "gpt-4.1-2025-04-14,"
+        "gpt-4.1-mini,"
+        "gpt-4.1-mini-2025-04-14,"
+        "gpt-4.1-nano,"
+        "gpt-4.1-nano-2025-04-14,"
         "gpt-3.5-turbo,"
         "gpt-3.5-turbo-16k,"
         "gpt-3.5-turbo-16k-0613,"
@@ -94,7 +133,150 @@ class HostedOpenAiConfig(BaseSettings):
         "gpt-3.5-turbo-0613,"
         "gpt-3.5-turbo-0125,"
         "gpt-3.5-turbo-instruct,"
-        "text-davinci-003",
+        "text-davinci-003,"
+        "chatgpt-4o-latest,"
+        "gpt-4o,"
+        "gpt-4o-2024-05-13,"
+        "gpt-4o-2024-08-06,"
+        "gpt-4o-2024-11-20,"
+        "gpt-4o-audio-preview,"
+        "gpt-4o-audio-preview-2025-06-03,"
+        "gpt-4o-mini,"
+        "gpt-4o-mini-2024-07-18,"
+        "o3-mini,"
+        "o3-mini-2025-01-31,"
+        "gpt-5-mini-2025-08-07,"
+        "gpt-5-mini,"
+        "o4-mini,"
+        "o4-mini-2025-04-16,"
+        "gpt-5-chat-latest,"
+        "gpt-5,"
+        "gpt-5-2025-08-07,"
+        "gpt-5-nano,"
+        "gpt-5-nano-2025-08-07",
+    )
+
+
+class HostedGeminiConfig(BaseSettings):
+    """
+    Configuration for fetching Gemini service
+    """
+
+    HOSTED_GEMINI_API_KEY: str | None = Field(
+        description="API key for hosted Gemini service",
+        default=None,
+    )
+
+    HOSTED_GEMINI_API_BASE: str | None = Field(
+        description="Base URL for hosted Gemini API",
+        default=None,
+    )
+
+    HOSTED_GEMINI_API_ORGANIZATION: str | None = Field(
+        description="Organization ID for hosted Gemini service",
+        default=None,
+    )
+
+    HOSTED_GEMINI_TRIAL_ENABLED: bool = Field(
+        description="Enable trial access to hosted Gemini service",
+        default=False,
+    )
+
+    HOSTED_GEMINI_TRIAL_MODELS: str = Field(
+        description="Comma-separated list of available models for trial access",
+        default="gemini-2.5-flash,gemini-2.0-flash,gemini-2.0-flash-lite,",
+    )
+
+    HOSTED_GEMINI_PAID_ENABLED: bool = Field(
+        description="Enable paid access to hosted gemini service",
+        default=False,
+    )
+
+    HOSTED_GEMINI_PAID_MODELS: str = Field(
+        description="Comma-separated list of available models for paid access",
+        default="gemini-2.5-flash,gemini-2.0-flash,gemini-2.0-flash-lite,",
+    )
+
+
+class HostedXAIConfig(BaseSettings):
+    """
+    Configuration for fetching XAI service
+    """
+
+    HOSTED_XAI_API_KEY: str | None = Field(
+        description="API key for hosted XAI service",
+        default=None,
+    )
+
+    HOSTED_XAI_API_BASE: str | None = Field(
+        description="Base URL for hosted XAI API",
+        default=None,
+    )
+
+    HOSTED_XAI_API_ORGANIZATION: str | None = Field(
+        description="Organization ID for hosted XAI service",
+        default=None,
+    )
+
+    HOSTED_XAI_TRIAL_ENABLED: bool = Field(
+        description="Enable trial access to hosted XAI service",
+        default=False,
+    )
+
+    HOSTED_XAI_TRIAL_MODELS: str = Field(
+        description="Comma-separated list of available models for trial access",
+        default="grok-3,grok-3-mini,grok-3-mini-fast",
+    )
+
+    HOSTED_XAI_PAID_ENABLED: bool = Field(
+        description="Enable paid access to hosted XAI service",
+        default=False,
+    )
+
+    HOSTED_XAI_PAID_MODELS: str = Field(
+        description="Comma-separated list of available models for paid access",
+        default="grok-3,grok-3-mini,grok-3-mini-fast",
+    )
+
+
+class HostedDeepseekConfig(BaseSettings):
+    """
+    Configuration for fetching Deepseek service
+    """
+
+    HOSTED_DEEPSEEK_API_KEY: str | None = Field(
+        description="API key for hosted Deepseek service",
+        default=None,
+    )
+
+    HOSTED_DEEPSEEK_API_BASE: str | None = Field(
+        description="Base URL for hosted Deepseek API",
+        default=None,
+    )
+
+    HOSTED_DEEPSEEK_API_ORGANIZATION: str | None = Field(
+        description="Organization ID for hosted Deepseek service",
+        default=None,
+    )
+
+    HOSTED_DEEPSEEK_TRIAL_ENABLED: bool = Field(
+        description="Enable trial access to hosted Deepseek service",
+        default=False,
+    )
+
+    HOSTED_DEEPSEEK_TRIAL_MODELS: str = Field(
+        description="Comma-separated list of available models for trial access",
+        default="deepseek-chat,deepseek-reasoner",
+    )
+
+    HOSTED_DEEPSEEK_PAID_ENABLED: bool = Field(
+        description="Enable paid access to hosted XAI service",
+        default=False,
+    )
+
+    HOSTED_DEEPSEEK_PAID_MODELS: str = Field(
+        description="Comma-separated list of available models for paid access",
+        default="deepseek-chat,deepseek-reasoner",
     )
 
 
@@ -144,16 +326,30 @@ class HostedAnthropicConfig(BaseSettings):
         default=False,
     )
 
-    HOSTED_ANTHROPIC_QUOTA_LIMIT: NonNegativeInt = Field(
-        description="Quota limit for hosted Anthropic service usage",
-        default=600000,
-    )
-
     HOSTED_ANTHROPIC_PAID_ENABLED: bool = Field(
         description="Enable paid access to hosted Anthropic service",
         default=False,
     )
 
+    HOSTED_ANTHROPIC_TRIAL_MODELS: str = Field(
+        description="Comma-separated list of available models for paid access",
+        default="claude-opus-4-20250514,"
+        "claude-sonnet-4-20250514,"
+        "claude-3-5-haiku-20241022,"
+        "claude-3-opus-20240229,"
+        "claude-3-7-sonnet-20250219,"
+        "claude-3-haiku-20240307",
+    )
+    HOSTED_ANTHROPIC_PAID_MODELS: str = Field(
+        description="Comma-separated list of available models for paid access",
+        default="claude-opus-4-20250514,"
+        "claude-sonnet-4-20250514,"
+        "claude-3-5-haiku-20241022,"
+        "claude-3-opus-20240229,"
+        "claude-3-7-sonnet-20250219,"
+        "claude-3-haiku-20240307",
+    )
+
 
 class HostedMinmaxConfig(BaseSettings):
     """
@@ -250,5 +446,8 @@ class HostedServiceConfig(
     HostedModerationConfig,
     # credit config
     HostedCreditConfig,
+    HostedGeminiConfig,
+    HostedXAIConfig,
+    HostedDeepseekConfig,
 ):
     pass

api/configs/middleware/vdb/weaviate_config.py

@@ -22,6 +22,11 @@ class WeaviateConfig(BaseSettings):
         default=True,
     )
 
+    WEAVIATE_GRPC_ENDPOINT: str | None = Field(
+        description="URL of the Weaviate gRPC server (e.g., 'grpc://localhost:50051' or 'grpcs://weaviate.example.com:443')",
+        default=None,
+    )
+
     WEAVIATE_BATCH_SIZE: PositiveInt = Field(
         description="Number of objects to be processed in a single batch operation (default is 100)",
         default=100,

api/controllers/common/errors.py

@@ -25,6 +25,12 @@ class UnsupportedFileTypeError(BaseHTTPException):
     code = 415
 
 
+class BlockedFileExtensionError(BaseHTTPException):
+    error_code = "file_extension_blocked"
+    description = "The file extension is blocked for security reasons."
+    code = 400
+
+
 class TooManyFilesError(BaseHTTPException):
     error_code = "too_many_files"
     description = "Only one file is allowed."

api/controllers/console/__init__.py

@@ -58,11 +58,13 @@ from .app import (
     mcp_server,
     message,
     model_config,
+    online_user,
     ops_trace,
     site,
     statistic,
     workflow,
     workflow_app_log,
+    workflow_comment,
     workflow_draft_variable,
     workflow_run,
     workflow_statistic,
@@ -106,10 +108,12 @@ from .datasets.rag_pipeline import (
 
 # Import explore controllers
 from .explore import (
+    banner,
     installed_app,
     parameter,
     recommended_app,
     saved_message,
+    trial,
 )
 
 # Import tag controllers
@@ -143,6 +147,7 @@ __all__ = [
     "apikey",
     "app",
     "audio",
+    "banner",
     "billing",
     "bp",
     "completion",
@@ -196,6 +201,7 @@ __all__ = [
     "statistic",
     "tags",
     "tool_providers",
+    "trial",
     "version",
     "website",
     "workflow",

api/controllers/console/admin.py

@@ -16,7 +16,7 @@ from controllers.console import api, console_ns
 from controllers.console.wraps import only_edition_cloud
 from extensions.ext_database import db
 from libs.token import extract_access_token
-from models.model import App, InstalledApp, RecommendedApp
+from models.model import App, ExporleBanner, InstalledApp, RecommendedApp, TrialApp
 
 
 def admin_required(view: Callable[P, R]):
@@ -52,6 +52,8 @@ class InsertExploreAppListApi(Resource):
                 "language": fields.String(required=True, description="Language code"),
                 "category": fields.String(required=True, description="App category"),
                 "position": fields.Integer(required=True, description="Display position"),
+                "can_trial": fields.Boolean(required=True, description="Can trial"),
+                "trial_limit": fields.Integer(required=True, description="Trial limit"),
             },
         )
     )
@@ -71,6 +73,8 @@ class InsertExploreAppListApi(Resource):
             .add_argument("language", type=supported_language, required=True, nullable=False, location="json")
             .add_argument("category", type=str, required=True, nullable=False, location="json")
             .add_argument("position", type=int, required=True, nullable=False, location="json")
+            .add_argument("can_trial", type=bool, required=True, nullable=False, location="json")
+            .add_argument("trial_limit", type=int, required=True, nullable=False, location="json")
         )
         args = parser.parse_args()
 
@@ -108,6 +112,20 @@ class InsertExploreAppListApi(Resource):
                 )
 
                 db.session.add(recommended_app)
+                if args["can_trial"]:
+                    trial_app = db.session.execute(
+                        select(TrialApp).where(TrialApp.app_id == args["app_id"])
+                    ).scalar_one_or_none()
+                    if not trial_app:
+                        db.session.add(
+                            TrialApp(
+                                app_id=args["app_id"],
+                                tenant_id=app.tenant_id,
+                                trial_limit=args["trial_limit"],
+                            )
+                        )
+                    else:
+                        trial_app.trial_limit = args["trial_limit"]
 
                 app.is_public = True
                 db.session.commit()
@@ -122,6 +140,20 @@ class InsertExploreAppListApi(Resource):
                 recommended_app.category = args["category"]
                 recommended_app.position = args["position"]
 
+                if args["can_trial"]:
+                    trial_app = db.session.execute(
+                        select(TrialApp).where(TrialApp.app_id == args["app_id"])
+                    ).scalar_one_or_none()
+                    if not trial_app:
+                        db.session.add(
+                            TrialApp(
+                                app_id=args["app_id"],
+                                tenant_id=app.tenant_id,
+                                trial_limit=args["trial_limit"],
+                            )
+                        )
+                    else:
+                        trial_app.trial_limit = args["trial_limit"]
                 app.is_public = True
 
                 db.session.commit()
@@ -167,7 +199,83 @@ class InsertExploreAppApi(Resource):
             for installed_app in installed_apps:
                 session.delete(installed_app)
 
+            trial_app = session.execute(
+                select(TrialApp).where(TrialApp.app_id == recommended_app.app_id)
+            ).scalar_one_or_none()
+            if trial_app:
+                session.delete(trial_app)
+
         db.session.delete(recommended_app)
         db.session.commit()
 
         return {"result": "success"}, 204
+
+
+@console_ns.route("/admin/insert-explore-banner")
+class InsertExploreBanner(Resource):
+    @api.doc("insert_explore_banner")
+    @api.doc(description="Insert an explore banner")
+    @api.expect(
+        api.model(
+            "InsertExploreBannerRequest",
+            {
+                "content": fields.String(required=True, description="Banner content"),
+                "link": fields.String(required=True, description="Banner link"),
+                "sort": fields.Integer(required=True, description="Banner sort"),
+            },
+        )
+    )
+    @api.response(200, "Banner inserted successfully")
+    @admin_required
+    @only_edition_cloud
+    def post(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument("category", type=str, required=True, nullable=False, location="json")
+        parser.add_argument("title", type=str, required=True, nullable=False, location="json")
+        parser.add_argument("description", type=str, required=True, nullable=False, location="json")
+        parser.add_argument("img-src", type=str, required=True, nullable=False, location="json")
+
+        parser.add_argument("language", type=str, required=True, nullable=False, location="json")
+        parser.add_argument("link", type=str, required=True, nullable=False, location="json")
+        parser.add_argument("sort", type=int, required=True, nullable=False, location="json")
+
+        args = parser.parse_args()
+
+        content = {
+            "category": args["category"],
+            "title": args["title"],
+            "description": args["description"],
+            "img-src": args["img-src"],
+        }
+
+        if not args["language"]:
+            args["language"] = "en-US"
+
+        banner = ExporleBanner(
+            content=content,
+            link=args["link"],
+            sort=args["sort"],
+            language=args["language"],
+        )
+        db.session.add(banner)
+        db.session.commit()
+
+        return {"result": "success"}, 200
+
+
+@console_ns.route("/admin/delete-explore-banner/<uuid:banner_id>")
+class DeleteExploreBanner(Resource):
+    @api.doc("delete_explore_banner")
+    @api.doc(description="Delete an explore banner")
+    @api.response(204, "Banner deleted successfully")
+    @admin_required
+    @only_edition_cloud
+    def delete(self, banner_id):
+        banner = db.session.execute(select(ExporleBanner).where(ExporleBanner.id == banner_id)).scalar_one_or_none()
+        if not banner:
+            raise NotFound(f"Banner '{banner_id}' is not found")
+
+        db.session.delete(banner)
+        db.session.commit()
+
+        return {"result": "success"}, 204

api/controllers/console/app/annotation.py

@@ -16,6 +16,7 @@ from fields.annotation_fields import (
     annotation_fields,
     annotation_hit_history_fields,
 )
+from libs.helper import uuid_value
 from libs.login import login_required
 from services.annotation_service import AppAnnotationService
 
@@ -175,8 +176,10 @@ class AnnotationApi(Resource):
         api.model(
             "CreateAnnotationRequest",
             {
-                "question": fields.String(required=True, description="Question text"),
-                "answer": fields.String(required=True, description="Answer text"),
+                "message_id": fields.String(description="Message ID (optional)"),
+                "question": fields.String(description="Question text (required when message_id not provided)"),
+                "answer": fields.String(description="Answer text (use 'answer' or 'content')"),
+                "content": fields.String(description="Content text (use 'answer' or 'content')"),
                 "annotation_reply": fields.Raw(description="Annotation reply data"),
             },
         )
@@ -193,11 +196,14 @@ class AnnotationApi(Resource):
         app_id = str(app_id)
         parser = (
             reqparse.RequestParser()
-            .add_argument("question", required=True, type=str, location="json")
-            .add_argument("answer", required=True, type=str, location="json")
+            .add_argument("message_id", required=False, type=uuid_value, location="json")
+            .add_argument("question", required=False, type=str, location="json")
+            .add_argument("answer", required=False, type=str, location="json")
+            .add_argument("content", required=False, type=str, location="json")
+            .add_argument("annotation_reply", required=False, type=dict, location="json")
         )
         args = parser.parse_args()
-        annotation = AppAnnotationService.insert_app_annotation_directly(args, app_id)
+        annotation = AppAnnotationService.up_insert_app_annotation_from_message(args, app_id)
         return annotation
 
     @setup_required

api/controllers/console/app/conversation.py

@@ -1,7 +1,5 @@
-from datetime import datetime
-
-import pytz
 import sqlalchemy as sa
+from flask import abort
 from flask_restx import Resource, marshal_with, reqparse
 from flask_restx.inputs import int_range
 from sqlalchemy import func, or_
@@ -19,7 +17,7 @@ from fields.conversation_fields import (
     conversation_pagination_fields,
     conversation_with_summary_pagination_fields,
 )
-from libs.datetime_utils import naive_utc_now
+from libs.datetime_utils import naive_utc_now, parse_time_range
 from libs.helper import DatetimeString
 from libs.login import current_account_with_tenant, login_required
 from models import Conversation, EndUser, Message, MessageAnnotation
@@ -90,25 +88,17 @@ class CompletionConversationApi(Resource):
 
         account = current_user
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             query = query.where(Conversation.created_at >= start_datetime_utc)
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=59)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
+            end_datetime_utc = end_datetime_utc.replace(second=59)
             query = query.where(Conversation.created_at < end_datetime_utc)
 
         # FIXME, the type ignore in this file
@@ -270,29 +260,21 @@ class ChatConversationApi(Resource):
 
         account = current_user
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             match args["sort_by"]:
                 case "updated_at" | "-updated_at":
                     query = query.where(Conversation.updated_at >= start_datetime_utc)
                 case "created_at" | "-created_at" | _:
                     query = query.where(Conversation.created_at >= start_datetime_utc)
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=59)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
+            end_datetime_utc = end_datetime_utc.replace(second=59)
             match args["sort_by"]:
                 case "updated_at" | "-updated_at":
                     query = query.where(Conversation.updated_at <= end_datetime_utc)

api/controllers/console/app/error.py

@@ -115,3 +115,9 @@ class InvokeRateLimitError(BaseHTTPException):
     error_code = "rate_limit_error"
     description = "Rate Limit Error"
     code = 429
+
+
+class NeedAddIdsError(BaseHTTPException):
+    error_code = "need_add_ids"
+    description = "Need to add ids."
+    code = 400

api/controllers/console/app/message.py

@@ -16,7 +16,6 @@ from controllers.console.app.wraps import get_app_model
 from controllers.console.explore.error import AppSuggestedQuestionsAfterAnswerDisabledError
 from controllers.console.wraps import (
     account_initialization_required,
-    cloud_edition_billing_resource_check,
     edit_permission_required,
     setup_required,
 )
@@ -24,12 +23,11 @@ from core.app.entities.app_invoke_entities import InvokeFrom
 from core.errors.error import ModelCurrentlyNotSupportError, ProviderTokenNotInitError, QuotaExceededError
 from core.model_runtime.errors.invoke import InvokeError
 from extensions.ext_database import db
-from fields.conversation_fields import annotation_fields, message_detail_fields
+from fields.conversation_fields import message_detail_fields
 from libs.helper import uuid_value
 from libs.infinite_scroll_pagination import InfiniteScrollPagination
 from libs.login import current_account_with_tenant, login_required
 from models.model import AppMode, Conversation, Message, MessageAnnotation, MessageFeedback
-from services.annotation_service import AppAnnotationService
 from services.errors.conversation import ConversationNotExistsError
 from services.errors.message import MessageNotExistsError, SuggestedQuestionsAfterAnswerDisabledError
 from services.message_service import MessageService
@@ -194,45 +192,6 @@ class MessageFeedbackApi(Resource):
         return {"result": "success"}
 
 
-@console_ns.route("/apps/<uuid:app_id>/annotations")
-class MessageAnnotationApi(Resource):
-    @api.doc("create_message_annotation")
-    @api.doc(description="Create message annotation")
-    @api.doc(params={"app_id": "Application ID"})
-    @api.expect(
-        api.model(
-            "MessageAnnotationRequest",
-            {
-                "message_id": fields.String(description="Message ID"),
-                "question": fields.String(required=True, description="Question text"),
-                "answer": fields.String(required=True, description="Answer text"),
-                "annotation_reply": fields.Raw(description="Annotation reply"),
-            },
-        )
-    )
-    @api.response(200, "Annotation created successfully", annotation_fields)
-    @api.response(403, "Insufficient permissions")
-    @marshal_with(annotation_fields)
-    @get_app_model
-    @setup_required
-    @login_required
-    @cloud_edition_billing_resource_check("annotation")
-    @account_initialization_required
-    @edit_permission_required
-    def post(self, app_model):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("message_id", required=False, type=uuid_value, location="json")
-            .add_argument("question", required=True, type=str, location="json")
-            .add_argument("answer", required=True, type=str, location="json")
-            .add_argument("annotation_reply", required=False, type=dict, location="json")
-        )
-        args = parser.parse_args()
-        annotation = AppAnnotationService.up_insert_app_annotation_from_message(args, app_model.id)
-
-        return annotation
-
-
 @console_ns.route("/apps/<uuid:app_id>/annotations/count")
 class MessageAnnotationCountApi(Resource):
     @api.doc("get_annotation_count")

api/controllers/console/app/online_user.py

@@ -0,0 +1,339 @@
+import json
+import time
+
+from werkzeug.wrappers import Request as WerkzeugRequest
+
+from extensions.ext_redis import redis_client
+from extensions.ext_socketio import sio
+from libs.passport import PassportService
+from libs.token import extract_access_token
+from services.account_service import AccountService
+
+SESSION_STATE_TTL_SECONDS = 3600
+WORKFLOW_ONLINE_USERS_PREFIX = "workflow_online_users:"
+WORKFLOW_LEADER_PREFIX = "workflow_leader:"
+WS_SID_MAP_PREFIX = "ws_sid_map:"
+
+
+def _workflow_key(workflow_id: str) -> str:
+    return f"{WORKFLOW_ONLINE_USERS_PREFIX}{workflow_id}"
+
+
+def _leader_key(workflow_id: str) -> str:
+    return f"{WORKFLOW_LEADER_PREFIX}{workflow_id}"
+
+
+def _sid_key(sid: str) -> str:
+    return f"{WS_SID_MAP_PREFIX}{sid}"
+
+
+def _refresh_session_state(workflow_id: str, sid: str) -> None:
+    """
+    Refresh TTLs for workflow + session keys so healthy sessions do not linger forever after crashes.
+    """
+    workflow_key = _workflow_key(workflow_id)
+    sid_key = _sid_key(sid)
+    if redis_client.exists(workflow_key):
+        redis_client.expire(workflow_key, SESSION_STATE_TTL_SECONDS)
+    if redis_client.exists(sid_key):
+        redis_client.expire(sid_key, SESSION_STATE_TTL_SECONDS)
+
+
+@sio.on("connect")
+def socket_connect(sid, environ, auth):
+    """
+    WebSocket connect event, do authentication here.
+    """
+    token = None
+    if auth and isinstance(auth, dict):
+        token = auth.get("token")
+
+    if not token:
+        try:
+            request_environ = WerkzeugRequest(environ)
+            token = extract_access_token(request_environ)
+        except Exception:
+            token = None
+
+    if not token:
+        return False
+
+    try:
+        decoded = PassportService().verify(token)
+        user_id = decoded.get("user_id")
+        if not user_id:
+            return False
+
+        with sio.app.app_context():
+            user = AccountService.load_logged_in_account(account_id=user_id)
+            if not user:
+                return False
+
+            sio.save_session(sid, {"user_id": user.id, "username": user.name, "avatar": user.avatar})
+
+            return True
+
+    except Exception:
+        return False
+
+
+@sio.on("user_connect")
+def handle_user_connect(sid, data):
+    """
+    Handle user connect event. Each session (tab) is treated as an independent collaborator.
+    """
+
+    workflow_id = data.get("workflow_id")
+    if not workflow_id:
+        return {"msg": "workflow_id is required"}, 400
+
+    session = sio.get_session(sid)
+    user_id = session.get("user_id")
+
+    if not user_id:
+        return {"msg": "unauthorized"}, 401
+
+    # Each session is stored independently with sid as key
+    session_info = {
+        "user_id": user_id,
+        "username": session.get("username", "Unknown"),
+        "avatar": session.get("avatar", None),
+        "sid": sid,
+        "connected_at": int(time.time()),  # Add timestamp to differentiate tabs
+    }
+
+    workflow_key = _workflow_key(workflow_id)
+    # Store session info with sid as key
+    redis_client.hset(workflow_key, sid, json.dumps(session_info))
+    redis_client.set(
+        _sid_key(sid),
+        json.dumps({"workflow_id": workflow_id, "user_id": user_id}),
+        ex=SESSION_STATE_TTL_SECONDS,
+    )
+    _refresh_session_state(workflow_id, sid)
+
+    # Leader election: first session becomes the leader
+    leader_sid = get_or_set_leader(workflow_id, sid)
+    is_leader = leader_sid == sid
+
+    sio.enter_room(sid, workflow_id)
+    broadcast_online_users(workflow_id)
+
+    # Notify this session of their leader status
+    sio.emit("status", {"isLeader": is_leader}, room=sid)
+
+    return {"msg": "connected", "user_id": user_id, "sid": sid, "isLeader": is_leader}
+
+
+@sio.on("disconnect")
+def handle_disconnect(sid):
+    """
+    Handle session disconnect event. Remove the specific session from online users.
+    """
+    mapping = redis_client.get(_sid_key(sid))
+    if mapping:
+        data = json.loads(mapping)
+        workflow_id = data["workflow_id"]
+
+        # Remove this specific session
+        redis_client.hdel(_workflow_key(workflow_id), sid)
+        redis_client.delete(_sid_key(sid))
+
+        # Handle leader re-election if the leader session disconnected
+        handle_leader_disconnect(workflow_id, sid)
+
+        broadcast_online_users(workflow_id)
+
+
+def _clear_session_state(workflow_id: str, sid: str) -> None:
+    redis_client.hdel(_workflow_key(workflow_id), sid)
+    redis_client.delete(_sid_key(sid))
+
+
+def _is_session_active(workflow_id: str, sid: str) -> bool:
+    if not sid:
+        return False
+
+    try:
+        if not sio.manager.is_connected(sid, "/"):
+            return False
+    except AttributeError:
+        return False
+
+    if not redis_client.hexists(_workflow_key(workflow_id), sid):
+        return False
+
+    if not redis_client.exists(_sid_key(sid)):
+        return False
+
+    return True
+
+
+def get_or_set_leader(workflow_id: str, sid: str) -> str:
+    """
+    Get current leader session or set this session as leader if no valid leader exists.
+    Returns the leader session id (sid).
+    """
+    raw_leader = redis_client.get(_leader_key(workflow_id))
+    current_leader = raw_leader.decode("utf-8") if isinstance(raw_leader, bytes) else raw_leader
+    leader_replaced = False
+
+    if current_leader and not _is_session_active(workflow_id, current_leader):
+        _clear_session_state(workflow_id, current_leader)
+        redis_client.delete(_leader_key(workflow_id))
+        current_leader = None
+        leader_replaced = True
+
+    if not current_leader:
+        redis_client.set(_leader_key(workflow_id), sid, ex=SESSION_STATE_TTL_SECONDS)  # Expire in 1 hour
+        if leader_replaced:
+            broadcast_leader_change(workflow_id, sid)
+        return sid
+
+    return current_leader
+
+
+def handle_leader_disconnect(workflow_id, disconnected_sid):
+    """
+    Handle leader re-election when a session disconnects.
+    If the disconnected session was the leader, elect a new leader from remaining sessions.
+    """
+    current_leader = redis_client.get(_leader_key(workflow_id))
+
+    if current_leader:
+        current_leader = current_leader.decode("utf-8") if isinstance(current_leader, bytes) else current_leader
+
+        if current_leader == disconnected_sid:
+            # Leader session disconnected, elect a new leader
+            sessions_json = redis_client.hgetall(_workflow_key(workflow_id))
+
+            if sessions_json:
+                # Get the first remaining session as new leader
+                new_leader_sid = list(sessions_json.keys())[0]
+                if isinstance(new_leader_sid, bytes):
+                    new_leader_sid = new_leader_sid.decode("utf-8")
+
+                redis_client.set(_leader_key(workflow_id), new_leader_sid, ex=SESSION_STATE_TTL_SECONDS)
+
+                # Notify all sessions about the new leader
+                broadcast_leader_change(workflow_id, new_leader_sid)
+            else:
+                # No sessions left, remove leader
+                redis_client.delete(_leader_key(workflow_id))
+
+
+def broadcast_leader_change(workflow_id, new_leader_sid):
+    """
+    Broadcast leader change to all sessions in the workflow.
+    """
+    sessions_json = redis_client.hgetall(_workflow_key(workflow_id))
+
+    for sid, session_info_json in sessions_json.items():
+        try:
+            sid_str = sid.decode("utf-8") if isinstance(sid, bytes) else sid
+            is_leader = sid_str == new_leader_sid
+            # Emit to each session whether they are the new leader
+            sio.emit("status", {"isLeader": is_leader}, room=sid_str)
+        except Exception:
+            continue
+
+
+def get_current_leader(workflow_id):
+    """
+    Get the current leader for a workflow.
+    """
+    leader = redis_client.get(_leader_key(workflow_id))
+    return leader.decode("utf-8") if leader and isinstance(leader, bytes) else leader
+
+
+def broadcast_online_users(workflow_id):
+    """
+    Broadcast online users to the workflow room.
+    Each session is shown as a separate user (even if same person has multiple tabs).
+    """
+    sessions_json = redis_client.hgetall(_workflow_key(workflow_id))
+    users = []
+
+    for sid, session_info_json in sessions_json.items():
+        try:
+            session_info = json.loads(session_info_json)
+            # Each session appears as a separate "user" in the UI
+            users.append(
+                {
+                    "user_id": session_info["user_id"],
+                    "username": session_info["username"],
+                    "avatar": session_info.get("avatar"),
+                    "sid": session_info["sid"],
+                    "connected_at": session_info.get("connected_at"),
+                }
+            )
+        except Exception:
+            continue
+
+    # Sort by connection time to maintain consistent order
+    users.sort(key=lambda x: x.get("connected_at") or 0)
+
+    # Get current leader session
+    leader_sid = get_current_leader(workflow_id)
+
+    sio.emit("online_users", {"workflow_id": workflow_id, "users": users, "leader": leader_sid}, room=workflow_id)
+
+
+@sio.on("collaboration_event")
+def handle_collaboration_event(sid, data):
+    """
+    Handle general collaboration events, include:
+    1. mouse_move
+    2. vars_and_features_update
+    3. sync_request (ask leader to update graph)
+    4. app_state_update
+    5. mcp_server_update
+    6. workflow_update
+    7. comments_update
+    8. node_panel_presence
+
+    """
+    mapping = redis_client.get(_sid_key(sid))
+
+    if not mapping:
+        return {"msg": "unauthorized"}, 401
+
+    mapping_data = json.loads(mapping)
+    workflow_id = mapping_data["workflow_id"]
+    user_id = mapping_data["user_id"]
+    _refresh_session_state(workflow_id, sid)
+
+    event_type = data.get("type")
+    event_data = data.get("data")
+    timestamp = data.get("timestamp", int(time.time()))
+
+    if not event_type:
+        return {"msg": "invalid event type"}, 400
+
+    sio.emit(
+        "collaboration_update",
+        {"type": event_type, "userId": user_id, "data": event_data, "timestamp": timestamp},
+        room=workflow_id,
+        skip_sid=sid,
+    )
+
+    return {"msg": "event_broadcasted"}
+
+
+@sio.on("graph_event")
+def handle_graph_event(sid, data):
+    """
+    Handle graph events - simple broadcast relay.
+    """
+    mapping = redis_client.get(_sid_key(sid))
+
+    if not mapping:
+        return {"msg": "unauthorized"}, 401
+
+    mapping_data = json.loads(mapping)
+    workflow_id = mapping_data["workflow_id"]
+    _refresh_session_state(workflow_id, sid)
+
+    sio.emit("graph_update", data, room=workflow_id, skip_sid=sid)
+
+    return {"msg": "graph_update_broadcasted"}

