external knowledge api

# Conflicts:
#	api/core/rag/datasource/retrieval_service.py
#	api/models/dataset.py
#	api/services/dataset_service.py

CONTRIBUTING_CN.md

@@ -36,7 +36,7 @@
   | 被团队成员标记为高优先级的功能    | 高优先级   |
   | 在 [community feedback board](https://github.com/langgenius/dify/discussions/categories/feedbacks) 内反馈的常见功能请求 | 中等优先级 |
   | 非核心功能和小幅改进                     | 低优先级    |
-  | 有价值但不紧急                                   | 未来功能  |
+  | 有价值当不紧急                                   | 未来功能  |
 
 ### 其他任何事情（例如 bug 报告、性能优化、拼写错误更正）：
 * 立即开始编码。
@@ -138,7 +138,7 @@ Dify 的后端使用 Python 编写，使用 [Flask](https://flask.palletsproject
 ├── models                // 描述数据模型和 API 响应的形状
 ├── public                // 如 favicon 等元资源
 ├── service               // 定义 API 操作的形状
-├── test
+├── test                  
 ├── types                 // 函数参数和返回值的描述
 └── utils                 // 共享的实用函数
 ```

api/configs/packaging/__init__.py

@@ -9,7 +9,7 @@ class PackagingInfo(BaseSettings):
 
     CURRENT_VERSION: str = Field(
         description="Dify version",
-        default="0.8.3",
+        default="0.8.2",
     )
 
     COMMIT_SHA: str = Field(

api/controllers/console/__init__.py

@@ -37,7 +37,7 @@ from .auth import activate, data_source_bearer_auth, data_source_oauth, forgot_p
 from .billing import billing
 
 # Import datasets controllers
-from .datasets import data_source, datasets, datasets_document, datasets_segments, file, hit_testing, website
+from .datasets import data_source, datasets, datasets_document, datasets_segments, external, file, hit_testing, website
 
 # Import explore controllers
 from .explore import (

api/controllers/console/datasets/datasets.py

@@ -110,6 +110,26 @@ class DatasetListApi(Resource):
             nullable=True,
             help="Invalid indexing technique.",
         )
+        parser.add_argument(
+            "external_api_template_id",
+            type=str,
+            nullable=True,
+            required=False,
+        )
+        parser.add_argument(
+            "provider",
+            type=str,
+            nullable=True,
+            choices=Dataset.PROVIDER_LIST,
+            required=False,
+            default="vendor",
+        )
+        parser.add_argument(
+            "external_knowledge_id",
+            type=str,
+            nullable=True,
+            required=False,
+        )
         args = parser.parse_args()
 
         # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
@@ -123,6 +143,9 @@ class DatasetListApi(Resource):
                 indexing_technique=args["indexing_technique"],
                 account=current_user,
                 permission=DatasetPermissionEnum.ONLY_ME,
+                provider=args["provider"],
+                external_api_template_id=args["external_api_template_id"],
+                external_knowledge_id=args["external_knowledge_id"],
             )
         except services.errors.dataset.DatasetNameDuplicateError:
             raise DatasetNameDuplicateError()

api/controllers/console/datasets/external.py

@@ -0,0 +1,254 @@
+from flask import request
+from flask_login import current_user
+from flask_restful import Resource, marshal, reqparse
+from werkzeug.exceptions import Forbidden, NotFound
+
+import services
+from controllers.console import api
+from controllers.console.app.error import ProviderNotInitializeError
+from controllers.console.datasets.error import DatasetNameDuplicateError
+from controllers.console.setup import setup_required
+from controllers.console.wraps import account_initialization_required
+from fields.dataset_fields import dataset_detail_fields
+from libs.login import login_required
+from services.external_knowledge_service import ExternalDatasetService
+
+
+def _validate_name(name):
+    if not name or len(name) < 1 or len(name) > 100:
+        raise ValueError("Name must be between 1 to 100 characters.")
+    return name
+
+
+def _validate_description_length(description):
+    if len(description) > 400:
+        raise ValueError("Description cannot exceed 400 characters.")
+    return description
+
+
+class ExternalApiTemplateListApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self):
+        page = request.args.get("page", default=1, type=int)
+        limit = request.args.get("limit", default=20, type=int)
+        search = request.args.get("keyword", default=None, type=str)
+
+        api_templates, total = ExternalDatasetService.get_external_api_templates(
+            page, limit, current_user.current_tenant_id, search
+        )
+        response = {
+            "data": [item.to_dict() for item in api_templates],
+            "has_more": len(api_templates) == limit,
+            "limit": limit,
+            "total": total,
+            "page": page,
+        }
+        return response, 200
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument(
+            "name",
+            nullable=False,
+            required=True,
+            help="Name is required. Name must be between 1 to 100 characters.",
+            type=_validate_name,
+        )
+        parser.add_argument(
+            "description",
+            nullable=False,
+            required=True,
+            help="Description is required. Description must be between 1 to 400 characters.",
+            type=_validate_description_length,
+        )
+        parser.add_argument(
+            "settings",
+            type=dict,
+            location="json",
+            nullable=False,
+            required=True,
+        )
+        args = parser.parse_args()
+
+        ExternalDatasetService.validate_api_list(args["settings"])
+
+        # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
+        if not current_user.is_dataset_editor:
+            raise Forbidden()
+
+        try:
+            api_template = ExternalDatasetService.create_api_template(
+                tenant_id=current_user.current_tenant_id, user_id=current_user.id, args=args
+            )
+        except services.errors.dataset.DatasetNameDuplicateError:
+            raise DatasetNameDuplicateError()
+
+        return api_template.to_dict(), 201
+
+
+class ExternalApiTemplateApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self, api_template_id):
+        api_template_id = str(api_template_id)
+        api_template = ExternalDatasetService.get_api_template(api_template_id)
+        if api_template is None:
+            raise NotFound("API template not found.")
+
+        return api_template.to_dict(), 200
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def patch(self, api_template_id):
+        api_template_id = str(api_template_id)
+
+        parser = reqparse.RequestParser()
+        parser.add_argument(
+            "name",
+            nullable=False,
+            required=True,
+            help="type is required. Name must be between 1 to 100 characters.",
+            type=_validate_name,
+        )
+        parser.add_argument(
+            "description",
+            nullable=False,
+            required=True,
+            help="description is required. Description must be between 1 to 400 characters.",
+            type=_validate_description_length,
+        )
+        parser.add_argument(
+            "settings",
+            type=dict,
+            location="json",
+            nullable=False,
+            required=True,
+        )
+        args = parser.parse_args()
+        ExternalDatasetService.validate_api_list(args["settings"])
+
+        api_template = ExternalDatasetService.update_api_template(
+            tenant_id=current_user.current_tenant_id,
+            user_id=current_user.id,
+            api_template_id=api_template_id,
+            args=args,
+        )
+
+        return api_template.to_dict(), 200
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def delete(self, api_template_id):
+        api_template_id = str(api_template_id)
+
+        # The role of the current user in the ta table must be admin, owner, or editor
+        if not current_user.is_editor or current_user.is_dataset_operator:
+            raise Forbidden()
+
+        ExternalDatasetService.delete_api_template(current_user.current_tenant_id, api_template_id)
+        return {"result": "success"}, 204
+
+
+class ExternalApiUseCheckApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self, api_template_id):
+        api_template_id = str(api_template_id)
+