api/controllers/console/app/statistic.py

@@ -1,9 +1,7 @@
-from datetime import datetime
 from decimal import Decimal
 
-import pytz
 import sqlalchemy as sa
-from flask import jsonify
+from flask import abort, jsonify
 from flask_restx import Resource, fields, reqparse
 
 from controllers.console import api, console_ns
@@ -11,6 +9,7 @@ from controllers.console.app.wraps import get_app_model
 from controllers.console.wraps import account_initialization_required, setup_required
 from core.app.entities.app_invoke_entities import InvokeFrom
 from extensions.ext_database import db
+from libs.datetime_utils import parse_time_range
 from libs.helper import DatetimeString
 from libs.login import current_account_with_tenant, login_required
 from models import AppMode, Message
@@ -56,26 +55,16 @@ WHERE
         arg_dict = {"tz": account.timezone, "app_id": app_model.id, "invoke_from": InvokeFrom.DEBUGGER}
         assert account.timezone is not None
 
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
-
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             sql_query += " AND created_at >= :start"
             arg_dict["start"] = start_datetime_utc
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
             sql_query += " AND created_at < :end"
             arg_dict["end"] = end_datetime_utc
 
@@ -120,8 +109,11 @@ class DailyConversationStatistic(Resource):
         )
         args = parser.parse_args()
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
+
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
         stmt = (
             sa.select(
@@ -134,18 +126,10 @@ class DailyConversationStatistic(Resource):
             .where(Message.app_id == app_model.id, Message.invoke_from != InvokeFrom.DEBUGGER)
         )
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        if start_datetime_utc:
             stmt = stmt.where(Message.created_at >= start_datetime_utc)
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
+        if end_datetime_utc:
             stmt = stmt.where(Message.created_at < end_datetime_utc)
 
         stmt = stmt.group_by("date").order_by("date")
@@ -198,26 +182,17 @@ WHERE
     AND invoke_from != :invoke_from"""
         arg_dict = {"tz": account.timezone, "app_id": app_model.id, "invoke_from": InvokeFrom.DEBUGGER}
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             sql_query += " AND created_at >= :start"
             arg_dict["start"] = start_datetime_utc
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
             sql_query += " AND created_at < :end"
             arg_dict["end"] = end_datetime_utc
 
@@ -273,26 +248,17 @@ WHERE
     AND invoke_from != :invoke_from"""
         arg_dict = {"tz": account.timezone, "app_id": app_model.id, "invoke_from": InvokeFrom.DEBUGGER}
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             sql_query += " AND created_at >= :start"
             arg_dict["start"] = start_datetime_utc
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
             sql_query += " AND created_at < :end"
             arg_dict["end"] = end_datetime_utc
 
@@ -357,26 +323,17 @@ FROM
             AND m.invoke_from != :invoke_from"""
         arg_dict = {"tz": account.timezone, "app_id": app_model.id, "invoke_from": InvokeFrom.DEBUGGER}
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             sql_query += " AND c.created_at >= :start"
             arg_dict["start"] = start_datetime_utc
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
             sql_query += " AND c.created_at < :end"
             arg_dict["end"] = end_datetime_utc
 
@@ -446,26 +403,17 @@ WHERE
     AND m.invoke_from != :invoke_from"""
         arg_dict = {"tz": account.timezone, "app_id": app_model.id, "invoke_from": InvokeFrom.DEBUGGER}
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             sql_query += " AND m.created_at >= :start"
             arg_dict["start"] = start_datetime_utc
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
             sql_query += " AND m.created_at < :end"
             arg_dict["end"] = end_datetime_utc
 
@@ -525,26 +473,17 @@ WHERE
     AND invoke_from != :invoke_from"""
         arg_dict = {"tz": account.timezone, "app_id": app_model.id, "invoke_from": InvokeFrom.DEBUGGER}
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             sql_query += " AND created_at >= :start"
             arg_dict["start"] = start_datetime_utc
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
             sql_query += " AND created_at < :end"
             arg_dict["end"] = end_datetime_utc
 
@@ -602,26 +541,17 @@ WHERE
     AND invoke_from != :invoke_from"""
         arg_dict = {"tz": account.timezone, "app_id": app_model.id, "invoke_from": InvokeFrom.DEBUGGER}
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_datetime_utc = start_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_datetime_utc, end_datetime_utc = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
+        if start_datetime_utc:
             sql_query += " AND created_at >= :start"
             arg_dict["start"] = start_datetime_utc
 
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_datetime_utc = end_datetime_timezone.astimezone(utc_timezone)
-
+        if end_datetime_utc:
             sql_query += " AND created_at < :end"
             arg_dict["end"] = end_datetime_utc
 

api/controllers/console/app/workflow.py

@@ -5,10 +5,12 @@ from typing import cast
 
 from flask import abort, request
 from flask_restx import Resource, fields, inputs, marshal_with, reqparse
+from pydantic_core import ValidationError
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import Forbidden, InternalServerError, NotFound
 
 import services
+from configs import dify_config
 from controllers.console import api, console_ns
 from controllers.console.app.error import ConversationCompletedError, DraftWorkflowNotExist, DraftWorkflowNotSync
 from controllers.console.app.wraps import get_app_model
@@ -21,7 +23,9 @@ from core.file.models import File
 from core.helper.trace_id_helper import get_external_trace_id
 from core.workflow.graph_engine.manager import GraphEngineManager
 from extensions.ext_database import db
+from extensions.ext_redis import redis_client
 from factories import file_factory, variable_factory
+from fields.online_user_fields import online_user_list_fields
 from fields.workflow_fields import workflow_fields, workflow_pagination_fields
 from fields.workflow_run_fields import workflow_run_node_execution_fields
 from libs import helper
@@ -99,10 +103,22 @@ class DraftWorkflowApi(Resource):
                 "hash": fields.String(description="Workflow hash for validation"),
                 "environment_variables": fields.List(fields.Raw, required=True, description="Environment variables"),
                 "conversation_variables": fields.List(fields.Raw, description="Conversation variables"),
+                "memory_blocks": fields.List(fields.Raw, description="Memory blocks"),
             },
         )
     )
-    @api.response(200, "Draft workflow synced successfully", workflow_fields)
+    @api.response(
+        200,
+        "Draft workflow synced successfully",
+        api.model(
+            "SyncDraftWorkflowResponse",
+            {
+                "result": fields.String,
+                "hash": fields.String,
+                "updated_at": fields.String,
+            },
+        ),
+    )
     @api.response(400, "Invalid workflow configuration")
     @api.response(403, "Permission denied")
     @edit_permission_required
@@ -122,6 +138,8 @@ class DraftWorkflowApi(Resource):
                 .add_argument("hash", type=str, required=False, location="json")
                 .add_argument("environment_variables", type=list, required=True, location="json")
                 .add_argument("conversation_variables", type=list, required=False, location="json")
+                .add_argument("force_upload", type=bool, required=False, default=False, location="json")
+                .add_argument("memory_blocks", type=list, required=False, location="json")
             )
             args = parser.parse_args()
         elif "text/plain" in content_type:
@@ -139,6 +157,8 @@ class DraftWorkflowApi(Resource):
                     "hash": data.get("hash"),
                     "environment_variables": data.get("environment_variables"),
                     "conversation_variables": data.get("conversation_variables"),
+                    "memory_blocks": data.get("memory_blocks"),
+                    "force_upload": data.get("force_upload", False),
                 }
             except json.JSONDecodeError:
                 return {"message": "Invalid JSON data"}, 400
@@ -155,6 +175,10 @@ class DraftWorkflowApi(Resource):
             conversation_variables = [
                 variable_factory.build_conversation_variable_from_mapping(obj) for obj in conversation_variables_list
             ]
+            memory_blocks_list = args.get("memory_blocks") or []
+            from core.memory.entities import MemoryBlockSpec
+
+            memory_blocks = [MemoryBlockSpec.model_validate(obj) for obj in memory_blocks_list]
             workflow = workflow_service.sync_draft_workflow(
                 app_model=app_model,
                 graph=args["graph"],
@@ -163,9 +187,13 @@ class DraftWorkflowApi(Resource):
                 account=current_user,
                 environment_variables=environment_variables,
                 conversation_variables=conversation_variables,
+                force_upload=args.get("force_upload", False),
+                memory_blocks=memory_blocks,
             )
         except WorkflowHashNotEqualError:
             raise DraftWorkflowNotSync()
+        except ValidationError as e:
+            return {"message": str(e)}, 400
 
         return {
             "result": "success",
@@ -734,6 +762,45 @@ class ConvertToWorkflowApi(Resource):
         }
 
 
+@console_ns.route("/apps/<uuid:app_id>/workflows/draft/config")
+class WorkflowConfigApi(Resource):
+    """Resource for workflow configuration."""
+
+    @api.doc("get_workflow_config")
+    @api.doc(description="Get workflow configuration")
+    @api.doc(params={"app_id": "Application ID"})
+    @api.response(200, "Workflow configuration retrieved successfully")
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @get_app_model(mode=[AppMode.ADVANCED_CHAT, AppMode.WORKFLOW])
+    def get(self, app_model: App):
+        return {
+            "parallel_depth_limit": dify_config.WORKFLOW_PARALLEL_DEPTH_LIMIT,
+        }
+
+
+class WorkflowFeaturesApi(Resource):
+    """Update draft workflow features."""
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @get_app_model(mode=[AppMode.ADVANCED_CHAT, AppMode.WORKFLOW])
+    def post(self, app_model: App):
+        parser = reqparse.RequestParser()
+        parser.add_argument("features", type=dict, required=True, location="json")
+        args = parser.parse_args()
+
+        features = args.get("features")
+
+        # Update draft workflow features
+        workflow_service = WorkflowService()
+        workflow_service.update_draft_workflow_features(app_model=app_model, features=features, account=current_user)
+
+        return {"result": "success"}
+
+
 @console_ns.route("/apps/<uuid:app_id>/workflows")
 class PublishedAllWorkflowApi(Resource):
     @api.doc("get_all_published_workflows")
@@ -915,3 +982,105 @@ class DraftWorkflowNodeLastRunApi(Resource):
         if node_exec is None:
             raise NotFound("last run not found")
         return node_exec
+
+
+class WorkflowOnlineUsersApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(online_user_list_fields)
+    def get(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument("workflow_ids", type=str, required=True, location="args")
+        args = parser.parse_args()
+
+        workflow_ids = [id.strip() for id in args["workflow_ids"].split(",")]
+
+        results = []
+        for workflow_id in workflow_ids:
+            users_json = redis_client.hgetall(f"workflow_online_users:{workflow_id}")
+
+            users = []
+            for _, user_info_json in users_json.items():
+                try:
+                    users.append(json.loads(user_info_json))
+                except Exception:
+                    continue
+            results.append({"workflow_id": workflow_id, "users": users})
+
+        return {"data": results}
+
+
+api.add_resource(
+    DraftWorkflowApi,
+    "/apps/<uuid:app_id>/workflows/draft",
+)
+api.add_resource(
+    WorkflowConfigApi,
+    "/apps/<uuid:app_id>/workflows/draft/config",
+)
+api.add_resource(
+    WorkflowFeaturesApi,
+    "/apps/<uuid:app_id>/workflows/draft/features",
+)
+api.add_resource(
+    AdvancedChatDraftWorkflowRunApi,
+    "/apps/<uuid:app_id>/advanced-chat/workflows/draft/run",
+)
+api.add_resource(
+    DraftWorkflowRunApi,
+    "/apps/<uuid:app_id>/workflows/draft/run",
+)
+api.add_resource(
+    WorkflowTaskStopApi,
+    "/apps/<uuid:app_id>/workflow-runs/tasks/<string:task_id>/stop",
+)
+api.add_resource(
+    DraftWorkflowNodeRunApi,
+    "/apps/<uuid:app_id>/workflows/draft/nodes/<string:node_id>/run",
+)
+api.add_resource(
+    AdvancedChatDraftRunIterationNodeApi,
+    "/apps/<uuid:app_id>/advanced-chat/workflows/draft/iteration/nodes/<string:node_id>/run",
+)
+api.add_resource(
+    WorkflowDraftRunIterationNodeApi,
+    "/apps/<uuid:app_id>/workflows/draft/iteration/nodes/<string:node_id>/run",
+)
+api.add_resource(
+    AdvancedChatDraftRunLoopNodeApi,
+    "/apps/<uuid:app_id>/advanced-chat/workflows/draft/loop/nodes/<string:node_id>/run",
+)
+api.add_resource(
+    WorkflowDraftRunLoopNodeApi,
+    "/apps/<uuid:app_id>/workflows/draft/loop/nodes/<string:node_id>/run",
+)
+api.add_resource(
+    PublishedWorkflowApi,
+    "/apps/<uuid:app_id>/workflows/publish",
+)
+api.add_resource(
+    PublishedAllWorkflowApi,
+    "/apps/<uuid:app_id>/workflows",
+)
+api.add_resource(
+    DefaultBlockConfigsApi,
+    "/apps/<uuid:app_id>/workflows/default-workflow-block-configs",
+)
+api.add_resource(
+    DefaultBlockConfigApi,
+    "/apps/<uuid:app_id>/workflows/default-workflow-block-configs/<string:block_type>",
+)
+api.add_resource(
+    ConvertToWorkflowApi,
+    "/apps/<uuid:app_id>/convert-to-workflow",
+)
+api.add_resource(
+    WorkflowByIdApi,
+    "/apps/<uuid:app_id>/workflows/<string:workflow_id>",
+)
+api.add_resource(
+    DraftWorkflowNodeLastRunApi,
+    "/apps/<uuid:app_id>/workflows/draft/nodes/<string:node_id>/last-run",
+)
+api.add_resource(WorkflowOnlineUsersApi, "/apps/workflows/online-users")

api/controllers/console/app/workflow_comment.py

@@ -0,0 +1,240 @@
+import logging
+
+from flask_restx import Resource, fields, marshal_with, reqparse
+
+from controllers.console import api
+from controllers.console.app.wraps import get_app_model
+from controllers.console.wraps import account_initialization_required, setup_required
+from fields.member_fields import account_with_role_fields
+from fields.workflow_comment_fields import (
+    workflow_comment_basic_fields,
+    workflow_comment_create_fields,
+    workflow_comment_detail_fields,
+    workflow_comment_reply_create_fields,
+    workflow_comment_reply_update_fields,
+    workflow_comment_resolve_fields,
+    workflow_comment_update_fields,
+)
+from libs.login import current_user, login_required
+from models import App
+from services.account_service import TenantService
+from services.workflow_comment_service import WorkflowCommentService
+
+logger = logging.getLogger(__name__)
+
+
+class WorkflowCommentListApi(Resource):
+    """API for listing and creating workflow comments."""
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with(workflow_comment_basic_fields, envelope="data")
+    def get(self, app_model: App):
+        """Get all comments for a workflow."""
+        comments = WorkflowCommentService.get_comments(tenant_id=current_user.current_tenant_id, app_id=app_model.id)
+
+        return comments
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with(workflow_comment_create_fields)
+    def post(self, app_model: App):
+        """Create a new workflow comment."""
+        parser = reqparse.RequestParser()
+        parser.add_argument("position_x", type=float, required=True, location="json")
+        parser.add_argument("position_y", type=float, required=True, location="json")
+        parser.add_argument("content", type=str, required=True, location="json")
+        parser.add_argument("mentioned_user_ids", type=list, location="json", default=[])
+        args = parser.parse_args()
+
+        result = WorkflowCommentService.create_comment(
+            tenant_id=current_user.current_tenant_id,
+            app_id=app_model.id,
+            created_by=current_user.id,
+            content=args.content,
+            position_x=args.position_x,
+            position_y=args.position_y,
+            mentioned_user_ids=args.mentioned_user_ids,
+        )
+
+        return result, 201
+
+
+class WorkflowCommentDetailApi(Resource):
+    """API for managing individual workflow comments."""
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with(workflow_comment_detail_fields)
+    def get(self, app_model: App, comment_id: str):
+        """Get a specific workflow comment."""
+        comment = WorkflowCommentService.get_comment(
+            tenant_id=current_user.current_tenant_id, app_id=app_model.id, comment_id=comment_id
+        )
+
+        return comment
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with(workflow_comment_update_fields)
+    def put(self, app_model: App, comment_id: str):
+        """Update a workflow comment."""
+        parser = reqparse.RequestParser()
+        parser.add_argument("content", type=str, required=True, location="json")
+        parser.add_argument("position_x", type=float, required=False, location="json")
+        parser.add_argument("position_y", type=float, required=False, location="json")
+        parser.add_argument("mentioned_user_ids", type=list, location="json", default=[])
+        args = parser.parse_args()
+
+        result = WorkflowCommentService.update_comment(
+            tenant_id=current_user.current_tenant_id,
+            app_id=app_model.id,
+            comment_id=comment_id,
+            user_id=current_user.id,
+            content=args.content,
+            position_x=args.position_x,
+            position_y=args.position_y,
+            mentioned_user_ids=args.mentioned_user_ids,
+        )
+
+        return result
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    def delete(self, app_model: App, comment_id: str):
+        """Delete a workflow comment."""
+        WorkflowCommentService.delete_comment(
+            tenant_id=current_user.current_tenant_id,
+            app_id=app_model.id,
+            comment_id=comment_id,
+            user_id=current_user.id,
+        )
+
+        return {"result": "success"}, 204
+
+
+class WorkflowCommentResolveApi(Resource):
+    """API for resolving and reopening workflow comments."""
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with(workflow_comment_resolve_fields)
+    def post(self, app_model: App, comment_id: str):
+        """Resolve a workflow comment."""
+        comment = WorkflowCommentService.resolve_comment(
+            tenant_id=current_user.current_tenant_id,
+            app_id=app_model.id,
+            comment_id=comment_id,
+            user_id=current_user.id,
+        )
+
+        return comment
+
+
+class WorkflowCommentReplyApi(Resource):
+    """API for managing comment replies."""
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with(workflow_comment_reply_create_fields)
+    def post(self, app_model: App, comment_id: str):
+        """Add a reply to a workflow comment."""
+        # Validate comment access first
+        WorkflowCommentService.validate_comment_access(
+            comment_id=comment_id, tenant_id=current_user.current_tenant_id, app_id=app_model.id
+        )
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("content", type=str, required=True, location="json")
+        parser.add_argument("mentioned_user_ids", type=list, location="json", default=[])
+        args = parser.parse_args()
+
+        result = WorkflowCommentService.create_reply(
+            comment_id=comment_id,
+            content=args.content,
+            created_by=current_user.id,
+            mentioned_user_ids=args.mentioned_user_ids,
+        )
+
+        return result, 201
+
+
+class WorkflowCommentReplyDetailApi(Resource):
+    """API for managing individual comment replies."""
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with(workflow_comment_reply_update_fields)
+    def put(self, app_model: App, comment_id: str, reply_id: str):
+        """Update a comment reply."""
+        # Validate comment access first
+        WorkflowCommentService.validate_comment_access(
+            comment_id=comment_id, tenant_id=current_user.current_tenant_id, app_id=app_model.id
+        )
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("content", type=str, required=True, location="json")
+        parser.add_argument("mentioned_user_ids", type=list, location="json", default=[])
+        args = parser.parse_args()
+
+        reply = WorkflowCommentService.update_reply(
+            reply_id=reply_id, user_id=current_user.id, content=args.content, mentioned_user_ids=args.mentioned_user_ids
+        )
+
+        return reply
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    def delete(self, app_model: App, comment_id: str, reply_id: str):
+        """Delete a comment reply."""
+        # Validate comment access first
+        WorkflowCommentService.validate_comment_access(
+            comment_id=comment_id, tenant_id=current_user.current_tenant_id, app_id=app_model.id
+        )
+
+        WorkflowCommentService.delete_reply(reply_id=reply_id, user_id=current_user.id)
+
+        return {"result": "success"}, 204
+
+
+class WorkflowCommentMentionUsersApi(Resource):
+    """API for getting mentionable users for workflow comments."""
+
+    @login_required
+    @setup_required
+    @account_initialization_required
+    @get_app_model
+    @marshal_with({"users": fields.List(fields.Nested(account_with_role_fields))})
+    def get(self, app_model: App):
+        """Get all users in current tenant for mentions."""
+        members = TenantService.get_tenant_members(current_user.current_tenant)
+        return {"users": members}
+
+
+# Register API routes
+api.add_resource(WorkflowCommentListApi, "/apps/<uuid:app_id>/workflow/comments")
+api.add_resource(WorkflowCommentDetailApi, "/apps/<uuid:app_id>/workflow/comments/<string:comment_id>")
+api.add_resource(WorkflowCommentResolveApi, "/apps/<uuid:app_id>/workflow/comments/<string:comment_id>/resolve")
+api.add_resource(WorkflowCommentReplyApi, "/apps/<uuid:app_id>/workflow/comments/<string:comment_id>/replies")
+api.add_resource(
+    WorkflowCommentReplyDetailApi, "/apps/<uuid:app_id>/workflow/comments/<string:comment_id>/replies/<string:reply_id>"
+)
+api.add_resource(WorkflowCommentMentionUsersApi, "/apps/<uuid:app_id>/workflow/comments/mention-users")

api/controllers/console/app/workflow_draft_variable.py

@@ -19,8 +19,8 @@ from core.variables.segments import ArrayFileSegment, FileSegment, Segment
 from core.variables.types import SegmentType
 from core.workflow.constants import CONVERSATION_VARIABLE_NODE_ID, SYSTEM_VARIABLE_NODE_ID
 from extensions.ext_database import db
+from factories import variable_factory
 from factories.file_factory import build_from_mapping, build_from_mappings
-from factories.variable_factory import build_segment_with_type
 from libs.login import current_user, login_required
 from models import Account, App, AppMode
 from models.workflow import WorkflowDraftVariable
@@ -355,7 +355,7 @@ class VariableApi(Resource):
                 if len(raw_value) > 0 and not isinstance(raw_value[0], dict):
                     raise InvalidArgumentError(description=f"expected dict for files[0], got {type(raw_value)}")
                 raw_value = build_from_mappings(mappings=raw_value, tenant_id=app_model.tenant_id)
-            new_value = build_segment_with_type(variable.value_type, raw_value)
+            new_value = variable_factory.build_segment_with_type(variable.value_type, raw_value)
         draft_var_srv.update_variable(variable, name=new_name, value=new_value)
         db.session.commit()
         return variable
@@ -448,8 +448,35 @@ class ConversationVariableCollectionApi(Resource):
         db.session.commit()
         return _get_variable_list(app_model, CONVERSATION_VARIABLE_NODE_ID)
 