+        external_api_template_is_using = ExternalDatasetService.external_api_template_use_check(api_template_id)
+        return {"is_using": external_api_template_is_using}, 200
+
+
+class ExternalDatasetInitApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self):
+        # 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("api_template_id", type=str, required=True, nullable=True, location="json")
+        # parser.add_argument('name', nullable=False, required=True,
+        #                     help='name is required. Name must be between 1 to 100 characters.',
+        #                     type=_validate_name)
+        # parser.add_argument('description', type=str, required=True, nullable=True, location='json')
+        parser.add_argument("data_source", type=dict, required=True, nullable=True, location="json")
+        parser.add_argument("process_parameter", type=dict, required=True, nullable=True, location="json")
+
+        args = parser.parse_args()
+
+        # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
+        if not current_user.is_dataset_editor:
+            raise Forbidden()
+
+        # validate args
+        ExternalDatasetService.document_create_args_validate(
+            current_user.current_tenant_id, args["api_template_id"], args["process_parameter"]
+        )
+
+        try:
+            dataset, documents, batch = ExternalDatasetService.init_external_dataset(
+                tenant_id=current_user.current_tenant_id,
+                user_id=current_user.id,
+                args=args,
+            )
+        except Exception as ex:
+            raise ProviderNotInitializeError(ex.description)
+        response = {"dataset": dataset, "documents": documents, "batch": batch}
+
+        return response
+
+
+class ExternalDatasetCreateApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self):
+        # 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("external_api_template_id", type=str, required=True, nullable=False, location="json")
+        parser.add_argument("external_knowledge_id", type=str, required=True, nullable=False, location="json")
+        parser.add_argument(
+            "name",
+            nullable=False,
+            required=True,
+            help="name is required. Name must be between 1 to 100 characters.",
+            type=_validate_name,
+        )
+        parser.add_argument("description", type=str, required=True, nullable=True, location="json")
+
+        args = parser.parse_args()
+
+        # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
+        if not current_user.is_dataset_editor:
+            raise Forbidden()
+
+        try:
+            dataset = ExternalDatasetService.create_external_dataset(
+                tenant_id=current_user.current_tenant_id,
+                user_id=current_user.id,
+                args=args,
+            )
+        except services.errors.dataset.DatasetNameDuplicateError:
+            raise DatasetNameDuplicateError()
+
+        return marshal(dataset, dataset_detail_fields), 201
+
+
+api.add_resource(ExternalApiTemplateListApi, "/datasets/external-api-template")
+api.add_resource(ExternalApiTemplateApi, "/datasets/external-api-template/<uuid:api_template_id>")
+api.add_resource(ExternalApiUseCheckApi, "/datasets/external-api-template/<uuid:api_template_id>/use-check")

api/controllers/console/datasets/hit_testing.py

@@ -47,6 +47,7 @@ class HitTestingApi(Resource):
         parser = reqparse.RequestParser()
         parser.add_argument("query", type=str, location="json")
         parser.add_argument("retrieval_model", type=dict, required=False, location="json")
+        parser.add_argument("external_retrival_model", type=dict, required=False, location="json")
         args = parser.parse_args()
 
         HitTestingService.hit_testing_args_check(args)
@@ -57,6 +58,7 @@ class HitTestingApi(Resource):
                 query=args["query"],
                 account=current_user,
                 retrieval_model=args["retrieval_model"],
+                external_retrieval_model=args["external_retrival_model"],
                 limit=10,
             )
 

api/controllers/console/datasets/test_external.py

@@ -0,0 +1,49 @@
+from flask import request
+from flask_login import current_user
+from flask_restful import Resource, marshal, reqparse
+from werkzeug.exceptions import Forbidden, NotFound
+
+import services
+from controllers.console import api
+from controllers.console.app.error import ProviderNotInitializeError
+from controllers.console.datasets.error import DatasetNameDuplicateError
+from controllers.console.setup import setup_required
+from controllers.console.wraps import account_initialization_required
+from fields.dataset_fields import dataset_detail_fields
+from libs.login import login_required
+from services.external_knowledge_service import ExternalDatasetService
+
+class TestExternalApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument(
+            "top_k",
+            nullable=False,
+            required=True,
+            type=int,
+        )
+        parser.add_argument(
+            "score_threshold",
+            nullable=False,
+            required=True,
+            type=float,
+        )
+        args = parser.parse_args()
+        result = ExternalDatasetService.test_external_knowledge_retrival(
+            args["top_k"], args["score_threshold"]
+        )
+        response = {
+            "data": [item.to_dict() for item in api_templates],
+            "has_more": len(api_templates) == limit,
+            "limit": limit,
+            "total": total,
+            "page": page,
+        }
+        return response, 200
+
+
+
+api.add_resource(TestExternalApi, "/dify/external-knowledge/retrival-documents")

api/controllers/service_api/dataset/dataset.py

@@ -82,6 +82,26 @@ class DatasetListApi(DatasetApiResource):
             required=False,
             nullable=False,
         )
+        parser.add_argument(
+            "external_api_template_id",
+            type=str,
+            nullable=True,
+            required=False,
+            default="_validate_name",
+        )
+        parser.add_argument(
+            "provider",
+            type=str,
+            nullable=True,
+            required=False,
+            default="vendor",
+        )
+        parser.add_argument(
+            "external_knowledge_id",
+            type=str,
+            nullable=True,
+            required=False,
+        )
         args = parser.parse_args()
 
         try:
@@ -91,6 +111,9 @@ class DatasetListApi(DatasetApiResource):
                 indexing_technique=args["indexing_technique"],
                 account=current_user,
                 permission=args["permission"],
+                provider=args["provider"],
+                external_api_template_id=args["external_api_template_id"],
+                external_knowledge_id=args["external_knowledge_id"],
             )
         except services.errors.dataset.DatasetNameDuplicateError:
             raise DatasetNameDuplicateError()

api/core/model_runtime/model_providers/hunyuan/llm/_position.yaml

@@ -3,4 +3,3 @@
 - hunyuan-standard-256k
 - hunyuan-pro
 - hunyuan-turbo
-- hunyuan-vision

api/core/model_runtime/model_providers/hunyuan/llm/hunyuan-vision.yaml

@@ -1,39 +0,0 @@
-model: hunyuan-vision
-label:
-  zh_Hans: hunyuan-vision
-  en_US: hunyuan-vision
-model_type: llm
-features:
-  - agent-thought
-  - tool-call
-  - multi-tool-call
-  - stream-tool-call
-  - vision
-model_properties:
-  mode: chat
-  context_size: 8000
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-  - name: top_p
-    use_template: top_p
-  - name: max_tokens
-    use_template: max_tokens
-    default: 1024
-    min: 1
-    max: 8000
-  - name: enable_enhance
-    label:
-      zh_Hans: 功能增强
-      en_US: Enable Enhancement
-    type: boolean
-    help:
-      zh_Hans: 功能增强（如搜索）开关，关闭时将直接由主模型生成回复内容，可以降低响应时延（对于流式输出时的首字时延尤为明显）。但在少数场景里，回复效果可能会下降。
-      en_US: Allow the model to perform external search to enhance the generation results.
-    required: false
-    default: true
-pricing:
-  input: '0.018'
-  output: '0.018'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/hunyuan/llm/llm.py

@@ -1,7 +1,6 @@
 import json
 import logging
 from collections.abc import Generator
-from typing import cast
 
 from tencentcloud.common import credential
 from tencentcloud.common.exception import TencentCloudSDKException
@@ -12,12 +11,9 @@ from tencentcloud.hunyuan.v20230901 import hunyuan_client, models
 from core.model_runtime.entities.llm_entities import LLMResult, LLMResultChunk, LLMResultChunkDelta
 from core.model_runtime.entities.message_entities import (
     AssistantPromptMessage,