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @get_app_model(mode=AppMode.ADVANCED_CHAT)
+    def post(self, app_model: App):
+        # The role of the current user in the ta table must be admin, owner, or editor
+        if not current_user.is_editor:
+            raise Forbidden()
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("conversation_variables", type=list, required=True, location="json")
+        args = parser.parse_args()
+
+        workflow_service = WorkflowService()
+
+        conversation_variables_list = args.get("conversation_variables") or []
+        conversation_variables = [
+            variable_factory.build_conversation_variable_from_mapping(obj) for obj in conversation_variables_list
+        ]
+
+        workflow_service.update_draft_workflow_conversation_variables(
+            app_model=app_model,
+            account=current_user,
+            conversation_variables=conversation_variables,
+        )
+
+        return {"result": "success"}
+
 
-@console_ns.route("/apps/<uuid:app_id>/workflows/draft/system-variables")
 class SystemVariableCollectionApi(Resource):
     @api.doc("get_system_variables")
     @api.doc(description="Get system variables for workflow")
@@ -499,3 +526,44 @@ class EnvironmentVariableCollectionApi(Resource):
             )
 
         return {"items": env_vars_list}
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @get_app_model(mode=[AppMode.ADVANCED_CHAT, AppMode.WORKFLOW])
+    def post(self, app_model: App):
+        # The role of the current user in the ta table must be admin, owner, or editor
+        if not current_user.is_editor:
+            raise Forbidden()
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("environment_variables", type=list, required=True, location="json")
+        args = parser.parse_args()
+
+        workflow_service = WorkflowService()
+
+        environment_variables_list = args.get("environment_variables") or []
+        environment_variables = [
+            variable_factory.build_environment_variable_from_mapping(obj) for obj in environment_variables_list
+        ]
+
+        workflow_service.update_draft_workflow_environment_variables(
+            app_model=app_model,
+            account=current_user,
+            environment_variables=environment_variables,
+        )
+
+        return {"result": "success"}
+
+
+api.add_resource(
+    WorkflowVariableCollectionApi,
+    "/apps/<uuid:app_id>/workflows/draft/variables",
+)
+api.add_resource(NodeVariableCollectionApi, "/apps/<uuid:app_id>/workflows/draft/nodes/<string:node_id>/variables")
+api.add_resource(VariableApi, "/apps/<uuid:app_id>/workflows/draft/variables/<uuid:variable_id>")
+api.add_resource(VariableResetApi, "/apps/<uuid:app_id>/workflows/draft/variables/<uuid:variable_id>/reset")
+
+api.add_resource(ConversationVariableCollectionApi, "/apps/<uuid:app_id>/workflows/draft/conversation-variables")
+api.add_resource(SystemVariableCollectionApi, "/apps/<uuid:app_id>/workflows/draft/system-variables")
+api.add_resource(EnvironmentVariableCollectionApi, "/apps/<uuid:app_id>/workflows/draft/environment-variables")

api/controllers/console/app/workflow_statistic.py

@@ -1,7 +1,4 @@
-from datetime import datetime
-
-import pytz
-from flask import jsonify
+from flask import abort, jsonify
 from flask_restx import Resource, reqparse
 from sqlalchemy.orm import sessionmaker
 
@@ -9,6 +6,7 @@ from controllers.console import api, console_ns
 from controllers.console.app.wraps import get_app_model
 from controllers.console.wraps import account_initialization_required, setup_required
 from extensions.ext_database import db
+from libs.datetime_utils import parse_time_range
 from libs.helper import DatetimeString
 from libs.login import current_account_with_tenant, login_required
 from models.enums import WorkflowRunTriggeredFrom
@@ -43,23 +41,11 @@ class WorkflowDailyRunsStatistic(Resource):
         args = parser.parse_args()
 
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        start_date = None
-        end_date = None
-
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_date = start_datetime_timezone.astimezone(utc_timezone)
-
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_date = end_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_date, end_date = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
         response_data = self._workflow_run_repo.get_daily_runs_statistics(
             tenant_id=app_model.tenant_id,
@@ -100,23 +86,11 @@ class WorkflowDailyTerminalsStatistic(Resource):
         args = parser.parse_args()
 
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        start_date = None
-        end_date = None
-
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_date = start_datetime_timezone.astimezone(utc_timezone)
-
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_date = end_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_date, end_date = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
         response_data = self._workflow_run_repo.get_daily_terminals_statistics(
             tenant_id=app_model.tenant_id,
@@ -157,23 +131,11 @@ class WorkflowDailyTokenCostStatistic(Resource):
         args = parser.parse_args()
 
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        start_date = None
-        end_date = None
-
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_date = start_datetime_timezone.astimezone(utc_timezone)
-
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_date = end_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_date, end_date = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
         response_data = self._workflow_run_repo.get_daily_token_cost_statistics(
             tenant_id=app_model.tenant_id,
@@ -214,23 +176,11 @@ class WorkflowAverageAppInteractionStatistic(Resource):
         args = parser.parse_args()
 
         assert account.timezone is not None
-        timezone = pytz.timezone(account.timezone)
-        utc_timezone = pytz.utc
 
-        start_date = None
-        end_date = None
-
-        if args["start"]:
-            start_datetime = datetime.strptime(args["start"], "%Y-%m-%d %H:%M")
-            start_datetime = start_datetime.replace(second=0)
-            start_datetime_timezone = timezone.localize(start_datetime)
-            start_date = start_datetime_timezone.astimezone(utc_timezone)
-
-        if args["end"]:
-            end_datetime = datetime.strptime(args["end"], "%Y-%m-%d %H:%M")
-            end_datetime = end_datetime.replace(second=0)
-            end_datetime_timezone = timezone.localize(end_datetime)
-            end_date = end_datetime_timezone.astimezone(utc_timezone)
+        try:
+            start_date, end_date = parse_time_range(args["start"], args["end"], account.timezone)
+        except ValueError as e:
+            abort(400, description=str(e))
 
         response_data = self._workflow_run_repo.get_average_app_interaction_statistics(
             tenant_id=app_model.tenant_id,

api/controllers/console/app/wraps.py

@@ -23,6 +23,15 @@ def _load_app_model(app_id: str) -> App | None:
     return app_model
 
 
+def _load_app_model_with_trial(app_id: str) -> App | None:
+    app_model = (
+        db.session.query(App)
+        .where(App.id == app_id, App.status == "normal")
+        .first()
+    )
+    return app_model
+
+
 def get_app_model(view: Callable[P, R] | None = None, *, mode: Union[AppMode, list[AppMode], None] = None):
     def decorator(view_func: Callable[P1, R1]):
         @wraps(view_func)
@@ -62,3 +71,44 @@ def get_app_model(view: Callable[P, R] | None = None, *, mode: Union[AppMode, li
         return decorator
     else:
         return decorator(view)
+
+
+def get_app_model_with_trial(view: Callable[P, R] | None = None, *, mode: Union[AppMode, list[AppMode], None] = None):
+    def decorator(view_func: Callable[P, R]):
+        @wraps(view_func)
+        def decorated_view(*args: P.args, **kwargs: P.kwargs):
+            if not kwargs.get("app_id"):
+                raise ValueError("missing app_id in path parameters")
+
+            app_id = kwargs.get("app_id")
+            app_id = str(app_id)
+
+            del kwargs["app_id"]
+
+            app_model = _load_app_model_with_trial(app_id)
+
+            if not app_model:
+                raise AppNotFoundError()
+
+            app_mode = AppMode.value_of(app_model.mode)
+
+            if mode is not None:
+                if isinstance(mode, list):
+                    modes = mode
+                else:
+                    modes = [mode]
+
+                if app_mode not in modes:
+                    mode_values = {m.value for m in modes}
+                    raise AppNotFoundError(f"App mode is not in the supported list: {mode_values}")
+
+            kwargs["app_model"] = app_model
+
+            return view_func(*args, **kwargs)
+
+        return decorated_view
+
+    if view is None:
+        return decorator
+    else:
+        return decorator(view)

api/controllers/console/auth/login.py

@@ -25,10 +25,13 @@ from controllers.console.wraps import email_password_login_enabled, setup_requir
 from events.tenant_event import tenant_was_created
 from libs.helper import email, extract_remote_ip
 from libs.login import current_account_with_tenant
+from libs.passport import PassportService
 from libs.token import (
+    check_csrf_token,
     clear_access_token_from_cookie,
     clear_csrf_token_from_cookie,
     clear_refresh_token_from_cookie,
+    extract_access_token,
     extract_refresh_token,
     set_access_token_to_cookie,
     set_csrf_token_to_cookie,
@@ -289,3 +292,18 @@ class RefreshTokenApi(Resource):
             return response
         except Exception as e:
             return {"result": "fail", "message": str(e)}, 401
+
+
+# this api helps frontend to check whether user is authenticated
+# TODO: remove in the future. frontend should redirect to login page by catching 401 status
+@console_ns.route("/login/status")
+class LoginStatus(Resource):
+    def get(self):
+        token = extract_access_token(request)
+        res = True
+        try:
+            validated = PassportService().verify(token=token)
+            check_csrf_token(request=request, user_id=validated.get("user_id"))
+        except:
+            res = False
+        return {"logged_in": res}

api/controllers/console/billing/billing.py

@@ -2,6 +2,7 @@ from flask_restx import Resource, reqparse
 
 from controllers.console import console_ns
 from controllers.console.wraps import account_initialization_required, only_edition_cloud, setup_required
+from enums.cloud_plan import CloudPlan
 from libs.login import current_account_with_tenant, login_required
 from services.billing_service import BillingService
 
@@ -16,7 +17,13 @@ class Subscription(Resource):
         current_user, current_tenant_id = current_account_with_tenant()
         parser = (
             reqparse.RequestParser()
-            .add_argument("plan", type=str, required=True, location="args", choices=["professional", "team"])
+            .add_argument(
+                "plan",
+                type=str,
+                required=True,
+                location="args",
+                choices=[CloudPlan.PROFESSIONAL, CloudPlan.TEAM],
+            )
             .add_argument("interval", type=str, required=True, location="args", choices=["month", "year"])
         )
         args = parser.parse_args()

api/controllers/console/datasets/datasets_document.py

@@ -746,7 +746,7 @@ class DocumentApi(DocumentResource):
                 "name": document.name,
                 "created_from": document.created_from,
                 "created_by": document.created_by,
-                "created_at": document.created_at.timestamp(),
+                "created_at": int(document.created_at.timestamp()),
                 "tokens": document.tokens,
                 "indexing_status": document.indexing_status,
                 "completed_at": int(document.completed_at.timestamp()) if document.completed_at else None,
@@ -779,7 +779,7 @@ class DocumentApi(DocumentResource):
                 "name": document.name,
                 "created_from": document.created_from,
                 "created_by": document.created_by,
-                "created_at": document.created_at.timestamp(),
+                "created_at": int(document.created_at.timestamp()),
                 "tokens": document.tokens,
                 "indexing_status": document.indexing_status,
                 "completed_at": int(document.completed_at.timestamp()) if document.completed_at else None,

api/controllers/console/explore/banner.py

@@ -0,0 +1,43 @@
+from flask import request
+from flask_restx import Resource
+
+from controllers.console import api
+from controllers.console.explore.wraps import explore_banner_enabled
+from extensions.ext_database import db
+from models.model import ExporleBanner
+
+
+class BannerApi(Resource):
+    """Resource for banner list."""
+
+    @explore_banner_enabled
+    def get(self):
+        """Get banner list."""
+        language = request.args.get("language", "en-US")
+
+        # Build base query for enabled banners
+        base_query = db.session.query(ExporleBanner).where(ExporleBanner.status == "enabled")
+        
+        # Try to get banners in the requested language
+        banners = base_query.where(ExporleBanner.language == language).order_by(ExporleBanner.sort).all()
+        
+        # Fallback to en-US if no banners found and language is not en-US
+        if not banners and language != "en-US":
+            banners = base_query.where(ExporleBanner.language == "en-US").order_by(ExporleBanner.sort).all()
+        # Convert banners to serializable format
+        result = []
+        for banner in banners:
+            banner_data = {
+                "id": banner.id,
+                "content": banner.content,  # Already parsed as JSON by SQLAlchemy
+                "link": banner.link,
+                "sort": banner.sort,
+                "status": banner.status,
+                "created_at": banner.created_at.isoformat() if banner.created_at else None,
+            }
+            result.append(banner_data)
+
+        return result
+
+
+api.add_resource(BannerApi, "/explore/banners")

api/controllers/console/explore/error.py

@@ -29,3 +29,25 @@ class AppAccessDeniedError(BaseHTTPException):
     error_code = "access_denied"
     description = "App access denied."
     code = 403
+
+
+class TrialAppNotAllowed(BaseHTTPException):
+    """*403* `Trial App Not Allowed`
+
+    Raise if the user has reached the trial app limit.
+    """
+
+    error_code = "trial_app_not_allowed"
+    code = 403
+    description = "the app is not allowed to be trial."
+
+
+class TrialAppLimitExceeded(BaseHTTPException):
+    """*403* `Trial App Limit Exceeded`
+
+    Raise if the user has exceeded the trial app limit.
+    """
+
+    error_code = "trial_app_limit_exceeded"
+    code = 403
+    description = "The user has exceeded the trial app limit."

api/controllers/console/explore/recommended_app.py

@@ -27,6 +27,7 @@ recommended_app_fields = {
     "category": fields.String,
     "position": fields.Integer,
     "is_listed": fields.Boolean,
+    "can_trial": fields.Boolean,
 }
 
 recommended_app_list_fields = {

api/controllers/console/explore/trial.py

@@ -0,0 +1,514 @@
+import logging
+from typing import Any, cast
+
+from flask import request
+from flask_restx import Resource, marshal, marshal_with, reqparse
+from werkzeug.exceptions import Forbidden, InternalServerError, NotFound
+
+import services
+from controllers.common import fields
+from controllers.common.fields import build_site_model
+from controllers.console import api
+from controllers.console.app.error import (
+    AppUnavailableError,
+    AudioTooLargeError,
+    CompletionRequestError,
+    ConversationCompletedError,
+    NeedAddIdsError,
+    NoAudioUploadedError,
+    ProviderModelCurrentlyNotSupportError,
+    ProviderNotInitializeError,
+    ProviderNotSupportSpeechToTextError,
+    ProviderQuotaExceededError,
+    UnsupportedAudioTypeError,
+)
+from controllers.console.app.wraps import get_app_model_with_trial
+from controllers.console.explore.error import (
+    AppSuggestedQuestionsAfterAnswerDisabledError,
+    NotChatAppError,
+    NotCompletionAppError,
+    NotWorkflowAppError,
+)
+from controllers.console.explore.wraps import TrialAppResource, trial_feature_enable
+from controllers.service_api import service_api_ns
+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
+from core.app.entities.app_invoke_entities import InvokeFrom
+from core.errors.error import (
+    ModelCurrentlyNotSupportError,
+    ProviderTokenNotInitError,
+    QuotaExceededError,
+)
+from core.model_runtime.errors.invoke import InvokeError
+from core.workflow.graph_engine.manager import GraphEngineManager
+from extensions.ext_database import db
+from fields.app_fields import app_detail_fields_with_site
+from fields.dataset_fields import dataset_fields
+from fields.workflow_fields import workflow_fields
+from libs import helper
+from libs.helper import uuid_value
+from libs.login import current_user
+from models import Account
+from models.account import TenantStatus
+from models.model import AppMode, Site
+from models.workflow import Workflow
+from services.app_generate_service import AppGenerateService
+from services.app_service import AppService
+from services.audio_service import AudioService
+from services.dataset_service import DatasetService
+from services.errors.audio import (
+    AudioTooLargeServiceError,
+    NoAudioUploadedServiceError,
+    ProviderNotSupportSpeechToTextServiceError,
+    UnsupportedAudioTypeServiceError,
+)
+from services.errors.conversation import ConversationNotExistsError
+from services.errors.llm import InvokeRateLimitError
+from services.errors.message import (
+    MessageNotExistsError,
+    SuggestedQuestionsAfterAnswerDisabledError,
+)
+from services.message_service import MessageService
+from services.recommended_app_service import RecommendedAppService
+
+logger = logging.getLogger(__name__)
+
+
+class TrialAppWorkflowRunApi(TrialAppResource):
+    def post(self, trial_app):
+        """
+        Run workflow
+        """
+        app_model = trial_app
+        if not app_model:
+            raise NotWorkflowAppError()
+        app_mode = AppMode.value_of(app_model.mode)
+        if app_mode != AppMode.WORKFLOW:
+            raise NotWorkflowAppError()
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("inputs", type=dict, required=True, nullable=False, location="json")
+        parser.add_argument("files", type=list, required=False, location="json")
+        args = parser.parse_args()
+        assert current_user is not None
+        try:
+            app_id = app_model.id
+            user_id = current_user.id
+            response = AppGenerateService.generate(
+                app_model=app_model, user=current_user, args=args, invoke_from=InvokeFrom.EXPLORE, streaming=True
+            )
+            RecommendedAppService.add_trial_app_record(app_id, user_id)
+            return helper.compact_generate_response(response)
+        except ProviderTokenNotInitError as ex:
+            raise ProviderNotInitializeError(ex.description)
+        except QuotaExceededError:
+            raise ProviderQuotaExceededError()
+        except ModelCurrentlyNotSupportError:
+            raise ProviderModelCurrentlyNotSupportError()
+        except InvokeError as e:
+            raise CompletionRequestError(e.description)
+        except InvokeRateLimitError as ex:
+            raise InvokeRateLimitHttpError(ex.description)
+        except ValueError as e:
+            raise e
+        except Exception:
+            logger.exception("internal server error.")
+            raise InternalServerError()
+
+
+class TrialAppWorkflowTaskStopApi(TrialAppResource):
+    def post(self, trial_app, task_id: str):
+        """
+        Stop workflow task
+        """
+        app_model = trial_app
+        if not app_model:
+            raise NotWorkflowAppError()
+        app_mode = AppMode.value_of(app_model.mode)
+        if app_mode != AppMode.WORKFLOW:
+            raise NotWorkflowAppError()
+        assert current_user is not None
+
+        # Stop using both mechanisms for backward compatibility
+        # Legacy stop flag mechanism (without user check)
+        AppQueueManager.set_stop_flag_no_user_check(task_id)
+
+        # New graph engine command channel mechanism
+        GraphEngineManager.send_stop_command(task_id)
+
+        return {"result": "success"}
+
+
+class TrialChatApi(TrialAppResource):
+    @trial_feature_enable
+    def post(self, trial_app):
+        app_model = trial_app
+        app_mode = AppMode.value_of(app_model.mode)
+        if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
+            raise NotChatAppError()
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("inputs", type=dict, required=True, location="json")
+        parser.add_argument("query", type=str, required=True, location="json")
+        parser.add_argument("files", type=list, required=False, location="json")
+        parser.add_argument("conversation_id", type=uuid_value, location="json")
+        parser.add_argument("parent_message_id", type=uuid_value, required=False, location="json")
+        parser.add_argument("retriever_from", type=str, required=False, default="explore_app", location="json")
+        args = parser.parse_args()
+
+        args["auto_generate_name"] = False
+
+        try:
+            if not isinstance(current_user, Account):
+                raise ValueError("current_user must be an Account instance")
+            
+            # Get IDs before they might be detached from session
+            app_id = app_model.id
+            user_id = current_user.id
+            
+            response = AppGenerateService.generate(
+                app_model=app_model, user=current_user, args=args, invoke_from=InvokeFrom.EXPLORE, streaming=True
+            )
+            RecommendedAppService.add_trial_app_record(app_id, user_id)
+            return helper.compact_generate_response(response)
+        except services.errors.conversation.ConversationNotExistsError:
+            raise NotFound("Conversation Not Exists.")
+        except services.errors.conversation.ConversationCompletedError:
+            raise ConversationCompletedError()
+        except services.errors.app_model_config.AppModelConfigBrokenError:
+            logger.exception("App model config broken.")
+            raise AppUnavailableError()
+        except ProviderTokenNotInitError as ex:
+            raise ProviderNotInitializeError(ex.description)
+        except QuotaExceededError:
+            raise ProviderQuotaExceededError()
+        except ModelCurrentlyNotSupportError:
+            raise ProviderModelCurrentlyNotSupportError()
+        except InvokeError as e:
+            raise CompletionRequestError(e.description)
+        except InvokeRateLimitError as ex:
+            raise InvokeRateLimitHttpError(ex.description)
+        except ValueError as e:
+            raise e
+        except Exception:
+            logger.exception("internal server error.")
+            raise InternalServerError()
+
+
+class TrialMessageSuggestedQuestionApi(TrialAppResource):
+    @trial_feature_enable
+    def get(self, trial_app, message_id):
+        app_model = trial_app
+        app_mode = AppMode.value_of(app_model.mode)
+        if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
+            raise NotChatAppError()
+
+        message_id = str(message_id)
+
+        try:
+            if not isinstance(current_user, Account):
+                raise ValueError("current_user must be an Account instance")
+            questions = MessageService.get_suggested_questions_after_answer(
+                app_model=app_model, user=current_user, message_id=message_id, invoke_from=InvokeFrom.EXPLORE
+            )
+        except MessageNotExistsError:
+            raise NotFound("Message not found")
+        except ConversationNotExistsError:
+            raise NotFound("Conversation not found")
+        except SuggestedQuestionsAfterAnswerDisabledError:
+            raise AppSuggestedQuestionsAfterAnswerDisabledError()
+        except ProviderTokenNotInitError as ex:
+            raise ProviderNotInitializeError(ex.description)
+        except QuotaExceededError:
+            raise ProviderQuotaExceededError()
+        except ModelCurrentlyNotSupportError:
+            raise ProviderModelCurrentlyNotSupportError()
+        except InvokeError as e:
+            raise CompletionRequestError(e.description)
+        except Exception:
+            logger.exception("internal server error.")
+            raise InternalServerError()
+
+        return {"data": questions}
+
+
+class TrialChatAudioApi(TrialAppResource):
+    @trial_feature_enable
+    def post(self, trial_app):
+        app_model = trial_app
+
+        file = request.files["file"]
+
+        try:
+            if not isinstance(current_user, Account):
+                raise ValueError("current_user must be an Account instance")
+            
+            # Get IDs before they might be detached from session
+            app_id = app_model.id
+            user_id = current_user.id
+            
+            response = AudioService.transcript_asr(app_model=app_model, file=file, end_user=None)
+            RecommendedAppService.add_trial_app_record(app_id, user_id)
+            return response
+        except services.errors.app_model_config.AppModelConfigBrokenError:
+            logger.exception("App model config broken.")
+            raise AppUnavailableError()
+        except NoAudioUploadedServiceError:
+            raise NoAudioUploadedError()
+        except AudioTooLargeServiceError as e:
+            raise AudioTooLargeError(str(e))
+        except UnsupportedAudioTypeServiceError:
+            raise UnsupportedAudioTypeError()
+        except ProviderNotSupportSpeechToTextServiceError:
+            raise ProviderNotSupportSpeechToTextError()
+        except ProviderTokenNotInitError as ex:
+            raise ProviderNotInitializeError(ex.description)
+        except QuotaExceededError:
+            raise ProviderQuotaExceededError()
+        except ModelCurrentlyNotSupportError:
+            raise ProviderModelCurrentlyNotSupportError()
+        except InvokeError as e:
+            raise CompletionRequestError(e.description)
+        except ValueError as e:
+            raise e
+        except Exception as e:
+            logger.exception("internal server error.")
+            raise InternalServerError()
+
+
+class TrialChatTextApi(TrialAppResource):
+    @trial_feature_enable
+    def post(self, trial_app):
+        app_model = trial_app
+        try:
+            parser = reqparse.RequestParser()
+            parser.add_argument("message_id", type=str, required=False, location="json")
+            parser.add_argument("voice", type=str, location="json")
+            parser.add_argument("text", type=str, location="json")
+            parser.add_argument("streaming", type=bool, location="json")
+            args = parser.parse_args()
+
+            message_id = args.get("message_id", None)
+            text = args.get("text", None)
+            voice = args.get("voice", None)
+            if not isinstance(current_user, Account):
+                raise ValueError("current_user must be an Account instance")
+            
+            # Get IDs before they might be detached from session
+            app_id = app_model.id
+            user_id = current_user.id
+            
+            response = AudioService.transcript_tts(app_model=app_model, text=text, voice=voice, message_id=message_id)
+            RecommendedAppService.add_trial_app_record(app_id, user_id)
+            return response
+        except services.errors.app_model_config.AppModelConfigBrokenError:
+            logger.exception("App model config broken.")
+            raise AppUnavailableError()
+        except NoAudioUploadedServiceError:
+            raise NoAudioUploadedError()
+        except AudioTooLargeServiceError as e:
+            raise AudioTooLargeError(str(e))
+        except UnsupportedAudioTypeServiceError:
+            raise UnsupportedAudioTypeError()
+        except ProviderNotSupportSpeechToTextServiceError:
+            raise ProviderNotSupportSpeechToTextError()
+        except ProviderTokenNotInitError as ex:
+            raise ProviderNotInitializeError(ex.description)
+        except QuotaExceededError:
+            raise ProviderQuotaExceededError()
+        except ModelCurrentlyNotSupportError:
+            raise ProviderModelCurrentlyNotSupportError()
+        except InvokeError as e:
+            raise CompletionRequestError(e.description)
+        except ValueError as e:
+            raise e
+        except Exception as e:
+            logger.exception("internal server error.")
+            raise InternalServerError()
+
+
+class TrialCompletionApi(TrialAppResource):
+    @trial_feature_enable
+    def post(self, trial_app):
+        app_model = trial_app
+        if app_model.mode != "completion":
+            raise NotCompletionAppError()
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("inputs", type=dict, required=True, location="json")
+        parser.add_argument("query", type=str, location="json", default="")
+        parser.add_argument("files", type=list, required=False, location="json")
+        parser.add_argument("response_mode", type=str, choices=["blocking", "streaming"], location="json")
+        parser.add_argument("retriever_from", type=str, required=False, default="explore_app", location="json")
+        args = parser.parse_args()
+
+        streaming = args["response_mode"] == "streaming"
+        args["auto_generate_name"] = False
+
+        try:
+            if not isinstance(current_user, Account):
+                raise ValueError("current_user must be an Account instance")
+            
+            # Get IDs before they might be detached from session
+            app_id = app_model.id
+            user_id = current_user.id
+            
+            response = AppGenerateService.generate(
+                app_model=app_model, user=current_user, args=args, invoke_from=InvokeFrom.EXPLORE, streaming=streaming
+            )
+
+            RecommendedAppService.add_trial_app_record(app_id, user_id)
+            return helper.compact_generate_response(response)
+        except services.errors.conversation.ConversationNotExistsError:
+            raise NotFound("Conversation Not Exists.")
+        except services.errors.conversation.ConversationCompletedError:
+            raise ConversationCompletedError()
+        except services.errors.app_model_config.AppModelConfigBrokenError:
+            logger.exception("App model config broken.")
+            raise AppUnavailableError()
+        except ProviderTokenNotInitError as ex:
+            raise ProviderNotInitializeError(ex.description)
+        except QuotaExceededError:
+            raise ProviderQuotaExceededError()
+        except ModelCurrentlyNotSupportError:
+            raise ProviderModelCurrentlyNotSupportError()
+        except InvokeError as e:
+            raise CompletionRequestError(e.description)
+        except ValueError as e:
+            raise e
+        except Exception:
+            logger.exception("internal server error.")
+            raise InternalServerError()
+
+
+class TrialSitApi(Resource):
+    """Resource for trial app sites."""
+
+    @trial_feature_enable
+    @get_app_model_with_trial
+    @service_api_ns.marshal_with(build_site_model(service_api_ns))
+    def get(self, app_model):
+        """Retrieve app site info.
+
+        Returns the site configuration for the application including theme, icons, and text.
+        """
+        site = db.session.query(Site).where(Site.app_id == app_model.id).first()
+
+        if not site:
+            raise Forbidden()
+
+        assert app_model.tenant
+        if app_model.tenant.status == TenantStatus.ARCHIVE:
+            raise Forbidden()
+
+        return site
+
+
+class TrialAppParameterApi(Resource):
+    """Resource for app variables."""
+
+    @trial_feature_enable
+    @get_app_model_with_trial
+    @marshal_with(fields.parameters_fields)
+    def get(self, app_model):
+        """Retrieve app parameters."""
+
+        if app_model is None:
+            raise AppUnavailableError()
+
+        if app_model.mode in {AppMode.ADVANCED_CHAT, AppMode.WORKFLOW}:
+            workflow = app_model.workflow
+            if workflow is None:
+                raise AppUnavailableError()
+
+            features_dict = workflow.features_dict
+            user_input_form = workflow.user_input_form(to_old_structure=True)
+        else:
+            app_model_config = app_model.app_model_config
+            if app_model_config is None:
+                raise AppUnavailableError()
+
+            features_dict = app_model_config.to_dict()
+
+            user_input_form = features_dict.get("user_input_form", [])
+
+        return get_parameters_from_feature_dict(features_dict=features_dict, user_input_form=user_input_form)
+
+
+class AppApi(Resource):
+    @trial_feature_enable
+    @get_app_model_with_trial
+    @marshal_with(app_detail_fields_with_site)
+    def get(self, app_model):
+        """Get app detail"""
+
+        app_service = AppService()
+        app_model = app_service.get_app(app_model)
+
+        return app_model
+
+
+class AppWorkflowApi(Resource):
+    @trial_feature_enable
+    @get_app_model_with_trial
+    @marshal_with(workflow_fields)
+    def get(self, app_model):
+        """Get workflow detail"""
+        if not app_model.workflow_id:
+            raise AppUnavailableError()
+
+        workflow = (
+            db.session.query(Workflow)
+            .where(
+                Workflow.id == app_model.workflow_id,
+            )
+            .first()
+        )
+        return workflow
+
+
+class DatasetListApi(Resource):
+    @trial_feature_enable
+    @get_app_model_with_trial
+    def get(self, app_model):
+        page = request.args.get("page", default=1, type=int)
+        limit = request.args.get("limit", default=20, type=int)
+        ids = request.args.getlist("ids")
+      
+        tenant_id = app_model.tenant_id
+        if ids:
+            datasets, total = DatasetService.get_datasets_by_ids(ids, tenant_id)
+        else:
+            raise NeedAddIdsError()
+
+        data = cast(list[dict[str, Any]], marshal(datasets, dataset_fields))
+
+        response = {"data": data, "has_more": len(datasets) == limit, "limit": limit, "total": total, "page": page}
+        return response
+
+
+api.add_resource(TrialChatApi, "/trial-apps/<uuid:app_id>/chat-messages", endpoint="trial_app_chat_completion")
+
+api.add_resource(
+    TrialMessageSuggestedQuestionApi,
+    "/trial-apps/<uuid:app_id>/messages/<uuid:message_id>/suggested-questions",
+    endpoint="trial_app_suggested_question",
+)
+
+api.add_resource(TrialChatAudioApi, "/trial-apps/<uuid:app_id>/audio-to-text", endpoint="trial_app_audio")
+api.add_resource(TrialChatTextApi, "/trial-apps/<uuid:app_id>/text-to-audio", endpoint="trial_app_text")
+
+api.add_resource(TrialCompletionApi, "/trial-apps/<uuid:app_id>/completion-messages", endpoint="trial_app_completion")
+
+api.add_resource(TrialSitApi, "/trial-apps/<uuid:app_id>/site")
+
+api.add_resource(TrialAppParameterApi, "/trial-apps/<uuid:app_id>/parameters", endpoint="trial_app_parameters")
+
+api.add_resource(AppApi, "/trial-apps/<uuid:app_id>", endpoint="trial_app")
+
+api.add_resource(TrialAppWorkflowRunApi, "/trial-apps/<uuid:app_id>/workflows/run", endpoint="trial_app_workflow_run")
+api.add_resource(TrialAppWorkflowTaskStopApi, "/trial-apps/<uuid:app_id>/workflows/tasks/<string:task_id>/stop")
+
+api.add_resource(AppWorkflowApi, "/trial-apps/<uuid:app_id>/workflows", endpoint="trial_app_workflow")
+api.add_resource(DatasetListApi, "/trial-apps/<uuid:app_id>/datasets", endpoint="trial_app_datasets")

api/controllers/console/explore/wraps.py

@@ -2,14 +2,16 @@ from collections.abc import Callable
 from functools import wraps
 from typing import Concatenate, ParamSpec, TypeVar
 
+from flask import abort
+from flask_login import current_user
 from flask_restx import Resource
 from werkzeug.exceptions import NotFound
 
-from controllers.console.explore.error import AppAccessDeniedError
+from controllers.console.explore.error import AppAccessDeniedError, TrialAppLimitExceeded, TrialAppNotAllowed
 from controllers.console.wraps import account_initialization_required
 from extensions.ext_database import db
 from libs.login import current_account_with_tenant, login_required
-from models import InstalledApp
+from models import AccountTrialAppRecord, App, InstalledApp, TrialApp
 from services.enterprise.enterprise_service import EnterpriseService
 from services.feature_service import FeatureService
 
@@ -71,6 +73,59 @@ def user_allowed_to_access_app(view: Callable[Concatenate[InstalledApp, P], R] |
     return decorator
 
 
+def trial_app_required(view: Callable[Concatenate[App, P], R] | None = None):
+    def decorator(view: Callable[Concatenate[App, P], R]):
+        @wraps(view)
+        def decorated(app_id: str, *args: P.args, **kwargs: P.kwargs):
+            trial_app = db.session.query(TrialApp).where(TrialApp.app_id == str(app_id)).first()
+
+            if trial_app is None:
+                raise TrialAppNotAllowed()
+            app = trial_app.app
+
+            if app is None:
+                raise TrialAppNotAllowed()
+
+            account_trial_app_record = (
+                db.session.query(AccountTrialAppRecord)
+                .where(AccountTrialAppRecord.account_id == current_user.id, AccountTrialAppRecord.app_id == app_id)
+                .first()
+            )
+            if account_trial_app_record:
+                if account_trial_app_record.count >= trial_app.trial_limit:
+                    raise TrialAppLimitExceeded()
+
+            return view(app, *args, **kwargs)
+
+        return decorated
+
+    if view:
+        return decorator(view)
+    return decorator
+
+
+def trial_feature_enable(view: Callable[..., R]) -> Callable[..., R]:
+    @wraps(view)
+    def decorated(*args, **kwargs):
+        features = FeatureService.get_system_features()
+        if not features.enable_trial_app:
+            abort(403, "Trial app feature is not enabled.")
+        return view(*args, **kwargs)
+
+    return decorated
+
+
+def explore_banner_enabled(view: Callable[..., R]) -> Callable[..., R]:
+    @wraps(view)
+    def decorated(*args, **kwargs):
+        features = FeatureService.get_system_features()
+        if not features.enable_explore_banner:
+            abort(403, "Explore banner feature is not enabled.")
+        return view(*args, **kwargs)
+
+    return decorated
+
+
 class InstalledAppResource(Resource):
     # must be reversed if there are multiple decorators
 
@@ -80,3 +135,13 @@ class InstalledAppResource(Resource):
         account_initialization_required,
         login_required,
     ]
+
+
+class TrialAppResource(Resource):
+    # must be reversed if there are multiple decorators
+
+    method_decorators = [
+        trial_app_required,
+        account_initialization_required,
+        login_required,
+    ]

api/controllers/console/files.py

@@ -8,6 +8,7 @@ import services
 from configs import dify_config
 from constants import DOCUMENT_EXTENSIONS
 from controllers.common.errors import (
+    BlockedFileExtensionError,
     FilenameNotExistsError,
     FileTooLargeError,
     NoFileUploadedError,
@@ -83,6 +84,8 @@ class FileApi(Resource):
             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)
 
         return upload_file, 201
 

api/controllers/console/workspace/account.py

@@ -32,6 +32,7 @@ from controllers.console.wraps import (
     only_edition_cloud,
     setup_required,
 )
+from core.file import helpers as file_helpers
 from extensions.ext_database import db
 from fields.member_fields import account_fields
 from libs.datetime_utils import naive_utc_now
@@ -128,6 +129,17 @@ class AccountNameApi(Resource):
 
 @console_ns.route("/account/avatar")
 class AccountAvatarApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument("avatar", type=str, required=True, location="args")
+        args = parser.parse_args()
+
+        avatar_url = file_helpers.get_signed_file_url(args["avatar"])
+        return {"avatar_url": avatar_url}
+
     @setup_required
     @login_required
     @account_initialization_required

api/controllers/console/workspace/workspace.py

@@ -21,6 +21,7 @@ from controllers.console.wraps import (
     cloud_edition_billing_resource_check,
     setup_required,
 )
+from enums.cloud_plan import CloudPlan
 from extensions.ext_database import db
 from libs.helper import TimestampField
 from libs.login import current_account_with_tenant, login_required
@@ -50,6 +51,9 @@ tenant_fields = {
     "in_trial": fields.Boolean,
     "trial_end_reason": fields.String,
     "custom_config": fields.Raw(attribute="custom_config"),
+    "trial_credits": fields.Integer,
+    "trial_credits_used": fields.Integer,
+    "next_credit_reset_date": fields.Integer,
 }
 
 tenants_fields = {
@@ -83,7 +87,7 @@ class TenantListApi(Resource):
                 "name": tenant.name,
                 "status": tenant.status,
                 "created_at": tenant.created_at,
-                "plan": features.billing.subscription.plan if features.billing.enabled else "sandbox",
+                "plan": features.billing.subscription.plan if features.billing.enabled else CloudPlan.SANDBOX,
                 "current": tenant.id == current_tenant_id if current_tenant_id else False,
             }
 

api/controllers/console/wraps.py

@@ -10,6 +10,7 @@ from flask import abort, request
 
 from configs import dify_config
 from controllers.console.workspace.error import AccountNotInitializedError
+from enums.cloud_plan import CloudPlan
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from libs.login import current_account_with_tenant
@@ -133,7 +134,7 @@ def cloud_edition_billing_knowledge_limit_check(resource: str):
             features = FeatureService.get_features(current_tenant_id)
             if features.billing.enabled:
                 if resource == "add_segment":
-                    if features.billing.subscription.plan == "sandbox":
+                    if features.billing.subscription.plan == CloudPlan.SANDBOX:
                         abort(
                             403,
                             "To unlock this feature and elevate your Dify experience, please upgrade to a paid plan.",

api/controllers/service_api/__init__.py

@@ -19,6 +19,7 @@ from .app import (
     annotation,
     app,
     audio,
+    chatflow_memory,
     completion,
     conversation,
     file,
@@ -40,6 +41,7 @@ __all__ = [
     "annotation",
     "app",
     "audio",
+    "chatflow_memory",
     "completion",
     "conversation",
     "dataset",

api/controllers/service_api/app/chatflow_memory.py

@@ -0,0 +1,109 @@
+from flask_restx import Resource, reqparse
+
+from controllers.service_api import api
+from controllers.service_api.wraps import FetchUserArg, WhereisUserArg, validate_app_token
+from core.memory.entities import MemoryBlock, MemoryCreatedBy
+from core.workflow.runtime.variable_pool import VariablePool
+from models import App, EndUser
+from services.chatflow_memory_service import ChatflowMemoryService
+from services.workflow_service import WorkflowService
+
+
+class MemoryListApi(Resource):
+    @validate_app_token(fetch_user_arg=FetchUserArg(fetch_from=WhereisUserArg.JSON, required=True))
+    def get(self, app_model: App, end_user: EndUser):
+        parser = reqparse.RequestParser()
+        parser.add_argument("conversation_id", required=False, type=str | None, default=None)
+        parser.add_argument("memory_id", required=False, type=str | None, default=None)
+        parser.add_argument("version", required=False, type=int | None, default=None)
+        args = parser.parse_args()
+        conversation_id: str | None = args.get("conversation_id")
+        memory_id = args.get("memory_id")
+        version = args.get("version")
+
+        if conversation_id:
+            result = ChatflowMemoryService.get_persistent_memories_with_conversation(
+                app_model, MemoryCreatedBy(end_user_id=end_user.id), conversation_id, version
+            )
+            session_memories = ChatflowMemoryService.get_session_memories_with_conversation(
+                app_model, MemoryCreatedBy(end_user_id=end_user.id), conversation_id, version
+            )
+            result = [*result, *session_memories]
+        else:
+            result = ChatflowMemoryService.get_persistent_memories(
+                app_model, MemoryCreatedBy(end_user_id=end_user.id), version
+            )
+
+        if memory_id:
+            result = [it for it in result if it.spec.id == memory_id]
+        return [it for it in result if it.spec.end_user_visible]
+
+
+class MemoryEditApi(Resource):
+    @validate_app_token(fetch_user_arg=FetchUserArg(fetch_from=WhereisUserArg.JSON, required=True))
+    def put(self, app_model: App, end_user: EndUser):
+        parser = reqparse.RequestParser()
+        parser.add_argument("id", type=str, required=True)
+        parser.add_argument("conversation_id", type=str | None, required=False, default=None)
+        parser.add_argument("node_id", type=str | None, required=False, default=None)
+        parser.add_argument("update", type=str, required=True)
+        args = parser.parse_args()
+        workflow = WorkflowService().get_published_workflow(app_model)
+        update = args.get("update")
+        conversation_id = args.get("conversation_id")
+        node_id = args.get("node_id")
+        if not isinstance(update, str):
+            return {"error": "Invalid update"}, 400
+        if not workflow:
+            return {"error": "Workflow not found"}, 404
+        memory_spec = next((it for it in workflow.memory_blocks if it.id == args["id"]), None)
+        if not memory_spec:
+            return {"error": "Memory not found"}, 404
+
+        # First get existing memory
+        existing_memory = ChatflowMemoryService.get_memory_by_spec(
+            spec=memory_spec,
+            tenant_id=app_model.tenant_id,
+            app_id=app_model.id,
+            created_by=MemoryCreatedBy(end_user_id=end_user.id),
+            conversation_id=conversation_id,
+            node_id=node_id,
+            is_draft=False,
+        )
+
+        # Create updated memory instance with incremented version
+        updated_memory = MemoryBlock(
+            spec=existing_memory.spec,
+            tenant_id=existing_memory.tenant_id,
+            app_id=existing_memory.app_id,
+            conversation_id=existing_memory.conversation_id,
+            node_id=existing_memory.node_id,
+            value=update,  # New value
+            version=existing_memory.version + 1,  # Increment version for update
+            edited_by_user=True,
+            created_by=existing_memory.created_by,
+        )
+
+        ChatflowMemoryService.save_memory(updated_memory, VariablePool(), False)
+        return "", 204
+
+
+class MemoryDeleteApi(Resource):
+    @validate_app_token(fetch_user_arg=FetchUserArg(fetch_from=WhereisUserArg.JSON, required=True))
+    def delete(self, app_model: App, end_user: EndUser):
+        parser = reqparse.RequestParser()
+        parser.add_argument("id", type=str, required=False, default=None)
+        args = parser.parse_args()
+        memory_id = args.get("id")
+
+        if memory_id:
+            ChatflowMemoryService.delete_memory(app_model, memory_id, MemoryCreatedBy(end_user_id=end_user.id))
+            return "", 204
+        else:
+            ChatflowMemoryService.delete_all_user_memories(app_model, MemoryCreatedBy(end_user_id=end_user.id))
+            return "", 200
+
+
+api.add_resource(MemoryListApi, "/memories")
+api.add_resource(MemoryEditApi, "/memory-edit")
+api.add_resource(MemoryDeleteApi, "/memories")

api/controllers/service_api/dataset/document.py

@@ -592,7 +592,7 @@ class DocumentApi(DatasetApiResource):
                 "name": document.name,
                 "created_from": document.created_from,
                 "created_by": document.created_by,
-                "created_at": document.created_at.timestamp(),
+                "created_at": int(document.created_at.timestamp()),
                 "tokens": document.tokens,
                 "indexing_status": document.indexing_status,
                 "completed_at": int(document.completed_at.timestamp()) if document.completed_at else None,
@@ -625,7 +625,7 @@ class DocumentApi(DatasetApiResource):
                 "name": document.name,
                 "created_from": document.created_from,
                 "created_by": document.created_by,
-                "created_at": document.created_at.timestamp(),
+                "created_at": int(document.created_at.timestamp()),
                 "tokens": document.tokens,
                 "indexing_status": document.indexing_status,
                 "completed_at": int(document.completed_at.timestamp()) if document.completed_at else None,

api/controllers/service_api/dataset/segment.py

@@ -2,6 +2,7 @@ from flask import request
 from flask_restx import marshal, reqparse
 from werkzeug.exceptions import NotFound
 
+from configs import dify_config
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import ProviderNotInitializeError
 from controllers.service_api.wraps import (
@@ -107,6 +108,10 @@ class SegmentApi(DatasetApiResource):
         # validate args
         args = segment_create_parser.parse_args()
         if args["segments"] is not None:
+            segments_limit = dify_config.DATASET_MAX_SEGMENTS_PER_REQUEST
+            if segments_limit > 0 and len(args["segments"]) > segments_limit:
+                raise ValueError(f"Exceeded maximum segments limit of {segments_limit}.")
+
             for args_item in args["segments"]:
                 SegmentService.segment_create_args_validate(args_item, document)
             segments = SegmentService.multi_create_segment(args["segments"], document, dataset)

api/controllers/service_api/wraps.py

@@ -13,6 +13,7 @@ from sqlalchemy import select, update
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import Forbidden, NotFound, Unauthorized
 
+from enums.cloud_plan import CloudPlan
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from libs.datetime_utils import naive_utc_now
@@ -66,6 +67,7 @@ def validate_app_token(view: Callable[P, R] | None = None, *, fetch_user_arg: Fe
 
             kwargs["app_model"] = app_model
 
+            # If caller needs end-user context, attach EndUser to current_user
             if fetch_user_arg:
                 if fetch_user_arg.fetch_from == WhereisUserArg.QUERY:
                     user_id = request.args.get("user")
@@ -74,7 +76,6 @@ def validate_app_token(view: Callable[P, R] | None = None, *, fetch_user_arg: Fe
                 elif fetch_user_arg.fetch_from == WhereisUserArg.FORM:
                     user_id = request.form.get("user")
                 else:
-                    # use default-user
                     user_id = None
 
                 if not user_id and fetch_user_arg.required:
@@ -89,6 +90,28 @@ def validate_app_token(view: Callable[P, R] | None = None, *, fetch_user_arg: Fe
                 # Set EndUser as current logged-in user for flask_login.current_user
                 current_app.login_manager._update_request_context_with_user(end_user)  # type: ignore
                 user_logged_in.send(current_app._get_current_object(), user=end_user)  # type: ignore
+            else:
+                # For service API without end-user context, ensure an Account is logged in
+                # so services relying on current_account_with_tenant() work correctly.
+                tenant_owner_info = (
+                    db.session.query(Tenant, Account)
+                    .join(TenantAccountJoin, Tenant.id == TenantAccountJoin.tenant_id)
+                    .join(Account, TenantAccountJoin.account_id == Account.id)
+                    .where(
+                        Tenant.id == app_model.tenant_id,
+                        TenantAccountJoin.role == "owner",
+                        Tenant.status == TenantStatus.NORMAL,
+                    )
+                    .one_or_none()
+                )
+
+                if tenant_owner_info:
+                    tenant_model, account = tenant_owner_info
+                    account.current_tenant = tenant_model
+                    current_app.login_manager._update_request_context_with_user(account)  # type: ignore
+                    user_logged_in.send(current_app._get_current_object(), user=current_user)  # type: ignore
+                else:
+                    raise Unauthorized("Tenant owner account not found or tenant is not active.")
 
             return view_func(*args, **kwargs)
 
@@ -138,7 +161,7 @@ def cloud_edition_billing_knowledge_limit_check(resource: str, api_token_type: s
             features = FeatureService.get_features(api_token.tenant_id)
             if features.billing.enabled:
                 if resource == "add_segment":
-                    if features.billing.subscription.plan == "sandbox":
+                    if features.billing.subscription.plan == CloudPlan.SANDBOX:
                         raise Forbidden(
                             "To unlock this feature and elevate your Dify experience, please upgrade to a paid plan."
                         )

api/controllers/web/__init__.py

@@ -18,6 +18,7 @@ web_ns = Namespace("web", description="Web application API operations", path="/"
 from . import (
     app,
     audio,
+    chatflow_memory,
     completion,
     conversation,
     feature,
@@ -39,6 +40,7 @@ __all__ = [
     "app",
     "audio",
     "bp",
+    "chatflow_memory",
     "completion",
     "conversation",
     "feature",

api/controllers/web/chatflow_memory.py

@@ -0,0 +1,108 @@
+from flask_restx import reqparse
+
+from controllers.web import api
+from controllers.web.wraps import WebApiResource
+from core.memory.entities import MemoryBlock, MemoryCreatedBy
+from core.workflow.runtime.variable_pool import VariablePool
+from models import App, EndUser
+from services.chatflow_memory_service import ChatflowMemoryService
+from services.workflow_service import WorkflowService
+
+
+class MemoryListApi(WebApiResource):
+    def get(self, app_model: App, end_user: EndUser):
+        parser = reqparse.RequestParser()
+        parser.add_argument("conversation_id", required=False, type=str | None, default=None)
+        parser.add_argument("memory_id", required=False, type=str | None, default=None)
+        parser.add_argument("version", required=False, type=int | None, default=None)
+        args = parser.parse_args()
+        conversation_id: str | None = args.get("conversation_id")
+        memory_id = args.get("memory_id")
+        version = args.get("version")
+
+        if conversation_id:
+            result = ChatflowMemoryService.get_persistent_memories_with_conversation(
+                app_model, MemoryCreatedBy(end_user_id=end_user.id), conversation_id, version
+            )
+            session_memories = ChatflowMemoryService.get_session_memories_with_conversation(
+                app_model, MemoryCreatedBy(end_user_id=end_user.id), conversation_id, version
+            )
+            result = [*result, *session_memories]
+        else:
+            result = ChatflowMemoryService.get_persistent_memories(
+                app_model, MemoryCreatedBy(end_user_id=end_user.id), version
+            )
+
+        if memory_id:
+            result = [it for it in result if it.spec.id == memory_id]
+        return [it for it in result if it.spec.end_user_visible]
+
+
+class MemoryEditApi(WebApiResource):
+    def put(self, app_model: App, end_user: EndUser):
+        parser = reqparse.RequestParser()
+        parser.add_argument("id", type=str, required=True)
+        parser.add_argument("conversation_id", type=str | None, required=False, default=None)
+        parser.add_argument("node_id", type=str | None, required=False, default=None)
+        parser.add_argument("update", type=str, required=True)
+        args = parser.parse_args()
+        workflow = WorkflowService().get_published_workflow(app_model)
+        update = args.get("update")
+        conversation_id = args.get("conversation_id")
+        node_id = args.get("node_id")
+        if not isinstance(update, str):
+            return {"error": "Update must be a string"}, 400
+        if not workflow:
+            return {"error": "Workflow not found"}, 404
+        memory_spec = next((it for it in workflow.memory_blocks if it.id == args["id"]), None)
+        if not memory_spec:
+            return {"error": "Memory not found"}, 404
+        if not memory_spec.end_user_editable:
+            return {"error": "Memory not editable"}, 403
+
+        # First get existing memory
+        existing_memory = ChatflowMemoryService.get_memory_by_spec(
+            spec=memory_spec,
+            tenant_id=app_model.tenant_id,
+            app_id=app_model.id,
+            created_by=MemoryCreatedBy(end_user_id=end_user.id),
+            conversation_id=conversation_id,
+            node_id=node_id,
+            is_draft=False,
+        )
+
+        # Create updated memory instance with incremented version
+        updated_memory = MemoryBlock(
+            spec=existing_memory.spec,
+            tenant_id=existing_memory.tenant_id,
+            app_id=existing_memory.app_id,
+            conversation_id=existing_memory.conversation_id,
+            node_id=existing_memory.node_id,
+            value=update,  # New value
+            version=existing_memory.version + 1,  # Increment version for update
+            edited_by_user=True,
+            created_by=existing_memory.created_by,
+        )
+
+        ChatflowMemoryService.save_memory(updated_memory, VariablePool(), False)
+        return "", 204
+
+
+class MemoryDeleteApi(WebApiResource):
+    def delete(self, app_model: App, end_user: EndUser):
+        parser = reqparse.RequestParser()
+        parser.add_argument("id", type=str, required=False, default=None)
+        args = parser.parse_args()
+        memory_id = args.get("id")
+
+        if memory_id:
+            ChatflowMemoryService.delete_memory(app_model, memory_id, MemoryCreatedBy(end_user_id=end_user.id))
+            return "", 204
+        else:
+            ChatflowMemoryService.delete_all_user_memories(app_model, MemoryCreatedBy(end_user_id=end_user.id))
+            return "", 200
+
+
+api.add_resource(MemoryListApi, "/memories")
+api.add_resource(MemoryEditApi, "/memory-edit")
+api.add_resource(MemoryDeleteApi, "/memories")

api/core/app/apps/advanced_chat/app_runner.py

@@ -1,10 +1,11 @@
 import logging
 import time
-from collections.abc import Mapping
+from collections.abc import Mapping, MutableMapping, Sequence
 from typing import Any, cast
 
 from sqlalchemy import select
 from sqlalchemy.orm import Session
+from typing_extensions import override
 
 from core.app.apps.advanced_chat.app_config_manager import AdvancedChatAppConfig
 from core.app.apps.base_app_queue_manager import AppQueueManager
@@ -20,12 +21,16 @@ from core.app.entities.queue_entities import (
     QueueTextChunkEvent,
 )
 from core.app.features.annotation_reply.annotation_reply import AnnotationReplyFeature
+from core.memory.entities import MemoryCreatedBy, MemoryScope
+from core.model_runtime.entities import AssistantPromptMessage, UserPromptMessage
 from core.moderation.base import ModerationError
 from core.moderation.input_moderation import InputModeration
 from core.variables.variables import VariableUnion
 from core.workflow.enums import WorkflowType
 from core.workflow.graph_engine.command_channels.redis_channel import RedisChannel
+from core.workflow.graph_engine.layers.base import GraphEngineLayer
 from core.workflow.graph_engine.layers.persistence import PersistenceWorkflowInfo, WorkflowPersistenceLayer
+from core.workflow.graph_events import GraphRunSucceededEvent
 from core.workflow.repositories.workflow_execution_repository import WorkflowExecutionRepository
 from core.workflow.repositories.workflow_node_execution_repository import WorkflowNodeExecutionRepository
 from core.workflow.runtime import GraphRuntimeState, VariablePool
@@ -38,6 +43,8 @@ from models import Workflow
 from models.enums import UserFrom
 from models.model import App, Conversation, Message, MessageAnnotation
 from models.workflow import ConversationVariable
+from services.chatflow_history_service import ChatflowHistoryService
+from services.chatflow_memory_service import ChatflowMemoryService
 
 logger = logging.getLogger(__name__)
 