-    ImagePromptMessageContent,
     PromptMessage,
-    PromptMessageContentType,
     PromptMessageTool,
     SystemPromptMessage,
-    TextPromptMessageContent,
     ToolPromptMessage,
     UserPromptMessage,
 )
@@ -147,25 +143,6 @@ class HunyuanLargeLanguageModel(LargeLanguageModel):
                 tool_execute_result = {"result": message.content}
                 content = json.dumps(tool_execute_result, ensure_ascii=False)
                 dict_list.append({"Role": message.role.value, "Content": content, "ToolCallId": message.tool_call_id})
-            elif isinstance(message, UserPromptMessage):
-                message = cast(UserPromptMessage, message)
-                if isinstance(message.content, str):
-                    dict_list.append({"Role": message.role.value, "Content": message.content})
-                else:
-                    sub_messages = []
-                    for message_content in message.content:
-                        if message_content.type == PromptMessageContentType.TEXT:
-                            message_content = cast(TextPromptMessageContent, message_content)
-                            sub_message_dict = {"Type": "text", "Text": message_content.data}
-                            sub_messages.append(sub_message_dict)
-                        elif message_content.type == PromptMessageContentType.IMAGE:
-                            message_content = cast(ImagePromptMessageContent, message_content)
-                            sub_message_dict = {
-                                "Type": "image_url",
-                                "ImageUrl": {"Url": message_content.data},
-                            }
-                            sub_messages.append(sub_message_dict)
-                    dict_list.append({"Role": message.role.value, "Contents": sub_messages})
             else:
                 dict_list.append({"Role": message.role.value, "Content": message.content})
         return dict_list

api/core/model_runtime/model_providers/jina/text_embedding/text_embedding.py

@@ -57,7 +57,7 @@ class JinaTextEmbeddingModel(TextEmbeddingModel):
         data = {"model": model, "input": [transform_jina_input_text(model, text) for text in texts]}
 
         if model == "jina-embeddings-v3":
-            data["task"] = "text-matching"
+            data["task_type"] = "retrieval.passage"
 
         try:
             response = post(url, headers=headers, data=dumps(data))

api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-0613.yaml

@@ -31,4 +31,3 @@ pricing:
   output: '0.002'
   unit: '0.001'
   currency: USD
-deprecated: true

api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-16k-0613.yaml

@@ -31,4 +31,3 @@ pricing:
   output: '0.004'
   unit: '0.001'
   currency: USD
-deprecated: true

api/core/model_runtime/model_providers/tongyi/llm/_position.yaml

@@ -1,51 +0,0 @@
-- qwen-vl-max-0809
-- qwen-vl-max-0201
-- qwen-vl-max
-- qwen-max-latest
-- qwen-max-1201
-- qwen-max-0919
-- qwen-max-0428
-- qwen-max-0403
-- qwen-max-0107
-- qwen-max
-- qwen-max-longcontext
-- qwen-plus-latest
-- qwen-plus-0919
-- qwen-plus-0806
-- qwen-plus-0723
-- qwen-plus-0624
-- qwen-plus-0206
-- qwen-plus-chat
-- qwen-plus
-- qwen-vl-plus-0809
-- qwen-vl-plus-0201
-- qwen-vl-plus
-- qwen-turbo-latest
-- qwen-turbo-0919
-- qwen-turbo-0624
-- qwen-turbo-0206
-- qwen-turbo-chat
-- qwen-turbo
-- qwen2.5-72b-instruct
-- qwen2.5-32b-instruct
-- qwen2.5-14b-instruct
-- qwen2.5-7b-instruct
-- qwen2.5-3b-instruct
-- qwen2.5-1.5b-instruct
-- qwen2.5-0.5b-instruct
-- qwen2.5-coder-7b-instruct
-- qwen2-math-72b-instruct
-- qwen2-math-7b-instruct
-- qwen2-math-1.5b-instruct
-- qwen-long
-- qwen-math-plus-latest
-- qwen-math-plus-0919
-- qwen-math-plus-0816
-- qwen-math-plus
-- qwen-math-turbo-latest
-- qwen-math-turbo-0919
-- qwen-math-turbo
-- qwen-coder-turbo-latest
-- qwen-coder-turbo-0919
-- qwen-coder-turbo
-- farui-plus

api/core/model_runtime/model_providers/tongyi/llm/qwen-coder-turbo-0919.yaml

@@ -1,79 +0,0 @@
-model: qwen-coder-turbo-0919
-label:
-  en_US: qwen-coder-turbo-0919
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-coder-turbo-latest.yaml

@@ -1,79 +0,0 @@
-model: qwen-coder-turbo-latest
-label:
-  en_US: qwen-coder-turbo-latest
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-coder-turbo.yaml

@@ -1,79 +0,0 @@
-model: qwen-coder-turbo