@@ -61,11 +68,13 @@ class AdvancedChatAppRunner(WorkflowBasedAppRunner):
         app: App,
         workflow_execution_repository: WorkflowExecutionRepository,
         workflow_node_execution_repository: WorkflowNodeExecutionRepository,
+        graph_engine_layers: Sequence[GraphEngineLayer] = (),
     ):
         super().__init__(
             queue_manager=queue_manager,
             variable_loader=variable_loader,
             app_id=application_generate_entity.app_config.app_id,
+            graph_engine_layers=graph_engine_layers,
         )
         self.application_generate_entity = application_generate_entity
         self.conversation = conversation
@@ -78,6 +87,10 @@ class AdvancedChatAppRunner(WorkflowBasedAppRunner):
         self._workflow_node_execution_repository = workflow_node_execution_repository
 
     def run(self):
+        ChatflowMemoryService.wait_for_sync_memory_completion(
+            workflow=self._workflow, conversation_id=self.conversation.id
+        )
+
         app_config = self.application_generate_entity.app_config
         app_config = cast(AdvancedChatAppConfig, app_config)
 
@@ -140,6 +153,7 @@ class AdvancedChatAppRunner(WorkflowBasedAppRunner):
                 # Based on the definition of `VariableUnion`,
                 # `list[Variable]` can be safely used as `list[VariableUnion]` since they are compatible.
                 conversation_variables=conversation_variables,
+                memory_blocks=self._fetch_memory_blocks(),
             )
 
             # init graph
@@ -195,12 +209,39 @@ class AdvancedChatAppRunner(WorkflowBasedAppRunner):
         )
 
         workflow_entry.graph_engine.layer(persistence_layer)
+        for layer in self._graph_engine_layers:
+            workflow_entry.graph_engine.layer(layer)
 
         generator = workflow_entry.run()
 
         for event in generator:
             self._handle_event(workflow_entry, event)
 
+        try:
+            self._check_app_memory_updates(variable_pool)
+        except Exception as e:
+            logger.exception("Failed to check app memory updates", exc_info=e)
+
+    @override
+    def _handle_event(self, workflow_entry: WorkflowEntry, event: Any) -> None:
+        super()._handle_event(workflow_entry, event)
+        if isinstance(event, GraphRunSucceededEvent):
+            workflow_outputs = event.outputs
+            if not workflow_outputs:
+                logger.warning("Chatflow output is empty.")
+                return
+            assistant_message = workflow_outputs.get("answer")
+            if not assistant_message:
+                logger.warning("Chatflow output does not contain 'answer'.")
+                return
+            if not isinstance(assistant_message, str):
+                logger.warning("Chatflow output 'answer' is not a string.")
+                return
+            try:
+                self._sync_conversation_to_chatflow_tables(assistant_message)
+            except Exception as e:
+                logger.exception("Failed to sync conversation to memory tables", exc_info=e)
+
     def handle_input_moderation(
         self,
         app_record: App,
@@ -398,3 +439,67 @@ class AdvancedChatAppRunner(WorkflowBasedAppRunner):
 
         # Return combined list
         return existing_variables + new_variables
+
+    def _fetch_memory_blocks(self) -> Mapping[str, str]:
+        """fetch all memory blocks for current app"""
+
+        memory_blocks_dict: MutableMapping[str, str] = {}
+        is_draft = self.application_generate_entity.invoke_from == InvokeFrom.DEBUGGER
+        conversation_id = self.conversation.id
+        memory_block_specs = self._workflow.memory_blocks
+        # Get runtime memory values
+        memories = ChatflowMemoryService.get_memories_by_specs(
+            memory_block_specs=memory_block_specs,
+            tenant_id=self._workflow.tenant_id,
+            app_id=self._workflow.app_id,
+            node_id=None,
+            conversation_id=conversation_id,
+            is_draft=is_draft,
+            created_by=self._get_created_by(),
+        )
+
+        # Build memory_id -> value mapping
+        for memory in memories:
+            if memory.spec.scope == MemoryScope.APP:
+                # App level: use memory_id directly
+                memory_blocks_dict[memory.spec.id] = memory.value
+            else:  # NODE scope
+                node_id = memory.node_id
+                if not node_id:
+                    logger.warning("Memory block %s has no node_id, skip.", memory.spec.id)
+                    continue
+                key = f"{node_id}.{memory.spec.id}"
+                memory_blocks_dict[key] = memory.value
+
+        return memory_blocks_dict
+
+    def _sync_conversation_to_chatflow_tables(self, assistant_message: str):
+        ChatflowHistoryService.save_app_message(
+            prompt_message=UserPromptMessage(content=(self.application_generate_entity.query)),
+            conversation_id=self.conversation.id,
+            app_id=self._workflow.app_id,
+            tenant_id=self._workflow.tenant_id,
+        )
+        ChatflowHistoryService.save_app_message(
+            prompt_message=AssistantPromptMessage(content=assistant_message),
+            conversation_id=self.conversation.id,
+            app_id=self._workflow.app_id,
+            tenant_id=self._workflow.tenant_id,
+        )
+
+    def _check_app_memory_updates(self, variable_pool: VariablePool):
+        is_draft = self.application_generate_entity.invoke_from == InvokeFrom.DEBUGGER
+
+        ChatflowMemoryService.update_app_memory_if_needed(
+            workflow=self._workflow,
+            conversation_id=self.conversation.id,
+            variable_pool=variable_pool,
+            is_draft=is_draft,
+            created_by=self._get_created_by(),
+        )
+
+    def _get_created_by(self) -> MemoryCreatedBy:
+        if self.application_generate_entity.invoke_from in {InvokeFrom.DEBUGGER, InvokeFrom.EXPLORE}:
+            return MemoryCreatedBy(account_id=self.application_generate_entity.user_id)
+        else:
+            return MemoryCreatedBy(end_user_id=self.application_generate_entity.user_id)

api/core/app/apps/advanced_chat/generate_task_pipeline.py

@@ -1,3 +1,4 @@
+import json
 import logging
 import re
 import time
@@ -60,6 +61,7 @@ from core.app.task_pipeline.based_generate_task_pipeline import BasedGenerateTas
 from core.app.task_pipeline.message_cycle_manager import MessageCycleManager
 from core.base.tts import AppGeneratorTTSPublisher, AudioTrunk
 from core.model_runtime.entities.llm_entities import LLMUsage
+from core.model_runtime.utils.encoders import jsonable_encoder
 from core.ops.ops_trace_manager import TraceQueueManager
 from core.workflow.enums import WorkflowExecutionStatus
 from core.workflow.nodes import NodeType
@@ -391,6 +393,14 @@ class AdvancedChatAppGenerateTaskPipeline(GraphRuntimeStateSupport):
         if should_direct_answer:
             return
 
+        current_time = time.perf_counter()
+        if self._task_state.first_token_time is None and delta_text.strip():
+            self._task_state.first_token_time = current_time
+            self._task_state.is_streaming_response = True
+
+        if delta_text.strip():
+            self._task_state.last_token_time = current_time
+
         # Only publish tts message at text chunk streaming
         if tts_publisher and queue_message:
             tts_publisher.publish(queue_message)
@@ -772,7 +782,33 @@ class AdvancedChatAppGenerateTaskPipeline(GraphRuntimeStateSupport):
         message.answer = answer_text
         message.updated_at = naive_utc_now()
         message.provider_response_latency = time.perf_counter() - self._base_task_pipeline.start_at
-        message.message_metadata = self._task_state.metadata.model_dump_json()
+
+        # Set usage first before dumping metadata
+        if graph_runtime_state and graph_runtime_state.llm_usage:
+            usage = graph_runtime_state.llm_usage
+            message.message_tokens = usage.prompt_tokens
+            message.message_unit_price = usage.prompt_unit_price
+            message.message_price_unit = usage.prompt_price_unit
+            message.answer_tokens = usage.completion_tokens
+            message.answer_unit_price = usage.completion_unit_price
+            message.answer_price_unit = usage.completion_price_unit
+            message.total_price = usage.total_price
+            message.currency = usage.currency
+            self._task_state.metadata.usage = usage
+        else:
+            usage = LLMUsage.empty_usage()
+            self._task_state.metadata.usage = usage
+
+        # Add streaming metrics to usage if available
+        if self._task_state.is_streaming_response and self._task_state.first_token_time:
+            start_time = self._base_task_pipeline.start_at
+            first_token_time = self._task_state.first_token_time
+            last_token_time = self._task_state.last_token_time or first_token_time
+            usage.time_to_first_token = round(first_token_time - start_time, 3)
+            usage.time_to_generate = round(last_token_time - first_token_time, 3)
+
+        metadata = self._task_state.metadata.model_dump()
+        message.message_metadata = json.dumps(jsonable_encoder(metadata))
         message_files = [
             MessageFile(
                 message_id=message.id,
@@ -790,20 +826,6 @@ class AdvancedChatAppGenerateTaskPipeline(GraphRuntimeStateSupport):
         ]
         session.add_all(message_files)
 
-        if graph_runtime_state and graph_runtime_state.llm_usage:
-            usage = graph_runtime_state.llm_usage
-            message.message_tokens = usage.prompt_tokens
-            message.message_unit_price = usage.prompt_unit_price
-            message.message_price_unit = usage.prompt_price_unit
-            message.answer_tokens = usage.completion_tokens
-            message.answer_unit_price = usage.completion_unit_price
-            message.answer_price_unit = usage.completion_price_unit
-            message.total_price = usage.total_price
-            message.currency = usage.currency
-            self._task_state.metadata.usage = usage
-        else:
-            self._task_state.metadata.usage = LLMUsage.empty_usage()
-
     def _seed_graph_runtime_state_from_queue_manager(self) -> None:
         """Bootstrap the cached runtime state from the queue manager when present."""
         candidate = self._base_task_pipeline.queue_manager.graph_runtime_state

api/core/app/apps/agent_chat/app_runner.py

@@ -144,7 +144,7 @@ class AgentChatAppRunner(AppRunner):
             prompt_template_entity=app_config.prompt_template,
             inputs=dict(inputs),
             files=list(files),
-            query=query or "",
+            query=query,
             memory=memory,
         )
 
@@ -172,7 +172,7 @@ class AgentChatAppRunner(AppRunner):
             prompt_template_entity=app_config.prompt_template,
             inputs=dict(inputs),
             files=list(files),
-            query=query or "",
+            query=query,
             memory=memory,
         )
 

api/core/app/apps/base_app_runner.py

@@ -79,7 +79,7 @@ class AppRunner:
         prompt_template_entity: PromptTemplateEntity,
         inputs: Mapping[str, str],
         files: Sequence["File"],
-        query: str | None = None,
+        query: str = "",
         context: str | None = None,
         memory: TokenBufferMemory | None = None,
         image_detail_config: ImagePromptMessageContent.DETAIL | None = None,
@@ -105,7 +105,7 @@ class AppRunner:
                 app_mode=AppMode.value_of(app_record.mode),
                 prompt_template_entity=prompt_template_entity,
                 inputs=inputs,
-                query=query or "",
+                query=query,
                 files=files,
                 context=context,
                 memory=memory,

api/core/app/apps/message_based_app_generator.py

@@ -190,7 +190,7 @@ class MessageBasedAppGenerator(BaseAppGenerator):
             override_model_configs=json.dumps(override_model_configs) if override_model_configs else None,
             conversation_id=conversation.id,
             inputs=application_generate_entity.inputs,
-            query=application_generate_entity.query or "",
+            query=application_generate_entity.query,
             message="",
             message_tokens=0,
             message_unit_price=0,

api/core/app/apps/pipeline/pipeline_generator.py

@@ -40,19 +40,15 @@ from core.workflow.repositories.draft_variable_repository import DraftVariableSa
 from core.workflow.repositories.workflow_execution_repository import WorkflowExecutionRepository
 from core.workflow.repositories.workflow_node_execution_repository import WorkflowNodeExecutionRepository
 from core.workflow.variable_loader import DUMMY_VARIABLE_LOADER, VariableLoader
+from enums.cloud_plan import CloudPlan
 from extensions.ext_database import db
-from extensions.ext_redis import redis_client
 from libs.flask_utils import preserve_flask_contexts
 from models import Account, EndUser, Workflow, WorkflowNodeExecutionTriggeredFrom
 from models.dataset import Document, DocumentPipelineExecutionLog, Pipeline
 from models.enums import WorkflowRunTriggeredFrom
 from models.model import AppMode
 from services.datasource_provider_service import DatasourceProviderService
-from services.feature_service import FeatureService
-from services.file_service import FileService
 from services.workflow_draft_variable_service import DraftVarLoader, WorkflowDraftVariableService
-from tasks.rag_pipeline.priority_rag_pipeline_run_task import priority_rag_pipeline_run_task
-from tasks.rag_pipeline.rag_pipeline_run_task import rag_pipeline_run_task
 
 logger = logging.getLogger(__name__)
 
@@ -255,7 +251,7 @@ class PipelineGenerator(BaseAppGenerator):
             json_text = json.dumps(text)
             upload_file = FileService(db.engine).upload_text(json_text, name, user.id, dataset.tenant_id)
             features = FeatureService.get_features(dataset.tenant_id)
-            if features.billing.enabled and features.billing.subscription.plan == "sandbox":
+            if features.billing.enabled and features.billing.subscription.plan == CloudPlan.SANDBOX:
                 tenant_pipeline_task_key = f"tenant_pipeline_task:{dataset.tenant_id}"
                 tenant_self_pipeline_task_queue = f"tenant_self_pipeline_task_queue:{dataset.tenant_id}"
 

api/core/app/apps/workflow/app_runner.py

@@ -135,6 +135,8 @@ class WorkflowAppRunner(WorkflowBasedAppRunner):
         )
 
         workflow_entry.graph_engine.layer(persistence_layer)
+        for layer in self._graph_engine_layers:
+            workflow_entry.graph_engine.layer(layer)
 
         generator = workflow_entry.run()
 

api/core/app/apps/workflow_app_runner.py

@@ -1,5 +1,5 @@
 import time
-from collections.abc import Mapping
+from collections.abc import Mapping, Sequence
 from typing import Any, cast
 
 from core.app.apps.base_app_queue_manager import AppQueueManager, PublishFrom
@@ -27,6 +27,7 @@ from core.app.entities.queue_entities import (
 )
 from core.workflow.entities import GraphInitParams
 from core.workflow.graph import Graph
+from core.workflow.graph_engine.layers.base import GraphEngineLayer
 from core.workflow.graph_events import (
     GraphEngineEvent,
     GraphRunFailedEvent,
@@ -69,10 +70,12 @@ class WorkflowBasedAppRunner:
         queue_manager: AppQueueManager,
         variable_loader: VariableLoader = DUMMY_VARIABLE_LOADER,
         app_id: str,
+        graph_engine_layers: Sequence[GraphEngineLayer] = (),
     ):
         self._queue_manager = queue_manager
         self._variable_loader = variable_loader
         self._app_id = app_id
+        self._graph_engine_layers = graph_engine_layers
 
     def _init_graph(
         self,

api/core/app/entities/app_invoke_entities.py

@@ -129,7 +129,7 @@ class EasyUIBasedAppGenerateEntity(AppGenerateEntity):
     app_config: EasyUIBasedAppConfig = None  # type: ignore
     model_conf: ModelConfigWithCredentialsEntity
 
-    query: str | None = None
+    query: str = ""
 
     # pydantic configs
     model_config = ConfigDict(protected_namespaces=())

api/core/app/entities/task_entities.py

@@ -48,6 +48,9 @@ class WorkflowTaskState(TaskState):
     """
 
     answer: str = ""
+    first_token_time: float | None = None
+    last_token_time: float | None = None
+    is_streaming_response: bool = False
 
 
 class StreamEvent(StrEnum):

api/core/app/layers/pause_state_persist_layer.py

@@ -0,0 +1,71 @@
+from sqlalchemy import Engine
+from sqlalchemy.orm import sessionmaker
+
+from core.workflow.graph_engine.layers.base import GraphEngineLayer
+from core.workflow.graph_events.base import GraphEngineEvent
+from core.workflow.graph_events.graph import GraphRunPausedEvent
+from repositories.api_workflow_run_repository import APIWorkflowRunRepository
+from repositories.factory import DifyAPIRepositoryFactory
+
+
+class PauseStatePersistenceLayer(GraphEngineLayer):
+    def __init__(self, session_factory: Engine | sessionmaker, state_owner_user_id: str):
+        """Create a PauseStatePersistenceLayer.
+
+        The `state_owner_user_id` is used when creating state file for pause.
+        It generally should id of the creator of workflow.
+        """
+        if isinstance(session_factory, Engine):
+            session_factory = sessionmaker(session_factory)
+        self._session_maker = session_factory
+        self._state_owner_user_id = state_owner_user_id
+
+    def _get_repo(self) -> APIWorkflowRunRepository:
+        return DifyAPIRepositoryFactory.create_api_workflow_run_repository(self._session_maker)
+
+    def on_graph_start(self) -> None:
+        """
+        Called when graph execution starts.
+
+        This is called after the engine has been initialized but before any nodes
+        are executed. Layers can use this to set up resources or log start information.
+        """
+        pass
+
+    def on_event(self, event: GraphEngineEvent) -> None:
+        """
+        Called for every event emitted by the engine.
+
+        This method receives all events generated during graph execution, including:
+        - Graph lifecycle events (start, success, failure)
+        - Node execution events (start, success, failure, retry)
+        - Stream events for response nodes
+        - Container events (iteration, loop)
+
+        Args:
+            event: The event emitted by the engine
+        """
+        if not isinstance(event, GraphRunPausedEvent):
+            return
+
+        assert self.graph_runtime_state is not None
+        workflow_run_id: str | None = self.graph_runtime_state.system_variable.workflow_execution_id
+        assert workflow_run_id is not None
+        repo = self._get_repo()
+        repo.create_workflow_pause(
+            workflow_run_id=workflow_run_id,
+            state_owner_user_id=self._state_owner_user_id,
+            state=self.graph_runtime_state.dumps(),
+        )
+
+    def on_graph_end(self, error: Exception | None) -> None:
+        """
+        Called when graph execution ends.
+
+        This is called after all nodes have been executed or when execution is
+        aborted. Layers can use this to clean up resources or log final state.
+
+        Args:
+            error: The exception that caused execution to fail, or None if successful
+        """
+        pass

api/core/app/task_pipeline/easy_ui_based_generate_task_pipeline.py

@@ -121,7 +121,7 @@ class EasyUIBasedGenerateTaskPipeline(BasedGenerateTaskPipeline):
         if self._application_generate_entity.app_config.app_mode != AppMode.COMPLETION:
             # start generate conversation name thread
             self._conversation_name_generate_thread = self._message_cycle_manager.generate_conversation_name(
-                conversation_id=self._conversation_id, query=self._application_generate_entity.query or ""
+                conversation_id=self._conversation_id, query=self._application_generate_entity.query
             )
 
         generator = self._wrapper_process_stream_response(trace_manager=self._application_generate_entity.trace_manager)

api/core/app/task_pipeline/message_cycle_manager.py

@@ -140,7 +140,27 @@ class MessageCycleManager:
         if not self._application_generate_entity.app_config.additional_features:
             raise ValueError("Additional features not found")
         if self._application_generate_entity.app_config.additional_features.show_retrieve_source:
-            self._task_state.metadata.retriever_resources = event.retriever_resources
+            merged_resources = [r for r in self._task_state.metadata.retriever_resources or [] if r]
+            existing_ids = {(r.dataset_id, r.document_id) for r in merged_resources if r.dataset_id and r.document_id}
+
+            # Add new unique resources from the event
+            for resource in event.retriever_resources or []:
+                if not resource:
+                    continue
+
+                is_duplicate = (
+                    resource.dataset_id
+                    and resource.document_id
+                    and (resource.dataset_id, resource.document_id) in existing_ids
+                )
+
+                if not is_duplicate:
+                    merged_resources.append(resource)
+
+            for i, resource in enumerate(merged_resources, 1):
+                resource.position = i
+
+            self._task_state.metadata.retriever_resources = merged_resources
 
     def message_file_to_stream_response(self, event: QueueMessageFileEvent) -> MessageFileStreamResponse | None:
         """

api/core/entities/document_task.py

@@ -0,0 +1,13 @@
+from dataclasses import dataclass
+
+
+@dataclass
+class DocumentTask:
+    """Document task entity for document indexing operations.
+    
+    This class represents a document indexing task that can be queued
+    and processed by the document indexing system.
+    """
+    tenant_id: str
+    dataset_id: str
+    document_ids: list[str]

api/core/entities/mcp_provider.py

@@ -272,6 +272,7 @@ class MCPProviderEntity(BaseModel):
 
     def _decrypt_dict(self, data: dict[str, Any]) -> dict[str, Any]:
         """Generic method to decrypt dictionary fields"""
+
         if not data:
             return {}
 

api/core/file/models.py

@@ -74,6 +74,10 @@ class File(BaseModel):
         storage_key: str | None = None,
         dify_model_identity: str | None = FILE_MODEL_IDENTITY,
         url: str | None = None,
+        # Legacy compatibility fields - explicitly handle known extra fields
+        tool_file_id: str | None = None,
+        upload_file_id: str | None = None,
+        datasource_file_id: str | None = None,
     ):
         super().__init__(
             id=id,

api/core/helper/code_executor/javascript/javascript_transformer.py

@@ -6,10 +6,7 @@ from core.helper.code_executor.template_transformer import TemplateTransformer
 class NodeJsTemplateTransformer(TemplateTransformer):
     @classmethod
     def get_runner_script(cls) -> str:
-        runner_script = dedent(
-            f"""
-            // declare main function
-            {cls._code_placeholder}
+        runner_script = dedent(f"""            {cls._code_placeholder}
 
             // decode and prepare input object
             var inputs_obj = JSON.parse(Buffer.from('{cls._inputs_placeholder}', 'base64').toString('utf-8'))
@@ -21,6 +18,5 @@ class NodeJsTemplateTransformer(TemplateTransformer):
             var output_json = JSON.stringify(output_obj)
             var result = `<<RESULT>>${{output_json}}<<RESULT>>`
             console.log(result)
-            """
-        )
+            """)
         return runner_script

api/core/helper/code_executor/python3/python3_transformer.py

@@ -6,9 +6,7 @@ from core.helper.code_executor.template_transformer import TemplateTransformer
 class Python3TemplateTransformer(TemplateTransformer):
     @classmethod
     def get_runner_script(cls) -> str:
-        runner_script = dedent(f"""
-            # declare main function
-            {cls._code_placeholder}
+        runner_script = dedent(f"""            {cls._code_placeholder}
 
             import json
             from base64 import b64decode

api/core/helper/marketplace.py

@@ -29,6 +29,18 @@ def batch_fetch_plugin_manifests(plugin_ids: list[str]) -> Sequence[MarketplaceP
     return [MarketplacePluginDeclaration.model_validate(plugin) for plugin in response.json()["data"]["plugins"]]
 
 
+def batch_fetch_plugin_by_ids(plugin_ids: list[str]) -> list[dict]:
+    if not plugin_ids:
+        return []
+
+    url = str(marketplace_api_url / "api/v1/plugins/batch")
+    response = httpx.post(url, json={"plugin_ids": plugin_ids}, headers={"X-Dify-Version": dify_config.project.version})
+    response.raise_for_status()
+
+    data = response.json()
+    return data.get("data", {}).get("plugins", [])
+
+
 def batch_fetch_plugin_manifests_ignore_deserialization_error(
     plugin_ids: list[str],
 ) -> Sequence[MarketplacePluginDeclaration]:

api/core/hosting_configuration.py

@@ -56,6 +56,9 @@ class HostingConfiguration:
         self.provider_map[f"{DEFAULT_PLUGIN_ID}/minimax/minimax"] = self.init_minimax()
         self.provider_map[f"{DEFAULT_PLUGIN_ID}/spark/spark"] = self.init_spark()
         self.provider_map[f"{DEFAULT_PLUGIN_ID}/zhipuai/zhipuai"] = self.init_zhipuai()
+        self.provider_map[f"{DEFAULT_PLUGIN_ID}/gemini/google"] = self.init_gemini()
+        self.provider_map[f"{DEFAULT_PLUGIN_ID}/x/x"] = self.init_xai()
+        self.provider_map[f"{DEFAULT_PLUGIN_ID}/deepseek/deepseek"] = self.init_deepseek()
 
         self.moderation_config = self.init_moderation_config()
 
@@ -128,7 +131,7 @@ class HostingConfiguration:
         quotas: list[HostingQuota] = []
 
         if dify_config.HOSTED_OPENAI_TRIAL_ENABLED:
-            hosted_quota_limit = dify_config.HOSTED_OPENAI_QUOTA_LIMIT
+            hosted_quota_limit = 0
             trial_models = self.parse_restrict_models_from_env("HOSTED_OPENAI_TRIAL_MODELS")
             trial_quota = TrialHostingQuota(quota_limit=hosted_quota_limit, restrict_models=trial_models)
             quotas.append(trial_quota)
@@ -156,18 +159,49 @@ class HostingConfiguration:
             quota_unit=quota_unit,
         )
 
-    @staticmethod
-    def init_anthropic() -> HostingProvider:
-        quota_unit = QuotaUnit.TOKENS
+    def init_gemini(self) -> HostingProvider:
+        quota_unit = QuotaUnit.CREDITS
+        quotas: list[HostingQuota] = []
+
+        if dify_config.HOSTED_GEMINI_TRIAL_ENABLED:
+            hosted_quota_limit = 0
+            trial_models = self.parse_restrict_models_from_env("HOSTED_GEMINI_TRIAL_MODELS")
+            trial_quota = TrialHostingQuota(quota_limit=hosted_quota_limit, restrict_models=trial_models)
+            quotas.append(trial_quota)
+
+        if dify_config.HOSTED_GEMINI_PAID_ENABLED:
+            paid_models = self.parse_restrict_models_from_env("HOSTED_GEMINI_PAID_MODELS")
+            paid_quota = PaidHostingQuota(restrict_models=paid_models)
+            quotas.append(paid_quota)
+
+        if len(quotas) > 0:
+            credentials = {
+                "google_api_key": dify_config.HOSTED_GEMINI_API_KEY,
+            }
+
+            if dify_config.HOSTED_GEMINI_API_BASE:
+                credentials["google_base_url"] = dify_config.HOSTED_GEMINI_API_BASE
+
+            return HostingProvider(enabled=True, credentials=credentials, quota_unit=quota_unit, quotas=quotas)
+
+        return HostingProvider(
+            enabled=False,
+            quota_unit=quota_unit,
+        )
+
+    def init_anthropic(self) -> HostingProvider:
+        quota_unit = QuotaUnit.CREDITS
         quotas: list[HostingQuota] = []
 
         if dify_config.HOSTED_ANTHROPIC_TRIAL_ENABLED:
-            hosted_quota_limit = dify_config.HOSTED_ANTHROPIC_QUOTA_LIMIT
-            trial_quota = TrialHostingQuota(quota_limit=hosted_quota_limit)
+            hosted_quota_limit = 0
+            trail_models = self.parse_restrict_models_from_env("HOSTED_ANTHROPIC_TRIAL_MODELS")
+            trial_quota = TrialHostingQuota(quota_limit=hosted_quota_limit, restrict_models=trail_models)
             quotas.append(trial_quota)
 
         if dify_config.HOSTED_ANTHROPIC_PAID_ENABLED:
-            paid_quota = PaidHostingQuota()
+            paid_models = self.parse_restrict_models_from_env("HOSTED_ANTHROPIC_PAID_MODELS")
+            paid_quota = PaidHostingQuota(restrict_models=paid_models)
             quotas.append(paid_quota)
 
         if len(quotas) > 0:
@@ -185,6 +219,66 @@ class HostingConfiguration:
             quota_unit=quota_unit,
         )
 