-label:
-  en_US: qwen-coder-turbo
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-long.yaml

@@ -1,4 +1,3 @@
-# model docs: https://help.aliyun.com/zh/model-studio/getting-started/models#27b2b3a15d5c6
 model: qwen-long
 label:
   en_US: qwen-long

api/core/model_runtime/model_providers/tongyi/llm/qwen-math-plus-0816.yaml

@@ -1,79 +0,0 @@
-model: qwen-math-plus-0816
-label:
-  en_US: qwen-math-plus-0816
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 3072
-    min: 1
-    max: 3072
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-math-plus-0919.yaml

@@ -1,79 +0,0 @@
-model: qwen-math-plus-0919
-label:
-  en_US: qwen-math-plus-0919
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 3072
-    min: 1
-    max: 3072
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-math-plus-latest.yaml

@@ -1,79 +0,0 @@
-model: qwen-math-plus-latest
-label:
-  en_US: qwen-math-plus-latest
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 3072
-    min: 1
-    max: 3072
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-math-plus.yaml

@@ -1,79 +0,0 @@
-model: qwen-math-plus
-label:
-  en_US: qwen-math-plus
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 3072
-    min: 1
-    max: 3072
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-math-turbo-0919.yaml

@@ -1,79 +0,0 @@
-model: qwen-math-turbo-0919
-label:
-  en_US: qwen-math-turbo-0919
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 3072
-    min: 1
-    max: 3072
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-math-turbo-latest.yaml

@@ -1,79 +0,0 @@
-model: qwen-math-turbo-latest
-label:
-  en_US: qwen-math-turbo-latest
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 3072
-    min: 1
-    max: 3072
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-math-turbo.yaml

@@ -1,79 +0,0 @@
-model: qwen-math-turbo
-label:
-  en_US: qwen-math-turbo
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 3072
-    min: 1
-    max: 3072
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-max-0107.yaml

@@ -1,81 +0,0 @@
-model: qwen-max-0107
-label:
-  en_US: qwen-max-0107
-model_type: llm
-features:
-  - multi-tool-call
-  - agent-thought
-  - stream-tool-call
-model_properties:
-  mode: chat
-  context_size: 8000
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.04'
-  output: '0.12'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-max-0403.yaml

@@ -8,7 +8,7 @@ features:
   - stream-tool-call
 model_properties:
   mode: chat
-  context_size: 8000
+  context_size: 8192
 parameter_rules:
   - name: temperature
     use_template: temperature

api/core/model_runtime/model_providers/tongyi/llm/qwen-max-0428.yaml

@@ -8,7 +8,7 @@ features:
   - stream-tool-call
 model_properties:
   mode: chat
-  context_size: 8000
+  context_size: 8192
 parameter_rules:
   - name: temperature
     use_template: temperature

api/core/model_runtime/model_providers/tongyi/llm/qwen-max-0919.yaml

@@ -1,81 +0,0 @@
-model: qwen-max-0919
-label:
-  en_US: qwen-max-0919
-model_type: llm
-features:
-  - multi-tool-call
-  - agent-thought
-  - stream-tool-call
-model_properties:
-  mode: chat
-  context_size: 32768
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.02'
-  output: '0.06'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-max-1201.yaml

@@ -79,4 +79,3 @@ pricing:
   output: '0.12'
   unit: '0.001'
   currency: RMB
-deprecated: true

api/core/model_runtime/model_providers/tongyi/llm/qwen-max-latest.yaml

@@ -1,81 +0,0 @@
-model: qwen-max-latest
-label:
-  en_US: qwen-max-latest
-model_type: llm
-features:
-  - multi-tool-call
-  - agent-thought
-  - stream-tool-call
-model_properties:
-  mode: chat
-  context_size: 32768
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.02'
-  output: '0.06'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-max-longcontext.yaml

@@ -8,7 +8,7 @@ features:
   - stream-tool-call
 model_properties:
   mode: chat
-  context_size: 32000
+  context_size: 32768
 parameter_rules:
   - name: temperature
     use_template: temperature
@@ -22,9 +22,9 @@ parameter_rules:
   - name: max_tokens
     use_template: max_tokens
     type: int
-    default: 8000
+    default: 2000
     min: 1
-    max: 8000
+    max: 2000
     help:
       zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
       en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.

api/core/model_runtime/model_providers/tongyi/llm/qwen-max.yaml

@@ -8,7 +8,7 @@ features:
   - stream-tool-call
 model_properties:
   mode: chat
-  context_size: 8000
+  context_size: 8192
 parameter_rules:
   - name: temperature
     use_template: temperature
@@ -75,7 +75,7 @@ parameter_rules:
   - name: response_format
     use_template: response_format
 pricing:
-  input: '0.02'
-  output: '0.06'
+  input: '0.04'
+  output: '0.12'
   unit: '0.001'
   currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0206.yaml

@@ -1,79 +0,0 @@
-model: qwen-plus-0206
-label:
-  en_US: qwen-plus-0206
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: completion
-  context_size: 32000
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8000
-    min: 1
-    max: 8000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0624.yaml

@@ -1,79 +0,0 @@
-model: qwen-plus-0624
-label:
-  en_US: qwen-plus-0624
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: completion
-  context_size: 32000
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8000
-    min: 1
-    max: 8000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0723.yaml

@@ -1,79 +0,0 @@
-model: qwen-plus-0723
-label:
-  en_US: qwen-plus-0723
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: completion
-  context_size: 32000
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8000
-    min: 1
-    max: 8000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0806.yaml

@@ -1,79 +0,0 @@
-model: qwen-plus-0806
-label:
-  en_US: qwen-plus-0806
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: completion
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0919.yaml

@@ -1,79 +0,0 @@
-model: qwen-plus-0919
-label:
-  en_US: qwen-plus-0919
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: completion
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.0008'
-  output: '0.002'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-chat.yaml

@@ -79,4 +79,3 @@ pricing:
   output: '0.012'
   unit: '0.001'
   currency: RMB
-deprecated: true

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-latest.yaml

@@ -1,79 +0,0 @@
-model: qwen-plus-latest
-label:
-  en_US: qwen-plus-latest
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.0008'
-  output: '0.002'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-plus.yaml

@@ -3,12 +3,10 @@ label:
   en_US: qwen-plus
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
-  mode: chat
-  context_size: 131072
+  mode: completion
+  context_size: 32768
 parameter_rules:
   - name: temperature
     use_template: temperature
@@ -22,9 +20,9 @@ parameter_rules:
   - name: max_tokens
     use_template: max_tokens
     type: int
-    default: 8192
+    default: 2000
     min: 1
-    max: 8192
+    max: 2000
     help:
       zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
       en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
@@ -75,7 +73,7 @@ parameter_rules:
   - name: response_format
     use_template: response_format
 pricing:
-  input: '0.0008'
-  output: '0.002'
+  input: '0.004'
+  output: '0.012'
   unit: '0.001'
   currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0206.yaml

@@ -1,79 +0,0 @@
-model: qwen-turbo-0206
-label:
-  en_US: qwen-turbo-0206
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 8000
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0624.yaml

@@ -1,79 +0,0 @@
-model: qwen-turbo-0624
-label:
-  en_US: qwen-turbo-0624
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 8000
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0919.yaml

@@ -1,79 +0,0 @@
-model: qwen-turbo-0919
-label:
-  en_US: qwen-turbo-0919
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.0003'
-  output: '0.0006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-chat.yaml

@@ -79,4 +79,3 @@ pricing:
   output: '0.006'
   unit: '0.001'
   currency: RMB
-deprecated: true

api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-latest.yaml

@@ -1,79 +0,0 @@
-model: qwen-turbo-latest
-label:
-  en_US: qwen-turbo-latest
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.0006'
-  output: '0.0003'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo.yaml

@@ -3,12 +3,10 @@ label:
   en_US: qwen-turbo
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
-  mode: chat
-  context_size: 8000
+  mode: completion
+  context_size: 8192
 parameter_rules:
   - name: temperature
     use_template: temperature
@@ -22,9 +20,9 @@ parameter_rules:
   - name: max_tokens
     use_template: max_tokens
     type: int
-    default: 2000
+    default: 1500
     min: 1
-    max: 2000
+    max: 1500
     help:
       zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
       en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
@@ -75,7 +73,7 @@ parameter_rules:
   - name: response_format
     use_template: response_format
 pricing:
-  input: '0.0006'
-  output: '0.0003'
+  input: '0.002'
+  output: '0.006'
   unit: '0.001'
   currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0201.yaml

@@ -1,48 +0,0 @@
-model: qwen-vl-max-0201
-label:
-  en_US: qwen-vl-max-0201
-model_type: llm
-features:
-  - vision
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 8192
-parameter_rules:
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.02'
-  output: '0.02'
-  unit: '0.001'
-  currency: RMB
-deprecated: true

api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0809.yaml

@@ -1,57 +0,0 @@
-model: qwen-vl-max-0809
-label:
-  en_US: qwen-vl-max-0809
-model_type: llm
-features:
-  - vision
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 32000
-parameter_rules:
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: max_tokens
-    required: false
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.02'
-  output: '0.02'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max.yaml

@@ -7,7 +7,7 @@ features:
   - agent-thought
 model_properties:
   mode: chat
-  context_size: 32000