+    def init_xai(self) -> HostingProvider:
+        quota_unit = QuotaUnit.CREDITS
+        quotas: list[HostingQuota] = []
+
+        if dify_config.HOSTED_XAI_TRIAL_ENABLED:
+            hosted_quota_limit = 0
+            trail_models = self.parse_restrict_models_from_env("HOSTED_XAI_TRIAL_MODELS")
+            trial_quota = TrialHostingQuota(quota_limit=hosted_quota_limit, restrict_models=trail_models)
+            quotas.append(trial_quota)
+
+        if dify_config.HOSTED_XAI_PAID_ENABLED:
+            paid_models = self.parse_restrict_models_from_env("HOSTED_XAI_PAID_MODELS")
+            paid_quota = PaidHostingQuota(restrict_models=paid_models)
+            quotas.append(paid_quota)
+
+        if len(quotas) > 0:
+            credentials = {
+                "api_key": dify_config.HOSTED_XAI_API_KEY,
+            }
+
+            if dify_config.HOSTED_XAI_API_BASE:
+                credentials["endpoint_url"] = dify_config.HOSTED_XAI_API_BASE
+
+            return HostingProvider(enabled=True, credentials=credentials, quota_unit=quota_unit, quotas=quotas)
+
+        return HostingProvider(
+            enabled=False,
+            quota_unit=quota_unit,
+        )
+
+    def init_deepseek(self) -> HostingProvider:
+        quota_unit = QuotaUnit.CREDITS
+        quotas: list[HostingQuota] = []
+
+        if dify_config.HOSTED_DEEPSEEK_TRIAL_ENABLED:
+            hosted_quota_limit = 0
+            trail_models = self.parse_restrict_models_from_env("HOSTED_DEEPSEEK_TRIAL_MODELS")
+            trial_quota = TrialHostingQuota(quota_limit=hosted_quota_limit, restrict_models=trail_models)
+            quotas.append(trial_quota)
+
+        if dify_config.HOSTED_DEEPSEEK_PAID_ENABLED:
+            paid_models = self.parse_restrict_models_from_env("HOSTED_DEEPSEEK_PAID_MODELS")
+            paid_quota = PaidHostingQuota(restrict_models=paid_models)
+            quotas.append(paid_quota)
+
+        if len(quotas) > 0:
+            credentials = {
+                "api_key": dify_config.HOSTED_DEEPSEEK_API_KEY,
+            }
+
+            if dify_config.HOSTED_DEEPSEEK_API_BASE:
+                credentials["endpoint_url"] = dify_config.HOSTED_DEEPSEEK_API_BASE
+
+            return HostingProvider(enabled=True, credentials=credentials, quota_unit=quota_unit, quotas=quotas)
+
+        return HostingProvider(
+            enabled=False,
+            quota_unit=quota_unit,
+        )
+
     @staticmethod
     def init_minimax() -> HostingProvider:
         quota_unit = QuotaUnit.TOKENS

api/core/llm_generator/llm_generator.py

@@ -14,10 +14,12 @@ from core.llm_generator.prompts import (
     JAVASCRIPT_CODE_GENERATOR_PROMPT_TEMPLATE,
     LLM_MODIFY_CODE_SYSTEM,
     LLM_MODIFY_PROMPT_SYSTEM,
+    MEMORY_UPDATE_PROMPT,
     PYTHON_CODE_GENERATOR_PROMPT_TEMPLATE,
     SYSTEM_STRUCTURED_OUTPUT_GENERATE,
     WORKFLOW_RULE_CONFIG_PROMPT_GENERATE_TEMPLATE,
 )
+from core.memory.entities import MemoryBlock, MemoryBlockSpec
 from core.model_manager import ModelManager
 from core.model_runtime.entities.llm_entities import LLMResult
 from core.model_runtime.entities.message_entities import PromptMessage, SystemPromptMessage, UserPromptMessage
@@ -28,6 +30,7 @@ from core.ops.ops_trace_manager import TraceQueueManager, TraceTask
 from core.ops.utils import measure_time
 from core.prompt.utils.prompt_template_parser import PromptTemplateParser
 from core.workflow.entities.workflow_node_execution import WorkflowNodeExecutionMetadataKey
+from core.workflow.runtime.variable_pool import VariablePool
 from extensions.ext_database import db
 from extensions.ext_storage import storage
 from models import App, Message, WorkflowNodeExecutionModel
@@ -562,3 +565,35 @@ class LLMGenerator:
                 "Failed to invoke LLM model, model: %s", json.dumps(model_config.get("name")), exc_info=True
             )
             return {"error": f"An unexpected error occurred: {str(e)}"}
+
+    @staticmethod
+    def update_memory_block(
+        tenant_id: str,
+        visible_history: Sequence[tuple[str, str]],
+        variable_pool: VariablePool,
+        memory_block: MemoryBlock,
+        memory_spec: MemoryBlockSpec,
+    ) -> str:
+        model_instance = ModelManager().get_model_instance(
+            tenant_id=tenant_id,
+            provider=memory_spec.model.provider,
+            model=memory_spec.model.name,
+            model_type=ModelType.LLM,
+        )
+        formatted_history = ""
+        for sender, message in visible_history:
+            formatted_history += f"{sender}: {message}\n"
+        filled_instruction = variable_pool.convert_template(memory_spec.instruction).text
+        formatted_prompt = PromptTemplateParser(MEMORY_UPDATE_PROMPT).format(
+            inputs={
+                "formatted_history": formatted_history,
+                "current_value": memory_block.value,
+                "instruction": filled_instruction,
+            }
+        )
+        llm_result = model_instance.invoke_llm(
+            prompt_messages=[UserPromptMessage(content=formatted_prompt)],
+            model_parameters=memory_spec.model.completion_params,
+            stream=False,
+        )
+        return llm_result.message.get_text_content()

api/core/llm_generator/prompts.py

@@ -422,3 +422,18 @@ INSTRUCTION_GENERATE_TEMPLATE_PROMPT = """The output of this prompt is not as ex
 You should edit the prompt according to the IDEAL OUTPUT."""
 
 INSTRUCTION_GENERATE_TEMPLATE_CODE = """Please fix the errors in the {{#error_message#}}."""
+
+MEMORY_UPDATE_PROMPT = """
+Based on the following conversation history, update the memory content:
+
+Conversation history:
+{{formatted_history}}
+
+Current memory:
+{{current_value}}
+
+Update instruction:
+{{instruction}}
+
+Please output only the updated memory content, no other text like greeting:
+"""

api/core/mcp/session/client_session.py

@@ -109,12 +109,16 @@ class ClientSession(
         self._message_handler = message_handler or _default_message_handler
 
     def initialize(self) -> types.InitializeResult:
-        sampling = types.SamplingCapability()
-        roots = types.RootsCapability(
-            # TODO: Should this be based on whether we
-            # _will_ send notifications, or only whether
-            # they're supported?
-            listChanged=True,
+        # Only set capabilities if non-default callbacks are provided
+        # This prevents servers from attempting callbacks when we don't actually support them
+        sampling = types.SamplingCapability() if self._sampling_callback is not _default_sampling_callback else None
+        roots = (
+            types.RootsCapability(
+                # Only enable listChanged if we have a custom callback
+                listChanged=True,
+            )
+            if self._list_roots_callback is not _default_list_roots_callback
+            else None
         )
 
         result = self.send_request(

api/core/memory/entities.py

@@ -0,0 +1,119 @@
+from __future__ import annotations
+
+from enum import StrEnum
+from typing import TYPE_CHECKING, Optional
+from uuid import uuid4
+
+from pydantic import BaseModel, Field
+
+if TYPE_CHECKING:
+    from core.app.app_config.entities import ModelConfig
+
+
+class MemoryScope(StrEnum):
+    """Memory scope determined by node_id field"""
+    APP = "app"  # node_id is None
+    NODE = "node"  # node_id is not None
+
+
+class MemoryTerm(StrEnum):
+    """Memory term determined by conversation_id field"""
+    SESSION = "session"  # conversation_id is not None
+    PERSISTENT = "persistent"  # conversation_id is None
+
+
+class MemoryStrategy(StrEnum):
+    ON_TURNS = "on_turns"
+
+
+class MemoryScheduleMode(StrEnum):
+    SYNC = "sync"
+    ASYNC = "async"
+
+
+class MemoryBlockSpec(BaseModel):
+    """Memory block specification for workflow configuration"""
+    id: str = Field(
+        default_factory=lambda: str(uuid4()),
+        description="Unique identifier for the memory block",
+    )
+    name: str = Field(description="Display name of the memory block")
+    description: str = Field(default="", description="Description of the memory block")
+    template: str = Field(description="Initial template content for the memory")
+    instruction: str = Field(description="Instructions for updating the memory")
+    scope: MemoryScope = Field(description="Scope of the memory (app or node level)")
+    term: MemoryTerm = Field(description="Term of the memory (session or persistent)")
+    strategy: MemoryStrategy = Field(description="Update strategy for the memory")
+    update_turns: int = Field(gt=0, description="Number of turns between updates")
+    preserved_turns: int = Field(gt=0, description="Number of conversation turns to preserve")
+    schedule_mode: MemoryScheduleMode = Field(description="Synchronous or asynchronous update mode")
+    model: ModelConfig = Field(description="Model configuration for memory updates")
+    end_user_visible: bool = Field(default=False, description="Whether memory is visible to end users")
+    end_user_editable: bool = Field(default=False, description="Whether memory is editable by end users")
+
+
+class MemoryCreatedBy(BaseModel):
+    end_user_id: str | None = None
+    account_id: str | None = None
+
+
+class MemoryBlock(BaseModel):
+    """Runtime memory block instance
+
+    Design Rules:
+    - app_id = None: Global memory (future feature, not implemented yet)
+    - app_id = str: App-specific memory
+    - conversation_id = None: Persistent memory (cross-conversation)
+    - conversation_id = str: Session memory (conversation-specific)
+    - node_id = None: App-level scope
+    - node_id = str: Node-level scope
+
+    These rules implicitly determine scope and term without redundant storage.
+    """
+    spec: MemoryBlockSpec
+    tenant_id: str
+    value: str
+    app_id: str
+    conversation_id: Optional[str] = None
+    node_id: Optional[str] = None
+    edited_by_user: bool = False
+    created_by: MemoryCreatedBy
+    version: int = Field(description="Memory block version number")
+
+
+class MemoryValueData(BaseModel):
+    value: str
+    edited_by_user: bool = False
+
+
+class ChatflowConversationMetadata(BaseModel):
+    """Metadata for chatflow conversation with visible message count"""
+    type: str = "mutable_visible_window"
+    visible_count: int = Field(gt=0, description="Number of visible messages to keep")
+
+
+class MemoryBlockWithConversation(MemoryBlock):
+    """MemoryBlock with optional conversation metadata for session memories"""
+    conversation_metadata: ChatflowConversationMetadata = Field(
+        description="Conversation metadata, only present for session memories"
+    )
+
+    @classmethod
+    def from_memory_block(
+        cls,
+        memory_block: MemoryBlock,
+        conversation_metadata: ChatflowConversationMetadata
+    ) -> MemoryBlockWithConversation:
+        """Create MemoryBlockWithConversation from MemoryBlock"""
+        return cls(
+            spec=memory_block.spec,
+            tenant_id=memory_block.tenant_id,
+            value=memory_block.value,
+            app_id=memory_block.app_id,
+            conversation_id=memory_block.conversation_id,
+            node_id=memory_block.node_id,
+            edited_by_user=memory_block.edited_by_user,
+            created_by=memory_block.created_by,
+            version=memory_block.version,
+            conversation_metadata=conversation_metadata
+        )

api/core/memory/errors.py

@@ -0,0 +1,6 @@
+class MemorySyncTimeoutError(Exception):
+    def __init__(self, app_id: str, conversation_id: str):
+        self.app_id = app_id
+        self.conversation_id = conversation_id
+        self.message = "Memory synchronization timeout after 50 seconds"
+        super().__init__(self.message)

api/core/model_runtime/entities/llm_entities.py

@@ -38,6 +38,8 @@ class LLMUsageMetadata(TypedDict, total=False):
     prompt_price: Union[float, str]
     completion_price: Union[float, str]
     latency: float
+    time_to_first_token: float
+    time_to_generate: float
 
 
 class LLMUsage(ModelUsage):
@@ -57,6 +59,8 @@ class LLMUsage(ModelUsage):
     total_price: Decimal
     currency: str
     latency: float
+    time_to_first_token: float | None = None
+    time_to_generate: float | None = None
 
     @classmethod
     def empty_usage(cls):
@@ -73,6 +77,8 @@ class LLMUsage(ModelUsage):
             total_price=Decimal("0.0"),
             currency="USD",
             latency=0.0,
+            time_to_first_token=None,
+            time_to_generate=None,
         )
 
     @classmethod
@@ -108,6 +114,8 @@ class LLMUsage(ModelUsage):
             prompt_price=Decimal(str(metadata.get("prompt_price", 0))),
             completion_price=Decimal(str(metadata.get("completion_price", 0))),
             latency=metadata.get("latency", 0.0),
+            time_to_first_token=metadata.get("time_to_first_token"),
+            time_to_generate=metadata.get("time_to_generate"),
         )
 
     def plus(self, other: LLMUsage) -> LLMUsage:
@@ -133,6 +141,8 @@ class LLMUsage(ModelUsage):
                 total_price=self.total_price + other.total_price,
                 currency=other.currency,
                 latency=self.latency + other.latency,
+                time_to_first_token=other.time_to_first_token,
+                time_to_generate=other.time_to_generate,
             )
 
     def __add__(self, other: LLMUsage) -> LLMUsage:

api/core/ops/entities/trace_entity.py

@@ -62,6 +62,9 @@ class MessageTraceInfo(BaseTraceInfo):
     file_list: Union[str, dict[str, Any], list] | None = None
     message_file_data: Any | None = None
     conversation_mode: str
+    gen_ai_server_time_to_first_token: float | None = None
+    llm_streaming_time_to_generate: float | None = None
+    is_streaming_request: bool = False
 
 
 class ModerationTraceInfo(BaseTraceInfo):

api/core/ops/ops_trace_manager.py

@@ -14,7 +14,7 @@ from flask import current_app
 from sqlalchemy import select
 from sqlalchemy.orm import Session, sessionmaker
 
-from core.helper.encrypter import decrypt_token, encrypt_token, obfuscated_token
+from core.helper.encrypter import batch_decrypt_token, encrypt_token, obfuscated_token
 from core.ops.entities.config_entity import (
     OPS_FILE_PATH,
     TracingProviderEnum,
@@ -141,6 +141,8 @@ provider_config_map = OpsTraceProviderConfigMap()
 
 class OpsTraceManager:
     ops_trace_instances_cache: LRUCache = LRUCache(maxsize=128)
+    decrypted_configs_cache: LRUCache = LRUCache(maxsize=128)
+    _decryption_cache_lock = threading.RLock()
 
     @classmethod
     def encrypt_tracing_config(
@@ -161,7 +163,7 @@ class OpsTraceManager:
             provider_config_map[tracing_provider]["other_keys"],
         )
 
-        new_config = {}
+        new_config: dict[str, Any] = {}
         # Encrypt necessary keys
         for key in secret_keys:
             if key in tracing_config:
@@ -191,20 +193,41 @@ class OpsTraceManager:
         :param tracing_config: tracing config
         :return:
         """
-        config_class, secret_keys, other_keys = (
-            provider_config_map[tracing_provider]["config_class"],
-            provider_config_map[tracing_provider]["secret_keys"],
-            provider_config_map[tracing_provider]["other_keys"],
+        config_json = json.dumps(tracing_config, sort_keys=True)
+        decrypted_config_key = (
+            tenant_id,
+            tracing_provider,
+            config_json,
         )
-        new_config = {}
-        for key in secret_keys:
-            if key in tracing_config:
-                new_config[key] = decrypt_token(tenant_id, tracing_config[key])
 
-        for key in other_keys:
-            new_config[key] = tracing_config.get(key, "")
+        # First check without lock for performance
+        cached_config = cls.decrypted_configs_cache.get(decrypted_config_key)
+        if cached_config is not None:
+            return dict(cached_config)
 
-        return config_class(**new_config).model_dump()
+        with cls._decryption_cache_lock:
+            # Second check (double-checked locking) to prevent race conditions
+            cached_config = cls.decrypted_configs_cache.get(decrypted_config_key)
+            if cached_config is not None:
+                return dict(cached_config)
+
+            config_class, secret_keys, other_keys = (
+                provider_config_map[tracing_provider]["config_class"],
+                provider_config_map[tracing_provider]["secret_keys"],
+                provider_config_map[tracing_provider]["other_keys"],
+            )
+            new_config: dict[str, Any] = {}
+            keys_to_decrypt = [key for key in secret_keys if key in tracing_config]
+            if keys_to_decrypt:
+                decrypted_values = batch_decrypt_token(tenant_id, [tracing_config[key] for key in keys_to_decrypt])
+                new_config.update(zip(keys_to_decrypt, decrypted_values))
+
+            for key in other_keys:
+                new_config[key] = tracing_config.get(key, "")
+
+            decrypted_config = config_class(**new_config).model_dump()
+            cls.decrypted_configs_cache[decrypted_config_key] = decrypted_config
+            return dict(decrypted_config)
 
     @classmethod
     def obfuscated_decrypt_token(cls, tracing_provider: str, decrypt_tracing_config: dict):
@@ -219,7 +242,7 @@ class OpsTraceManager:
             provider_config_map[tracing_provider]["secret_keys"],
             provider_config_map[tracing_provider]["other_keys"],
         )
-        new_config = {}
+        new_config: dict[str, Any] = {}
         for key in secret_keys:
             if key in decrypt_tracing_config:
                 new_config[key] = obfuscated_token(decrypt_tracing_config[key])
@@ -596,6 +619,8 @@ class TraceTask:
             file_url = f"{self.file_base_url}/{message_file_data.url}" if message_file_data else ""
             file_list.append(file_url)
 
+        streaming_metrics = self._extract_streaming_metrics(message_data)
+
         metadata = {
             "conversation_id": message_data.conversation_id,
             "ls_provider": message_data.model_provider,
@@ -628,6 +653,9 @@ class TraceTask:
             metadata=metadata,
             message_file_data=message_file_data,
             conversation_mode=conversation_mode,
+            gen_ai_server_time_to_first_token=streaming_metrics.get("gen_ai_server_time_to_first_token"),
+            llm_streaming_time_to_generate=streaming_metrics.get("llm_streaming_time_to_generate"),
+            is_streaming_request=streaming_metrics.get("is_streaming_request", False),
         )
 
         return message_trace_info
@@ -853,6 +881,24 @@ class TraceTask:
 
         return generate_name_trace_info
 
+    def _extract_streaming_metrics(self, message_data) -> dict:
+        if not message_data.message_metadata:
+            return {}
+
+        try:
+            metadata = json.loads(message_data.message_metadata)
+            usage = metadata.get("usage", {})
+            time_to_first_token = usage.get("time_to_first_token")
+            time_to_generate = usage.get("time_to_generate")
+
+            return {
+                "gen_ai_server_time_to_first_token": time_to_first_token,
+                "llm_streaming_time_to_generate": time_to_generate,
+                "is_streaming_request": time_to_first_token is not None,
+            }
+        except (json.JSONDecodeError, AttributeError):
+            return {}
+
 
 trace_manager_timer: threading.Timer | None = None
 trace_manager_queue: queue.Queue = queue.Queue()

api/core/ops/tencent_trace/client.py

@@ -5,12 +5,18 @@ Tencent APM Trace Client - handles network operations, metrics, and API communic
 from __future__ import annotations
 
 import importlib
+import json
 import logging
 import os
 import socket
 from typing import TYPE_CHECKING
 from urllib.parse import urlparse
 
+try:
+    from importlib.metadata import version
+except ImportError:
+    from importlib_metadata import version  # type: ignore[import-not-found]
+
 if TYPE_CHECKING:
     from opentelemetry.metrics import Meter
     from opentelemetry.metrics._internal.instrument import Histogram
@@ -27,12 +33,27 @@ from opentelemetry.util.types import AttributeValue
 
 from configs import dify_config
 
-from .entities.tencent_semconv import LLM_OPERATION_DURATION
+from .entities.semconv import (
+    GEN_AI_SERVER_TIME_TO_FIRST_TOKEN,
+    GEN_AI_STREAMING_TIME_TO_GENERATE,
+    GEN_AI_TOKEN_USAGE,
+    GEN_AI_TRACE_DURATION,
+    LLM_OPERATION_DURATION,
+)
 from .entities.tencent_trace_entity import SpanData
 
 logger = logging.getLogger(__name__)
 
 
+def _get_opentelemetry_sdk_version() -> str:
+    """Get OpenTelemetry SDK version dynamically."""
+    try:
+        return version("opentelemetry-sdk")
+    except Exception:
+        logger.debug("Failed to get opentelemetry-sdk version, using default")
+        return "1.27.0"  # fallback version
+
+
 class TencentTraceClient:
     """Tencent APM trace client using OpenTelemetry OTLP exporter"""
 
@@ -57,6 +78,9 @@ class TencentTraceClient:
                 ResourceAttributes.SERVICE_VERSION: f"dify-{dify_config.project.version}-{dify_config.COMMIT_SHA}",
                 ResourceAttributes.DEPLOYMENT_ENVIRONMENT: f"{dify_config.DEPLOY_ENV}-{dify_config.EDITION}",
                 ResourceAttributes.HOST_NAME: socket.gethostname(),
+                ResourceAttributes.TELEMETRY_SDK_LANGUAGE: "python",
+                ResourceAttributes.TELEMETRY_SDK_NAME: "opentelemetry",
+                ResourceAttributes.TELEMETRY_SDK_VERSION: _get_opentelemetry_sdk_version(),
             }
         )
         # Prepare gRPC endpoint/metadata
@@ -80,18 +104,23 @@ class TencentTraceClient:
         )
         self.tracer_provider.add_span_processor(self.span_processor)
 
-        self.tracer = self.tracer_provider.get_tracer("dify.tencent_apm")
+        # use dify api version as tracer version
+        self.tracer = self.tracer_provider.get_tracer("dify-sdk", dify_config.project.version)
 
         # Store span contexts for parent-child relationships
         self.span_contexts: dict[int, trace_api.SpanContext] = {}
 
         self.meter: Meter | None = None
+        self.meter_provider: MeterProvider | None = None
         self.hist_llm_duration: Histogram | None = None
+        self.hist_token_usage: Histogram | None = None
+        self.hist_time_to_first_token: Histogram | None = None
+        self.hist_time_to_generate: Histogram | None = None
+        self.hist_trace_duration: Histogram | None = None
         self.metric_reader: MetricReader | None = None
 
         # Metrics exporter and instruments
         try:
-            from opentelemetry import metrics
             from opentelemetry.sdk.metrics import Histogram, MeterProvider
             from opentelemetry.sdk.metrics.export import AggregationTemporality, PeriodicExportingMetricReader
 
@@ -99,7 +128,7 @@ class TencentTraceClient:
             use_http_protobuf = protocol in {"http/protobuf", "http-protobuf"}
             use_http_json = protocol in {"http/json", "http-json"}
 
-            # Set preferred temporality for histograms to DELTA
+            # Tencent APM works best with delta aggregation temporality
             preferred_temporality: dict[type, AggregationTemporality] = {Histogram: AggregationTemporality.DELTA}
 
             def _create_metric_exporter(exporter_cls, **kwargs):
@@ -174,23 +203,66 @@ class TencentTraceClient:
                 )
 
             if metric_reader is not None:
+                # Use instance-level MeterProvider instead of global to support config changes
+                # without worker restart. Each TencentTraceClient manages its own MeterProvider.
                 provider = MeterProvider(resource=self.resource, metric_readers=[metric_reader])
-                metrics.set_meter_provider(provider)
-                self.meter = metrics.get_meter("dify-sdk", dify_config.project.version)
+                self.meter_provider = provider
+                self.meter = provider.get_meter("dify-sdk", dify_config.project.version)
+
+                # LLM operation duration histogram
                 self.hist_llm_duration = self.meter.create_histogram(
                     name=LLM_OPERATION_DURATION,
                     unit="s",
                     description="LLM operation duration (seconds)",
                 )
+
+                # Token usage histogram with exponential buckets
+                self.hist_token_usage = self.meter.create_histogram(
+                    name=GEN_AI_TOKEN_USAGE,
+                    unit="token",
+                    description="Number of tokens used in prompt and completions",
+                )
+
+                # Time to first token histogram
+                self.hist_time_to_first_token = self.meter.create_histogram(
+                    name=GEN_AI_SERVER_TIME_TO_FIRST_TOKEN,
+                    unit="s",
+                    description="Time to first token for streaming LLM responses (seconds)",
+                )
+
+                # Time to generate histogram
+                self.hist_time_to_generate = self.meter.create_histogram(
+                    name=GEN_AI_STREAMING_TIME_TO_GENERATE,
+                    unit="s",
+                    description="Total time to generate streaming LLM responses (seconds)",
+                )
+
+                # Trace duration histogram
+                self.hist_trace_duration = self.meter.create_histogram(
+                    name=GEN_AI_TRACE_DURATION,
+                    unit="s",
+                    description="End-to-end GenAI trace duration (seconds)",
+                )
+
                 self.metric_reader = metric_reader
             else:
                 self.meter = None
+                self.meter_provider = None
                 self.hist_llm_duration = None
+                self.hist_token_usage = None
+                self.hist_time_to_first_token = None
+                self.hist_time_to_generate = None
+                self.hist_trace_duration = None
                 self.metric_reader = None
         except Exception:
             logger.exception("[Tencent APM] Metrics initialization failed; metrics disabled")
             self.meter = None
+            self.meter_provider = None
             self.hist_llm_duration = None
+            self.hist_token_usage = None
+            self.hist_time_to_first_token = None
+            self.hist_time_to_generate = None
+            self.hist_trace_duration = None
             self.metric_reader = None
 
     def add_span(self, span_data: SpanData) -> None:
@@ -212,10 +284,158 @@ class TencentTraceClient:
             if attributes:
                 for k, v in attributes.items():
                     attrs[k] = str(v) if not isinstance(v, (str, int, float, bool)) else v  # type: ignore[assignment]
+
+            logger.info(
+                "[Tencent Metrics] Metric: %s | Value: %.4f | Attributes: %s",
+                LLM_OPERATION_DURATION,
+                latency_seconds,
+                json.dumps(attrs, ensure_ascii=False),
+            )
+
             self.hist_llm_duration.record(latency_seconds, attrs)  # type: ignore[attr-defined]
         except Exception:
             logger.debug("[Tencent APM] Failed to record LLM duration", exc_info=True)
 
+    def record_token_usage(
+        self,
+        token_count: int,
+        token_type: str,
+        operation_name: str,
+        request_model: str,
+        response_model: str,
+        server_address: str,
+        provider: str,
+    ) -> None:
+        """Record token usage histogram.
+
+        Args:
+            token_count: Number of tokens used
+            token_type: "input" or "output"
+            operation_name: Operation name (e.g., "chat")
+            request_model: Model used in request
+            response_model: Model used in response
+            server_address: Server address
+            provider: Model provider name
+        """
+        try:
+            if not hasattr(self, "hist_token_usage") or self.hist_token_usage is None:
+                return
+
+            attributes = {
+                "gen_ai.operation.name": operation_name,
+                "gen_ai.request.model": request_model,
+                "gen_ai.response.model": response_model,
+                "gen_ai.system": provider,
+                "gen_ai.token.type": token_type,
+                "server.address": server_address,
+            }
+
+            logger.info(
+                "[Tencent Metrics] Metric: %s | Value: %d | Attributes: %s",
+                GEN_AI_TOKEN_USAGE,
+                token_count,
+                json.dumps(attributes, ensure_ascii=False),
+            )
+
+            self.hist_token_usage.record(token_count, attributes)  # type: ignore[attr-defined]
+        except Exception:
+            logger.debug("[Tencent APM] Failed to record token usage", exc_info=True)
+
+    def record_time_to_first_token(
+        self, ttft_seconds: float, provider: str, model: str, operation_name: str = "chat"
+    ) -> None:
+        """Record time to first token histogram.
+
+        Args:
+            ttft_seconds: Time to first token in seconds
+            provider: Model provider name
+            model: Model name
+            operation_name: Operation name (default: "chat")
+        """
+        try:
+            if not hasattr(self, "hist_time_to_first_token") or self.hist_time_to_first_token is None:
+                return
+
+            attributes = {
+                "gen_ai.operation.name": operation_name,
+                "gen_ai.system": provider,
+                "gen_ai.request.model": model,
+                "gen_ai.response.model": model,
+                "stream": "true",
+            }
+
+            logger.info(
+                "[Tencent Metrics] Metric: %s | Value: %.4f | Attributes: %s",
+                GEN_AI_SERVER_TIME_TO_FIRST_TOKEN,
+                ttft_seconds,
+                json.dumps(attributes, ensure_ascii=False),
+            )
+
+            self.hist_time_to_first_token.record(ttft_seconds, attributes)  # type: ignore[attr-defined]
+        except Exception:
+            logger.debug("[Tencent APM] Failed to record time to first token", exc_info=True)
+
+    def record_time_to_generate(
+        self, ttg_seconds: float, provider: str, model: str, operation_name: str = "chat"
+    ) -> None:
+        """Record time to generate histogram.
+
+        Args:
+            ttg_seconds: Time to generate in seconds
+            provider: Model provider name
+            model: Model name
+            operation_name: Operation name (default: "chat")
+        """
+        try:
+            if not hasattr(self, "hist_time_to_generate") or self.hist_time_to_generate is None:
+                return
+
+            attributes = {
+                "gen_ai.operation.name": operation_name,
+                "gen_ai.system": provider,
+                "gen_ai.request.model": model,
+                "gen_ai.response.model": model,
+                "stream": "true",
+            }
+
+            logger.info(
+                "[Tencent Metrics] Metric: %s | Value: %.4f | Attributes: %s",
+                GEN_AI_STREAMING_TIME_TO_GENERATE,
+                ttg_seconds,
+                json.dumps(attributes, ensure_ascii=False),
+            )
+
+            self.hist_time_to_generate.record(ttg_seconds, attributes)  # type: ignore[attr-defined]
+        except Exception:
+            logger.debug("[Tencent APM] Failed to record time to generate", exc_info=True)
+
+    def record_trace_duration(self, duration_seconds: float, attributes: dict[str, str] | None = None) -> None:
+        """Record end-to-end trace duration histogram in seconds.
+
+        Args:
+            duration_seconds: Trace duration in seconds
+            attributes: Optional attributes (e.g., conversation_mode, app_id)
+        """
+        try:
+            if not hasattr(self, "hist_trace_duration") or self.hist_trace_duration is None:
+                return
+
+            attrs: dict[str, str] = {}
+            if attributes:
+                for k, v in attributes.items():
+                    attrs[k] = str(v) if not isinstance(v, (str, int, float, bool)) else v  # type: ignore[assignment]
+
+            logger.info(
+                "[Tencent Metrics] Metric: %s | Value: %.4f | Attributes: %s",
+                GEN_AI_TRACE_DURATION,
+                duration_seconds,
+                json.dumps(attrs, ensure_ascii=False),
+            )
+
+            self.hist_trace_duration.record(duration_seconds, attrs)  # type: ignore[attr-defined]
+        except Exception:
+            logger.debug("[Tencent APM] Failed to record trace duration", exc_info=True)
+
     def _create_and_export_span(self, span_data: SpanData) -> None:
         """Create span using OpenTelemetry Tracer API"""
         try:
@@ -296,11 +516,19 @@ class TencentTraceClient:
 
             if self.tracer_provider:
                 self.tracer_provider.shutdown()
+
+            # Shutdown instance-level meter provider
+            if self.meter_provider is not None:
+                try:
+                    self.meter_provider.shutdown()  # type: ignore[attr-defined]
+                except Exception:
+                    logger.debug("[Tencent APM] Error shutting down meter provider", exc_info=True)
+
             if self.metric_reader is not None:
                 try:
                     self.metric_reader.shutdown()  # type: ignore[attr-defined]
                 except Exception:
-                    pass
+                    logger.debug("[Tencent APM] Error shutting down metric reader", exc_info=True)
 
         except Exception:
             logger.exception("[Tencent APM] Error during client shutdown")

api/core/ops/tencent_trace/entities/semconv.py

@@ -47,6 +47,9 @@ GEN_AI_COMPLETION = "gen_ai.completion"
 
 GEN_AI_RESPONSE_FINISH_REASON = "gen_ai.response.finish_reason"
 
+# Streaming Span Attributes
+GEN_AI_IS_STREAMING_REQUEST = "llm.is_streaming"  # Same as OpenLLMetry semconv
+
 # Tool
 TOOL_NAME = "tool.name"
 
@@ -62,6 +65,19 @@ INSTRUMENTATION_LANGUAGE = "python"
 
 # Metrics
 LLM_OPERATION_DURATION = "gen_ai.client.operation.duration"
+GEN_AI_TOKEN_USAGE = "gen_ai.client.token.usage"
+GEN_AI_SERVER_TIME_TO_FIRST_TOKEN = "gen_ai.server.time_to_first_token"
+GEN_AI_STREAMING_TIME_TO_GENERATE = "gen_ai.streaming.time_to_generate"
+# The LLM trace duration which is exclusive to tencent apm
+GEN_AI_TRACE_DURATION = "gen_ai.trace.duration"
+
+# Token Usage Attributes
+GEN_AI_OPERATION_NAME = "gen_ai.operation.name"
+GEN_AI_REQUEST_MODEL = "gen_ai.request.model"
+GEN_AI_RESPONSE_MODEL = "gen_ai.response.model"
+GEN_AI_SYSTEM = "gen_ai.system"
+GEN_AI_TOKEN_TYPE = "gen_ai.token.type"
+SERVER_ADDRESS = "server.address"
 
 
 class GenAISpanKind(Enum):

api/core/ops/tencent_trace/span_builder.py

@@ -14,10 +14,11 @@ from core.ops.entities.trace_entity import (
     ToolTraceInfo,
     WorkflowTraceInfo,
 )
-from core.ops.tencent_trace.entities.tencent_semconv import (
+from core.ops.tencent_trace.entities.semconv import (
     GEN_AI_COMPLETION,
     GEN_AI_FRAMEWORK,
     GEN_AI_IS_ENTRY,
+    GEN_AI_IS_STREAMING_REQUEST,
     GEN_AI_MODEL_NAME,
     GEN_AI_PROMPT,
     GEN_AI_PROVIDER,
@@ -156,6 +157,25 @@ class TencentSpanBuilder:
         outputs = node_execution.outputs or {}
         usage_data = process_data.get("usage", {}) if "usage" in process_data else outputs.get("usage", {})
 
+        attributes = {
+            GEN_AI_SESSION_ID: trace_info.metadata.get("conversation_id", ""),
+            GEN_AI_SPAN_KIND: GenAISpanKind.GENERATION.value,
+            GEN_AI_FRAMEWORK: "dify",
+            GEN_AI_MODEL_NAME: process_data.get("model_name", ""),
+            GEN_AI_PROVIDER: process_data.get("model_provider", ""),
+            GEN_AI_USAGE_INPUT_TOKENS: str(usage_data.get("prompt_tokens", 0)),
+            GEN_AI_USAGE_OUTPUT_TOKENS: str(usage_data.get("completion_tokens", 0)),
+            GEN_AI_USAGE_TOTAL_TOKENS: str(usage_data.get("total_tokens", 0)),
+            GEN_AI_PROMPT: json.dumps(process_data.get("prompts", []), ensure_ascii=False),
+            GEN_AI_COMPLETION: str(outputs.get("text", "")),
+            GEN_AI_RESPONSE_FINISH_REASON: outputs.get("finish_reason", ""),
+            INPUT_VALUE: json.dumps(process_data.get("prompts", []), ensure_ascii=False),
+            OUTPUT_VALUE: str(outputs.get("text", "")),
+        }
+
+        if usage_data.get("time_to_first_token") is not None:
+            attributes[GEN_AI_IS_STREAMING_REQUEST] = "true"
+
         return SpanData(
             trace_id=trace_id,
             parent_span_id=workflow_span_id,
@@ -163,21 +183,7 @@ class TencentSpanBuilder:
             name="GENERATION",
             start_time=TencentSpanBuilder._get_time_nanoseconds(node_execution.created_at),
             end_time=TencentSpanBuilder._get_time_nanoseconds(node_execution.finished_at),
-            attributes={
-                GEN_AI_SESSION_ID: trace_info.metadata.get("conversation_id", ""),
-                GEN_AI_SPAN_KIND: GenAISpanKind.GENERATION.value,
-                GEN_AI_FRAMEWORK: "dify",
-                GEN_AI_MODEL_NAME: process_data.get("model_name", ""),
-                GEN_AI_PROVIDER: process_data.get("model_provider", ""),
-                GEN_AI_USAGE_INPUT_TOKENS: str(usage_data.get("prompt_tokens", 0)),
-                GEN_AI_USAGE_OUTPUT_TOKENS: str(usage_data.get("completion_tokens", 0)),
-                GEN_AI_USAGE_TOTAL_TOKENS: str(usage_data.get("total_tokens", 0)),
-                GEN_AI_PROMPT: json.dumps(process_data.get("prompts", []), ensure_ascii=False),
-                GEN_AI_COMPLETION: str(outputs.get("text", "")),
-                GEN_AI_RESPONSE_FINISH_REASON: outputs.get("finish_reason", ""),
-                INPUT_VALUE: json.dumps(process_data.get("prompts", []), ensure_ascii=False),
-                OUTPUT_VALUE: str(outputs.get("text", "")),
-            },
+            attributes=attributes,
             status=TencentSpanBuilder._get_workflow_node_status(node_execution),
         )
 
@@ -191,6 +197,19 @@ class TencentSpanBuilder:
         if trace_info.error:
             status = Status(StatusCode.ERROR, trace_info.error)
 
+        attributes = {
+            GEN_AI_SESSION_ID: trace_info.metadata.get("conversation_id", ""),
+            GEN_AI_USER_ID: str(user_id),
+            GEN_AI_SPAN_KIND: GenAISpanKind.WORKFLOW.value,
+            GEN_AI_FRAMEWORK: "dify",
+            GEN_AI_IS_ENTRY: "true",
+            INPUT_VALUE: str(trace_info.inputs or ""),
+            OUTPUT_VALUE: str(trace_info.outputs or ""),
+        }
+
+        if trace_info.is_streaming_request:
+            attributes[GEN_AI_IS_STREAMING_REQUEST] = "true"
+
         return SpanData(
             trace_id=trace_id,
             parent_span_id=None,
@@ -198,15 +217,7 @@ class TencentSpanBuilder:
             name="message",
             start_time=TencentSpanBuilder._get_time_nanoseconds(trace_info.start_time),
             end_time=TencentSpanBuilder._get_time_nanoseconds(trace_info.end_time),
-            attributes={
-                GEN_AI_SESSION_ID: trace_info.metadata.get("conversation_id", ""),
-                GEN_AI_USER_ID: str(user_id),
-                GEN_AI_SPAN_KIND: GenAISpanKind.WORKFLOW.value,
-                GEN_AI_FRAMEWORK: "dify",
-                GEN_AI_IS_ENTRY: "true",
-                INPUT_VALUE: str(trace_info.inputs or ""),
-                OUTPUT_VALUE: str(trace_info.outputs or ""),
-            },
+            attributes=attributes,
             status=status,
             links=links,
         )

api/core/ops/tencent_trace/tencent_trace.py

@@ -90,6 +90,9 @@ class TencentDataTrace(BaseTraceInstance):
 
             self._process_workflow_nodes(trace_info, trace_id)
 
+            # Record trace duration for entry span
+            self._record_workflow_trace_duration(trace_info)
+
         except Exception:
             logger.exception("[Tencent APM] Failed to process workflow trace")
 
@@ -107,6 +110,11 @@ class TencentDataTrace(BaseTraceInstance):
 
             self.trace_client.add_span(message_span)
 
+            self._record_message_llm_metrics(trace_info)
+
+            # Record trace duration for entry span
+            self._record_message_trace_duration(trace_info)
+
         except Exception:
             logger.exception("[Tencent APM] Failed to process message trace")
 
@@ -290,24 +298,219 @@ class TencentDataTrace(BaseTraceInstance):
     def _record_llm_metrics(self, node_execution: WorkflowNodeExecution) -> None:
         """Record LLM performance metrics"""
         try:
-            if not hasattr(self.trace_client, "record_llm_duration"):
-                return
-
             process_data = node_execution.process_data or {}
-            usage = process_data.get("usage", {})
-            latency_s = float(usage.get("latency", 0.0))
+            outputs = node_execution.outputs or {}
+            usage = process_data.get("usage", {}) if "usage" in process_data else outputs.get("usage", {})
 
-            if latency_s > 0:
-                attributes = {
-                    "provider": process_data.get("model_provider", ""),
-                    "model": process_data.get("model_name", ""),
-                    "span_kind": "GENERATION",
-                }
-                self.trace_client.record_llm_duration(latency_s, attributes)
+            model_provider = process_data.get("model_provider", "unknown")
+            model_name = process_data.get("model_name", "unknown")
+            model_mode = process_data.get("model_mode", "chat")
+
+            # Record LLM duration
+            if hasattr(self.trace_client, "record_llm_duration"):
+                latency_s = float(usage.get("latency", 0.0))
+
+                if latency_s > 0:
+                    # Determine if streaming from usage metrics
+                    is_streaming = usage.get("time_to_first_token") is not None
+
+                    attributes = {
+                        "gen_ai.system": model_provider,
+                        "gen_ai.response.model": model_name,
+                        "gen_ai.operation.name": model_mode,
+                        "stream": "true" if is_streaming else "false",
+                    }
+                    self.trace_client.record_llm_duration(latency_s, attributes)
+
+            # Record streaming metrics from usage
+            time_to_first_token = usage.get("time_to_first_token")
+            if time_to_first_token is not None and hasattr(self.trace_client, "record_time_to_first_token"):
+                ttft_seconds = float(time_to_first_token)
+                if ttft_seconds > 0:
+                    self.trace_client.record_time_to_first_token(
+                        ttft_seconds=ttft_seconds, provider=model_provider, model=model_name, operation_name=model_mode
+                    )
+
+            time_to_generate = usage.get("time_to_generate")
+            if time_to_generate is not None and hasattr(self.trace_client, "record_time_to_generate"):
+                ttg_seconds = float(time_to_generate)
+                if ttg_seconds > 0:
+                    self.trace_client.record_time_to_generate(
+                        ttg_seconds=ttg_seconds, provider=model_provider, model=model_name, operation_name=model_mode
+                    )
+
+            # Record token usage
+            if hasattr(self.trace_client, "record_token_usage"):
+                # Extract token counts
+                input_tokens = int(usage.get("prompt_tokens", 0))
+                output_tokens = int(usage.get("completion_tokens", 0))
+
+                if input_tokens > 0 or output_tokens > 0:
+                    server_address = f"{model_provider}"
+
+                    # Record input tokens
+                    if input_tokens > 0:
+                        self.trace_client.record_token_usage(
+                            token_count=input_tokens,
+                            token_type="input",
+                            operation_name=model_mode,
+                            request_model=model_name,
+                            response_model=model_name,
+                            server_address=server_address,
+                            provider=model_provider,
+                        )
+
+                    # Record output tokens
+                    if output_tokens > 0:
+                        self.trace_client.record_token_usage(
+                            token_count=output_tokens,
+                            token_type="output",
+                            operation_name=model_mode,
+                            request_model=model_name,
+                            response_model=model_name,
+                            server_address=server_address,
+                            provider=model_provider,
+                        )
 
         except Exception:
             logger.debug("[Tencent APM] Failed to record LLM metrics")
 
+    def _record_message_llm_metrics(self, trace_info: MessageTraceInfo) -> None:
+        """Record LLM metrics for message traces"""
+        try:
+            trace_metadata = trace_info.metadata or {}
+            message_data = trace_info.message_data or {}
+            provider_latency = 0.0
+            if isinstance(message_data, dict):
+                provider_latency = float(message_data.get("provider_response_latency", 0.0) or 0.0)
+            else:
+                provider_latency = float(getattr(message_data, "provider_response_latency", 0.0) or 0.0)
+
+            model_provider = trace_metadata.get("ls_provider") or (
+                message_data.get("model_provider", "") if isinstance(message_data, dict) else ""
+            )
+            model_name = trace_metadata.get("ls_model_name") or (
+                message_data.get("model_id", "") if isinstance(message_data, dict) else ""
+            )
+
+            # Record LLM duration
+            if provider_latency > 0 and hasattr(self.trace_client, "record_llm_duration"):
+                is_streaming = trace_info.is_streaming_request
+
+                duration_attributes = {
+                    "gen_ai.system": model_provider,
+                    "gen_ai.response.model": model_name,
+                    "gen_ai.operation.name": "chat",  # Message traces are always chat
+                    "stream": "true" if is_streaming else "false",
+                }
+                self.trace_client.record_llm_duration(provider_latency, duration_attributes)
+
+            # Record streaming metrics for message traces
+            if trace_info.is_streaming_request:
+                # Record time to first token
+                if trace_info.gen_ai_server_time_to_first_token is not None and hasattr(
+                    self.trace_client, "record_time_to_first_token"
+                ):
+                    ttft_seconds = float(trace_info.gen_ai_server_time_to_first_token)
+                    if ttft_seconds > 0:
+                        self.trace_client.record_time_to_first_token(
+                            ttft_seconds=ttft_seconds, provider=str(model_provider or ""), model=str(model_name or "")
+                        )
+
+                # Record time to generate
+                if trace_info.llm_streaming_time_to_generate is not None and hasattr(
+                    self.trace_client, "record_time_to_generate"
+                ):
+                    ttg_seconds = float(trace_info.llm_streaming_time_to_generate)
+                    if ttg_seconds > 0:
+                        self.trace_client.record_time_to_generate(
+                            ttg_seconds=ttg_seconds, provider=str(model_provider or ""), model=str(model_name or "")
+                        )
+
+            # Record token usage
+            if hasattr(self.trace_client, "record_token_usage"):
+                input_tokens = int(trace_info.message_tokens or 0)
+                output_tokens = int(trace_info.answer_tokens or 0)
+
+                if input_tokens > 0:
+                    self.trace_client.record_token_usage(
+                        token_count=input_tokens,
+                        token_type="input",
+                        operation_name="chat",
+                        request_model=str(model_name or ""),
+                        response_model=str(model_name or ""),
+                        server_address=str(model_provider or ""),
+                        provider=str(model_provider or ""),
+                    )
+
+                if output_tokens > 0:
+                    self.trace_client.record_token_usage(
+                        token_count=output_tokens,
+                        token_type="output",
+                        operation_name="chat",
+                        request_model=str(model_name or ""),
+                        response_model=str(model_name or ""),
+                        server_address=str(model_provider or ""),
+                        provider=str(model_provider or ""),
+                    )
+
+        except Exception:
+            logger.debug("[Tencent APM] Failed to record message LLM metrics")
+
+    def _record_workflow_trace_duration(self, trace_info: WorkflowTraceInfo) -> None:
+        """Record end-to-end workflow trace duration."""
+        try:
+            if not hasattr(self.trace_client, "record_trace_duration"):
+                return
+
+            # Calculate duration from start_time and end_time to match span duration
+            if trace_info.start_time and trace_info.end_time:
+                duration_s = (trace_info.end_time - trace_info.start_time).total_seconds()
+            else:
+                # Fallback to workflow_run_elapsed_time if timestamps not available
+                duration_s = float(trace_info.workflow_run_elapsed_time)
+
+            if duration_s > 0:
+                attributes = {
+                    "conversation_mode": "workflow",
+                    "workflow_status": trace_info.workflow_run_status,
+                }
+
+                # Add conversation_id if available
+                if trace_info.conversation_id:
+                    attributes["has_conversation"] = "true"
+                else:
+                    attributes["has_conversation"] = "false"
+
+                self.trace_client.record_trace_duration(duration_s, attributes)
+
+        except Exception:
+            logger.debug("[Tencent APM] Failed to record workflow trace duration")
+
+    def _record_message_trace_duration(self, trace_info: MessageTraceInfo) -> None:
+        """Record end-to-end message trace duration."""
+        try:
+            if not hasattr(self.trace_client, "record_trace_duration"):
+                return
+
+            # Calculate duration from start_time and end_time
+            if trace_info.start_time and trace_info.end_time:
+                duration = (trace_info.end_time - trace_info.start_time).total_seconds()
+
+                if duration > 0:
+                    attributes = {
+                        "conversation_mode": trace_info.conversation_mode,
+                    }
+
+                    # Add streaming flag if available
+                    if hasattr(trace_info, "is_streaming_request"):
+                        attributes["stream"] = "true" if trace_info.is_streaming_request else "false"
+
+                    self.trace_client.record_trace_duration(duration, attributes)
+
+        except Exception:
+            logger.debug("[Tencent APM] Failed to record message trace duration")
+
     def __del__(self):
         """Ensure proper cleanup on garbage collection."""
         try:

api/core/prompt/entities/advanced_prompt_entities.py

@@ -45,6 +45,12 @@ class MemoryConfig(BaseModel):
         enabled: bool
         size: int | None = None
 
+    mode: Literal["linear", "block"] | None = "linear"
+    block_id: list[str] | None = None
     role_prefix: RolePrefix | None = None
     window: WindowConfig
     query_prompt_template: str | None = None
+
+    @property
+    def is_block_mode(self) -> bool:
+        return self.mode == "block" and bool(self.block_id)

api/core/provider_manager.py

@@ -618,18 +618,18 @@ class ProviderManager:
                 )
 
             for quota in configuration.quotas:
-                if quota.quota_type == ProviderQuotaType.TRIAL:
+                if quota.quota_type in (ProviderQuotaType.TRIAL, ProviderQuotaType.PAID):
                     # Init trial provider records if not exists
-                    if ProviderQuotaType.TRIAL not in provider_quota_to_provider_record_dict:
+                    if quota.quota_type not in provider_quota_to_provider_record_dict:
                         try:
                             # FIXME ignore the type error, only TrialHostingQuota has limit need to change the logic
                             new_provider_record = Provider(
                                 tenant_id=tenant_id,
                                 # TODO: Use provider name with prefix after the data migration.
                                 provider_name=ModelProviderID(provider_name).provider_name,
-                                provider_type=ProviderType.SYSTEM,
-                                quota_type=ProviderQuotaType.TRIAL,
-                                quota_limit=quota.quota_limit,  # type: ignore
+                                provider_type=ProviderType.SYSTEM.value,
+                                quota_type=quota.quota_type,
+                                quota_limit=0,  # type: ignore
                                 quota_used=0,
                                 is_valid=True,
                             )
@@ -641,8 +641,8 @@ class ProviderManager:
                             stmt = select(Provider).where(
                                 Provider.tenant_id == tenant_id,
                                 Provider.provider_name == ModelProviderID(provider_name).provider_name,
-                                Provider.provider_type == ProviderType.SYSTEM,
-                                Provider.quota_type == ProviderQuotaType.TRIAL,
+                                Provider.provider_type == ProviderType.SYSTEM.value,
+                                Provider.quota_type == quota.quota_type,
                             )
                             existed_provider_record = db.session.scalar(stmt)
                             if not existed_provider_record:
@@ -652,7 +652,7 @@ class ProviderManager:
                                 existed_provider_record.is_valid = True
                                 db.session.commit()
 
-                            provider_name_to_provider_records_dict[provider_name].append(existed_provider_record)
+                            provider_name_to_provider_records_dict[provider_name].append(existed_provider_record)                        
 
         return provider_name_to_provider_records_dict
 
@@ -912,6 +912,22 @@ class ProviderManager:
                 provider_record
             )
         quota_configurations = []
+
+        if dify_config.EDITION == "CLOUD":
+            from services.credit_pool_service import CreditPoolService
+
+            trail_pool = CreditPoolService.get_pool(
+                tenant_id=tenant_id,
+                pool_type=ProviderQuotaType.TRIAL.value,
+            )
+            paid_pool = CreditPoolService.get_pool(
+                tenant_id=tenant_id,
+                pool_type=ProviderQuotaType.PAID.value,
+            )
+        else:
+            trail_pool = None
+            paid_pool = None
+
         for provider_quota in provider_hosting_configuration.quotas:
             if provider_quota.quota_type not in quota_type_to_provider_records_dict:
                 if provider_quota.quota_type == ProviderQuotaType.FREE:
@@ -932,16 +948,36 @@ class ProviderManager:
                     raise ValueError("quota_used is None")
                 if provider_record.quota_limit is None:
                     raise ValueError("quota_limit is None")
+                if provider_quota.quota_type == ProviderQuotaType.TRIAL and trail_pool is not None:
+                    quota_configuration = QuotaConfiguration(
+                        quota_type=provider_quota.quota_type,
+                        quota_unit=provider_hosting_configuration.quota_unit or QuotaUnit.TOKENS,
+                        quota_used=trail_pool.quota_used,
+                        quota_limit=trail_pool.quota_limit,
+                        is_valid=trail_pool.quota_limit > trail_pool.quota_used or trail_pool.quota_limit == -1,
+                        restrict_models=provider_quota.restrict_models,
+                    )
 
-                quota_configuration = QuotaConfiguration(
-                    quota_type=provider_quota.quota_type,
-                    quota_unit=provider_hosting_configuration.quota_unit or QuotaUnit.TOKENS,
-                    quota_used=provider_record.quota_used,
-                    quota_limit=provider_record.quota_limit,
-                    is_valid=provider_record.quota_limit > provider_record.quota_used
-                    or provider_record.quota_limit == -1,
-                    restrict_models=provider_quota.restrict_models,
-                )
+                elif provider_quota.quota_type == ProviderQuotaType.PAID and paid_pool is not None:
+                    quota_configuration = QuotaConfiguration(
+                        quota_type=provider_quota.quota_type,
+                        quota_unit=provider_hosting_configuration.quota_unit or QuotaUnit.TOKENS,
+                        quota_used=paid_pool.quota_used,
+                        quota_limit=paid_pool.quota_limit,
+                        is_valid=paid_pool.quota_limit > paid_pool.quota_used or paid_pool.quota_limit == -1,
+                        restrict_models=provider_quota.restrict_models,
+                    )
+
+                else:
+                    quota_configuration = QuotaConfiguration(
+                        quota_type=provider_quota.quota_type,
+                        quota_unit=provider_hosting_configuration.quota_unit or QuotaUnit.TOKENS,
+                        quota_used=provider_record.quota_used,
+                        quota_limit=provider_record.quota_limit,
+                        is_valid=provider_record.quota_limit > provider_record.quota_used
+                        or provider_record.quota_limit == -1,
+                        restrict_models=provider_quota.restrict_models,
+                    )
 
             quota_configurations.append(quota_configuration)
 

api/core/rag/datasource/vdb/oracle/oraclevector.py

@@ -289,7 +289,8 @@ class OracleVector(BaseVector):
                 words = pseg.cut(query)
                 current_entity = ""
                 for word, pos in words:
-                    if pos in {"nr", "Ng", "eng", "nz", "n", "ORG", "v"}:  # nr: 人名，ns: 地名，nt: 机构名
+                    # nr: person name, ns: place name, nt: organization name
+                    if pos in {"nr", "Ng", "eng", "nz", "n", "ORG", "v"}:
                         current_entity += word
                     else:
                         if current_entity:

api/core/rag/datasource/vdb/pyvastbase/vastbase_vector.py

@@ -213,7 +213,7 @@ class VastbaseVector(BaseVector):
 
             with self._get_cursor() as cur:
                 cur.execute(SQL_CREATE_TABLE.format(table_name=self.table_name, dimension=dimension))
-                # Vastbase 支持的向量维度取值范围为 [1,16000]
+                # Vastbase supports vector dimensions in range [1, 16000]
                 if dimension <= 16000:
                     cur.execute(SQL_CREATE_INDEX.format(table_name=self.table_name))
             redis_client.set(collection_exist_cache_key, 1, ex=3600)

api/core/rag/datasource/vdb/weaviate/weaviate_vector.py

@@ -39,11 +39,13 @@ class WeaviateConfig(BaseModel):
 
     Attributes:
         endpoint: Weaviate server endpoint URL
+        grpc_endpoint: Optional Weaviate gRPC server endpoint URL
         api_key: Optional API key for authentication
         batch_size: Number of objects to batch per insert operation
     """
 
     endpoint: str
+    grpc_endpoint: str | None = None
     api_key: str | None = None
     batch_size: int = 100
 
@@ -88,9 +90,22 @@ class WeaviateVector(BaseVector):
         http_secure = p.scheme == "https"
         http_port = p.port or (443 if http_secure else 80)
 
-        grpc_host = host
-        grpc_secure = http_secure
-        grpc_port = 443 if grpc_secure else 50051
+        # Parse gRPC configuration
+        if config.grpc_endpoint:
+            # Urls without scheme won't be parsed correctly in some python verions,
+            # see https://bugs.python.org/issue27657
+            grpc_endpoint_with_scheme = (
+                config.grpc_endpoint if "://" in config.grpc_endpoint else f"grpc://{config.grpc_endpoint}"
+            )
+            grpc_p = urlparse(grpc_endpoint_with_scheme)
+            grpc_host = grpc_p.hostname or "localhost"
+            grpc_port = grpc_p.port or (443 if grpc_p.scheme == "grpcs" else 50051)
+            grpc_secure = grpc_p.scheme == "grpcs"
+        else:
+            # Infer from HTTP endpoint as fallback
+            grpc_host = host
+            grpc_secure = http_secure
+            grpc_port = 443 if grpc_secure else 50051
 
         client = weaviate.connect_to_custom(
             http_host=host,
@@ -100,6 +115,7 @@ class WeaviateVector(BaseVector):
             grpc_port=grpc_port,
             grpc_secure=grpc_secure,
             auth_credentials=Auth.api_key(config.api_key) if config.api_key else None,
+            skip_init_checks=True,  # Skip PyPI version check to avoid unnecessary HTTP requests
         )
 
         if not client.is_ready():
@@ -431,6 +447,7 @@ class WeaviateVectorFactory(AbstractVectorFactory):
             collection_name=collection_name,
             config=WeaviateConfig(
                 endpoint=dify_config.WEAVIATE_ENDPOINT or "",
+                grpc_endpoint=dify_config.WEAVIATE_GRPC_ENDPOINT or "",
                 api_key=dify_config.WEAVIATE_API_KEY,
                 batch_size=dify_config.WEAVIATE_BATCH_SIZE,
             ),

api/core/rag/pipeline/queue.py

@@ -0,0 +1,89 @@
+import json
+from dataclasses import dataclass
+from typing import Any
+
+from extensions.ext_redis import redis_client
+
+TASK_WRAPPER_PREFIX = "__WRAPPER__:"
+
+
+@dataclass
+class TaskWrapper:
+    data: Any
+    
+    def serialize(self) -> str:
+        return json.dumps(self.data, ensure_ascii=False)
+    
+    @classmethod
+    def deserialize(cls, serialized_data: str) -> 'TaskWrapper':
+        data = json.loads(serialized_data)
+        return cls(data)
+
+
+class TenantSelfTaskQueue:
+    """
+    Simple queue for tenant self tasks, used for tenant self task isolation.
+    It uses Redis list to store tasks, and Redis key to store task waiting flag.
+    Support tasks that can be serialized by json.
+    """
+    DEFAULT_TASK_TTL = 60 * 60
+
+    def __init__(self, tenant_id: str, unique_key: str):
+        self.tenant_id = tenant_id
+        self.unique_key = unique_key
+        self.queue = f"tenant_self_{unique_key}_task_queue:{tenant_id}"
+        self.task_key = f"tenant_{unique_key}_task:{tenant_id}"
+
+    def get_task_key(self):
+        return redis_client.get(self.task_key)
+
+    def set_task_waiting_time(self, ttl: int | None = None):
+        ttl = ttl or self.DEFAULT_TASK_TTL
+        redis_client.setex(self.task_key, ttl, 1)
+
+    def delete_task_key(self):
+        redis_client.delete(self.task_key)
+
+    def push_tasks(self, tasks: list):
+        serialized_tasks = []
+        for task in tasks:
+            # Store str list directly, maintaining full compatibility for pipeline scenarios
+            if isinstance(task, str):
+                serialized_tasks.append(task)
+            else:
+                # Use TaskWrapper to do JSON serialization, add prefix for identification
+                wrapper = TaskWrapper(task)
+                serialized_data = wrapper.serialize()
+                serialized_tasks.append(f"{TASK_WRAPPER_PREFIX}{serialized_data}")
+        
+        redis_client.lpush(self.queue, *serialized_tasks)
+    
+    def pull_tasks(self, count: int = 1) -> list:
+        if count <= 0:
+            return []
+        
+        tasks = []
+        for _ in range(count):
+            serialized_task = redis_client.rpop(self.queue)
+            if not serialized_task:
+                break
+            
+            if isinstance(serialized_task, bytes):
+                serialized_task = serialized_task.decode('utf-8')
+            
+            # Check if use TaskWrapper or not
+            if serialized_task.startswith(TASK_WRAPPER_PREFIX):
+                try:
+                    wrapper_data = serialized_task[len(TASK_WRAPPER_PREFIX):]
+                    wrapper = TaskWrapper.deserialize(wrapper_data)
+                    tasks.append(wrapper.data)
+                except (json.JSONDecodeError, TypeError, ValueError):
+                    tasks.append(serialized_task)
+            else:
+                tasks.append(serialized_task)
+        
+        return tasks
+
+    def get_next_task(self):
+        tasks = self.pull_tasks(1)
+        return tasks[0] if tasks else None

api/core/tools/__base/tool.py

@@ -210,12 +210,13 @@ class Tool(ABC):
             meta=meta,
         )
 
-    def create_json_message(self, object: dict) -> ToolInvokeMessage:
+    def create_json_message(self, object: dict, suppress_output: bool = False) -> ToolInvokeMessage:
         """
         create a json message
         """
         return ToolInvokeMessage(
-            type=ToolInvokeMessage.MessageType.JSON, message=ToolInvokeMessage.JsonMessage(json_object=object)
+            type=ToolInvokeMessage.MessageType.JSON,
+            message=ToolInvokeMessage.JsonMessage(json_object=object, suppress_output=suppress_output),
         )
 
     def create_variable_message(

api/core/tools/entities/tool_entities.py

@@ -129,6 +129,7 @@ class ToolInvokeMessage(BaseModel):
 
     class JsonMessage(BaseModel):
         json_object: dict
+        suppress_output: bool = Field(default=False, description="Whether to suppress JSON output in result string")
 
     class BlobMessage(BaseModel):
         blob: bytes

api/core/tools/tool_engine.py

@@ -228,29 +228,41 @@ class ToolEngine:
         """
         Handle tool response
         """
-        result = ""
+        parts: list[str] = []
+        json_parts: list[str] = []
+
         for response in tool_response:
             if response.type == ToolInvokeMessage.MessageType.TEXT:
-                result += cast(ToolInvokeMessage.TextMessage, response.message).text
+                parts.append(cast(ToolInvokeMessage.TextMessage, response.message).text)
             elif response.type == ToolInvokeMessage.MessageType.LINK:
-                result += (
+                parts.append(
                     f"result link: {cast(ToolInvokeMessage.TextMessage, response.message).text}."
                     + " please tell user to check it."
                 )
             elif response.type in {ToolInvokeMessage.MessageType.IMAGE_LINK, ToolInvokeMessage.MessageType.IMAGE}:
-                result += (
+                parts.append(
                     "image has been created and sent to user already, "
                     + "you do not need to create it, just tell the user to check it now."
                 )
             elif response.type == ToolInvokeMessage.MessageType.JSON:
-                result += json.dumps(
-                    safe_json_value(cast(ToolInvokeMessage.JsonMessage, response.message).json_object),
-                    ensure_ascii=False,
+                json_message = cast(ToolInvokeMessage.JsonMessage, response.message)
+                if json_message.suppress_output:
+                    continue
+                json_parts.append(
+                    json.dumps(
+                        safe_json_value(cast(ToolInvokeMessage.JsonMessage, response.message).json_object),
+                        ensure_ascii=False,
+                    )
                 )
             else:
-                result += str(response.message)
+                parts.append(str(response.message))
 
-        return result
+        # Add JSON parts, avoiding duplicates from text parts.
+        if json_parts:
+            existing_parts = set(parts)
+            parts.extend(p for p in json_parts if p not in existing_parts)
+
+        return "".join(parts)
 
     @staticmethod
     def _extract_tool_response_binary_and_text(

api/core/tools/workflow_as_tool/tool.py

@@ -117,7 +117,7 @@ class WorkflowTool(Tool):
         self._latest_usage = self._derive_usage_from_result(data)
 
         yield self.create_text_message(json.dumps(outputs, ensure_ascii=False))
-        yield self.create_json_message(outputs)
+        yield self.create_json_message(outputs, suppress_output=True)
 
     @property
     def latest_usage(self) -> LLMUsage:

api/core/variables/segments.py

@@ -203,6 +203,48 @@ class ArrayFileSegment(ArraySegment):
         return ""
 
 
+class VersionedMemoryValue(BaseModel):
+    current_value: str = None  # type: ignore
+    versions: Mapping[str, str] = {}
+
+    model_config = ConfigDict(frozen=True)
+
+    def add_version(
+        self,
+        new_value: str,
+        version_name: str | None = None
+    ) -> "VersionedMemoryValue":
+        if version_name is None:
+            version_name = str(len(self.versions) + 1)
+        if version_name in self.versions:
+            raise ValueError(f"Version '{version_name}' already exists.")
+        self.current_value = new_value
+        return VersionedMemoryValue(
+            current_value=new_value,
+            versions={
+                version_name: new_value,
+                **self.versions,
+            }
+        )
+
+
+class VersionedMemorySegment(Segment):
+    value_type: SegmentType = SegmentType.VERSIONED_MEMORY
+    value: VersionedMemoryValue = None  # type: ignore
+
+    @property
+    def text(self) -> str:
+        return self.value.current_value
+
+    @property
+    def log(self) -> str:
+        return self.value.current_value
+
+    @property
+    def markdown(self) -> str:
+        return self.value.current_value
+
+
 class ArrayBooleanSegment(ArraySegment):
     value_type: SegmentType = SegmentType.ARRAY_BOOLEAN
     value: Sequence[bool]
@@ -248,6 +290,7 @@ SegmentUnion: TypeAlias = Annotated[
         | Annotated[ArrayObjectSegment, Tag(SegmentType.ARRAY_OBJECT)]
         | Annotated[ArrayFileSegment, Tag(SegmentType.ARRAY_FILE)]
         | Annotated[ArrayBooleanSegment, Tag(SegmentType.ARRAY_BOOLEAN)]
+        | Annotated[VersionedMemorySegment, Tag(SegmentType.VERSIONED_MEMORY)]
     ),
     Discriminator(get_segment_discriminator),
 ]

api/core/variables/types.py

@@ -41,6 +41,8 @@ class SegmentType(StrEnum):
     ARRAY_FILE = "array[file]"
     ARRAY_BOOLEAN = "array[boolean]"
 
+    VERSIONED_MEMORY = "versioned_memory"
+
     NONE = "none"
 
     GROUP = "group"

api/core/variables/variables.py

@@ -22,6 +22,7 @@ from .segments import (
     ObjectSegment,
     Segment,
     StringSegment,
+    VersionedMemorySegment,
     get_segment_discriminator,
 )
 from .types import SegmentType
@@ -106,6 +107,10 @@ class ArrayFileVariable(ArrayFileSegment, ArrayVariable):
     pass
 
 
+class VersionedMemoryVariable(VersionedMemorySegment, Variable):
+    pass
+
+
 class ArrayBooleanVariable(ArrayBooleanSegment, ArrayVariable):
     pass
 
@@ -161,6 +166,7 @@ VariableUnion: TypeAlias = Annotated[
         | Annotated[ArrayFileVariable, Tag(SegmentType.ARRAY_FILE)]
         | Annotated[ArrayBooleanVariable, Tag(SegmentType.ARRAY_BOOLEAN)]
         | Annotated[SecretVariable, Tag(SegmentType.SECRET)]
+        | Annotated[VersionedMemoryVariable, Tag(SegmentType.VERSIONED_MEMORY)]
     ),
     Discriminator(get_segment_discriminator),
 ]

api/core/workflow/constants.py

@@ -1,4 +1,5 @@
 SYSTEM_VARIABLE_NODE_ID = "sys"
 ENVIRONMENT_VARIABLE_NODE_ID = "env"
 CONVERSATION_VARIABLE_NODE_ID = "conversation"
+MEMORY_BLOCK_VARIABLE_NODE_ID = "memory_block"
 RAG_PIPELINE_VARIABLE_NODE_ID = "rag"

api/core/workflow/entities/__init__.py

@@ -4,6 +4,7 @@ from .agent import AgentNodeStrategyInit
 from .graph_init_params import GraphInitParams
 from .workflow_execution import WorkflowExecution
 from .workflow_node_execution import WorkflowNodeExecution
+from .workflow_pause import WorkflowPauseEntity
 
 __all__ = [
     "AgentNodeStrategyInit",
@@ -12,4 +13,5 @@ __all__ = [
     "VariablePool",
     "WorkflowExecution",
     "WorkflowNodeExecution",
+    "WorkflowPauseEntity",
 ]

api/core/workflow/entities/pause_reason.py

@@ -0,0 +1,49 @@
+from enum import StrEnum, auto
+from typing import Annotated, Any, ClassVar, TypeAlias
+
+from pydantic import BaseModel, Discriminator, Tag
+
+
+class _PauseReasonType(StrEnum):
+    HUMAN_INPUT_REQUIRED = auto()
+    SCHEDULED_PAUSE = auto()
+
+
+class _PauseReasonBase(BaseModel):
+    TYPE: ClassVar[_PauseReasonType]
+
+
+class HumanInputRequired(_PauseReasonBase):
+    TYPE = _PauseReasonType.HUMAN_INPUT_REQUIRED
+
+
+class SchedulingPause(_PauseReasonBase):
+    TYPE = _PauseReasonType.SCHEDULED_PAUSE
+
+    message: str
+
+
+def _get_pause_reason_discriminator(v: Any) -> _PauseReasonType | None:
+    if isinstance(v, _PauseReasonBase):
+        return v.TYPE
+    elif isinstance(v, dict):
+        reason_type_str = v.get("TYPE")
+        if reason_type_str is None:
+            return None
+        try:
+            reason_type = _PauseReasonType(reason_type_str)
+        except ValueError:
+            return None
+        return reason_type
+    else:
+        # return None if the discriminator value isn't found
+        return None
+
+
+PauseReason: TypeAlias = Annotated[
+    (
+        Annotated[HumanInputRequired, Tag(_PauseReasonType.HUMAN_INPUT_REQUIRED)]
+        | Annotated[SchedulingPause, Tag(_PauseReasonType.SCHEDULED_PAUSE)]
+    ),
+    Discriminator(_get_pause_reason_discriminator),
+]

api/core/workflow/entities/workflow_pause.py

@@ -0,0 +1,61 @@
+"""
+Domain entities for workflow pause management.
+
+This module contains the domain model for workflow pause, which is used
+by the core workflow module. These models are independent of the storage mechanism
+and don't contain implementation details like tenant_id, app_id, etc.
+"""
+
+from abc import ABC, abstractmethod
+from datetime import datetime
+
+
+class WorkflowPauseEntity(ABC):
+    """
+    Abstract base class for workflow pause entities.
+
+    This domain model represents a paused workflow execution state,
+    without implementation details like tenant_id, app_id, etc.
+    It provides the interface for managing workflow pause/resume operations
+    and state persistence through file storage.
+
+    The `WorkflowPauseEntity` is never reused. If a workflow execution pauses multiple times,
+    it will generate multiple `WorkflowPauseEntity` records.
+    """
+
+    @property
+    @abstractmethod
+    def id(self) -> str:
+        """The identifier of current WorkflowPauseEntity"""
+        pass
+
+    @property
+    @abstractmethod
+    def workflow_execution_id(self) -> str:
+        """The identifier of the workflow execution record the pause associated with.
+        Correspond to `WorkflowExecution.id`.
+        """
+
+    @abstractmethod
+    def get_state(self) -> bytes:
+        """
+        Retrieve the serialized workflow state from storage.
+
+        This method should load and return the workflow execution state
+        that was saved when the workflow was paused. The state contains
+        all necessary information to resume the workflow execution.
+
+        Returns:
+            bytes: The serialized workflow state containing
+            execution context, variable values, node states, etc.
+
+        """
+        ...
+
+    @property
+    @abstractmethod
+    def resumed_at(self) -> datetime | None:
+        """`resumed_at` return the resumption time of the current pause, or `None` if
+        the pause is not resumed yet.
+        """
+        pass

api/core/workflow/enums.py

@@ -92,13 +92,111 @@ class WorkflowType(StrEnum):
 
 
 class WorkflowExecutionStatus(StrEnum):
+    # State diagram for the workflw status:
+    # (@) means start, (*) means end
+    #
+    #       ┌------------------>------------------------->------------------->--------------┐
+    #       |                                                                               |
+    #       |                       ┌-----------------------<--------------------┐          |
+    #       ^                       |                                            |          |
+    #       |                       |                                            ^          |
+    #       |                       V                                            |          |
+    # ┌-----------┐        ┌-----------------------┐                       ┌-----------┐    V
+    # | Scheduled |------->|        Running        |---------------------->|   paused  |    |
+    # └-----------┘        └-----------------------┘                       └-----------┘    |
+    #       |                |       |       |    |                              |          |
+    #       |                |       |       |    |                              |          |
+    #       ^                |       |       |    V                              V          |
+    #       |                |       |       |    |                         ┌---------┐     |
+    #      (@)               |       |       |    └------------------------>| Stopped |<----┘
+    #                        |       |       |                              └---------┘
+    #                        |       |       |                                   |
+    #                        |       |       V                                   V
+    #                        |       |  ┌-----------┐                            |
+    #                        |       |  | Succeeded |------------->--------------┤
+    #                        |       |  └-----------┘                            |
+    #                        |       V                                           V
+    #                        |  +--------┐                                       |
+    #                        |  | Failed |---------------------->----------------┤
+    #                        |  └--------┘                                       |
+    #                        V                                                   V
+    #             ┌---------------------┐                                        |
+    #             | Partially Succeeded |---------------------->-----------------┘--------> (*)
+    #             └---------------------┘
+    #
+    # Mermaid diagram:
+    #
+    #     ---
+    #     title: State diagram for Workflow run state
+    #     ---
+    #     stateDiagram-v2
+    #         scheduled: Scheduled
+    #         running: Running
+    #         succeeded: Succeeded
+    #         failed: Failed
+    #         partial_succeeded: Partial Succeeded
+    #         paused: Paused
+    #         stopped: Stopped
+    #
+    #         [*] --> scheduled:
+    #         scheduled --> running: Start Execution
+    #         running --> paused: Human input required
+    #         paused --> running: human input added
+    #         paused --> stopped: User stops execution
+    #         running --> succeeded: Execution finishes without any error
+    #         running --> failed: Execution finishes with errors
+    #         running --> stopped: User stops execution
+    #         running --> partial_succeeded: some execution occurred and handled during execution
+    #
+    #         scheduled --> stopped: User stops execution
+    #
+    #         succeeded --> [*]
+    #         failed --> [*]
+    #         partial_succeeded --> [*]
+    #         stopped --> [*]
+
+    # `SCHEDULED` means that the workflow is scheduled to run, but has not
+    # started running yet. (maybe due to possible worker saturation.)
+    #
+    # This enum value is currently unused.
+    SCHEDULED = "scheduled"
+
+    # `RUNNING` means the workflow is exeuting.
     RUNNING = "running"
+
+    # `SUCCEEDED` means the execution of workflow succeed without any error.
     SUCCEEDED = "succeeded"
+
+    # `FAILED` means the execution of workflow failed without some errors.
     FAILED = "failed"
+
+    # `STOPPED` means the execution of workflow was stopped, either manually
+    # by the user, or automatically by the Dify application (E.G. the moderation
+    # mechanism.)
     STOPPED = "stopped"
+
+    # `PARTIAL_SUCCEEDED` indicates that some errors occurred during the workflow
+    # execution, but they were successfully handled (e.g., by using an error
+    # strategy such as "fail branch" or "default value").
     PARTIAL_SUCCEEDED = "partial-succeeded"
+
+    # `PAUSED` indicates that the workflow execution is temporarily paused
+    # (e.g., awaiting human input) and is expected to resume later.
     PAUSED = "paused"
 
+    def is_ended(self) -> bool:
+        return self in _END_STATE
+
+
+_END_STATE = frozenset(
+    [
+        WorkflowExecutionStatus.SUCCEEDED,
+        WorkflowExecutionStatus.FAILED,
+        WorkflowExecutionStatus.PARTIAL_SUCCEEDED,
+        WorkflowExecutionStatus.STOPPED,
+    ]
+)
+
 
 class WorkflowNodeExecutionMetadataKey(StrEnum):
     """

api/core/workflow/graph_engine/command_processing/command_handlers.py

@@ -3,6 +3,8 @@ from typing import final
 
 from typing_extensions import override
 
+from core.workflow.entities.pause_reason import SchedulingPause
+
 from ..domain.graph_execution import GraphExecution
 from ..entities.commands import AbortCommand, GraphEngineCommand, PauseCommand
 from .command_processor import CommandHandler
@@ -25,4 +27,7 @@ class PauseCommandHandler(CommandHandler):
     def handle(self, command: GraphEngineCommand, execution: GraphExecution) -> None:
         assert isinstance(command, PauseCommand)
         logger.debug("Pausing workflow %s: %s", execution.workflow_id, command.reason)
-        execution.pause(command.reason)
+        # Convert string reason to PauseReason if needed
+        reason = command.reason
+        pause_reason = SchedulingPause(message=reason)
+        execution.pause(pause_reason)

api/core/workflow/graph_engine/domain/graph_execution.py

@@ -8,6 +8,7 @@ from typing import Literal
 
 from pydantic import BaseModel, Field
 
+from core.workflow.entities.pause_reason import PauseReason
 from core.workflow.enums import NodeState
 
 from .node_execution import NodeExecution
@@ -41,7 +42,7 @@ class GraphExecutionState(BaseModel):
     completed: bool = Field(default=False)
     aborted: bool = Field(default=False)
     paused: bool = Field(default=False)
-    pause_reason: str | None = Field(default=None)
+    pause_reason: PauseReason | None = Field(default=None)
     error: GraphExecutionErrorState | None = Field(default=None)
     exceptions_count: int = Field(default=0)
     node_executions: list[NodeExecutionState] = Field(default_factory=list[NodeExecutionState])
@@ -106,7 +107,7 @@ class GraphExecution:
     completed: bool = False
     aborted: bool = False
     paused: bool = False
-    pause_reason: str | None = None
+    pause_reason: PauseReason | None = None
     error: Exception | None = None
     node_executions: dict[str, NodeExecution] = field(default_factory=dict[str, NodeExecution])
     exceptions_count: int = 0
@@ -130,7 +131,7 @@ class GraphExecution:
         self.aborted = True
         self.error = RuntimeError(f"Aborted: {reason}")
 
-    def pause(self, reason: str | None = None) -> None:
+    def pause(self, reason: PauseReason) -> None:
         """Pause the graph execution without marking it complete."""
         if self.completed:
             raise RuntimeError("Cannot pause execution that has completed")

api/core/workflow/graph_engine/entities/commands.py

@@ -36,4 +36,4 @@ class PauseCommand(GraphEngineCommand):
     """Command to pause a running workflow execution."""
 
     command_type: CommandType = Field(default=CommandType.PAUSE, description="Type of command")
-    reason: str | None = Field(default=None, description="Optional reason for pause")
+    reason: str = Field(default="unknown reason", description="reason for pause")

api/core/workflow/graph_engine/event_management/event_handlers.py

@@ -210,7 +210,7 @@ class EventHandler:
     def _(self, event: NodeRunPauseRequestedEvent) -> None:
         """Handle pause requests emitted by nodes."""
 
-        pause_reason = event.reason or "Awaiting human input"
+        pause_reason = event.reason
         self._graph_execution.pause(pause_reason)
         self._state_manager.finish_execution(event.node_id)
         if event.node_id in self._graph.nodes:

api/core/workflow/graph_engine/graph_engine.py

@@ -247,8 +247,11 @@ class GraphEngine:
 
             # Handle completion
             if self._graph_execution.is_paused:
+                pause_reason = self._graph_execution.pause_reason
+                assert pause_reason is not None, "pause_reason should not be None when execution is paused."
+                # Ensure we have a valid PauseReason for the event
                 paused_event = GraphRunPausedEvent(
-                    reason=self._graph_execution.pause_reason,
+                    reason=pause_reason,
                     outputs=self._graph_runtime_state.outputs,
                 )
                 self._event_manager.notify_layers(paused_event)