+  context_size: 8192
 parameter_rules:
   - name: top_p
     use_template: top_p
@@ -28,16 +28,6 @@ parameter_rules:
     help:
       zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
       en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: max_tokens
-    required: false
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
   - name: seed
     required: false
     type: int

api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus-0201.yaml

@@ -1,57 +0,0 @@
-model: qwen-vl-plus-0201
-label:
-  en_US: qwen-vl-plus-0201
-model_type: llm
-features:
-  - vision
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 8000
-parameter_rules:
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: max_tokens
-    required: false
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.02'
-  output: '0.02'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus-0809.yaml

@@ -1,57 +0,0 @@
-model: qwen-vl-plus-0809
-label:
-  en_US: qwen-vl-plus-0809
-model_type: llm
-features:
-  - vision
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 32768
-parameter_rules:
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: max_tokens
-    required: false
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.008'
-  output: '0.008'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus.yaml

@@ -7,7 +7,7 @@ features:
   - agent-thought
 model_properties:
   mode: chat
-  context_size: 8000
+  context_size: 32768
 parameter_rules:
   - name: top_p
     use_template: top_p
@@ -28,16 +28,6 @@ parameter_rules:
     help:
       zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
       en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: max_tokens
-    required: false
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
   - name: seed
     required: false
     type: int

api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-1.5b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2-math-1.5b-instruct
-label:
-  en_US: qwen2-math-1.5b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-72b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2-math-72b-instruct
-label:
-  en_US: qwen2-math-72b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-7b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2-math-7b-instruct
-label:
-  en_US: qwen2-math-7b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 4096
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 2000
-    min: 1
-    max: 2000
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-0.5b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-0.5b-instruct
-label:
-  en_US: qwen2.5-0.5b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 32768
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.000'
-  output: '0.000'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-1.5b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-1.5b-instruct
-label:
-  en_US: qwen2.5-1.5b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 32768
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.000'
-  output: '0.000'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-14b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-14b-instruct
-label:
-  en_US: qwen2.5-14b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.002'
-  output: '0.006'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-32b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-32b-instruct
-label:
-  en_US: qwen2.5-32b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.0035'
-  output: '0.007'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-3b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-3b-instruct
-label:
-  en_US: qwen2.5-3b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 32768
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.000'
-  output: '0.000'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-72b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-72b-instruct
-label:
-  en_US: qwen2.5-72b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.004'
-  output: '0.012'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-7b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-7b-instruct
-label:
-  en_US: qwen2.5-7b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.001'
-  output: '0.002'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-coder-7b-instruct.yaml

@@ -1,79 +0,0 @@
-model: qwen2.5-7b-instruct
-label:
-  en_US: qwen2.5-7b-instruct
-model_type: llm
-features:
-  - agent-thought
-model_properties:
-  mode: chat
-  context_size: 131072
-parameter_rules:
-  - name: temperature
-    use_template: temperature
-    type: float
-    default: 0.3
-    min: 0.0
-    max: 2.0
-    help:
-      zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。
-      en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain.
-  - name: max_tokens
-    use_template: max_tokens
-    type: int
-    default: 8192
-    min: 1
-    max: 8192
-    help:
-      zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。
-      en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time.
-  - name: top_p
-    use_template: top_p
-    type: float
-    default: 0.8
-    min: 0.1
-    max: 0.9
-    help:
-      zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。
-      en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated.
-  - name: top_k
-    type: int
-    min: 0
-    max: 99
-    label:
-      zh_Hans: 取样数量
-      en_US: Top k
-    help:
-      zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。
-      en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated.
-  - name: seed
-    required: false
-    type: int
-    default: 1234
-    label:
-      zh_Hans: 随机种子
-      en_US: Random seed
-    help:
-      zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。
-      en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time.
-  - name: repetition_penalty
-    required: false
-    type: float
-    default: 1.1
-    label:
-      en_US: Repetition penalty
-    help:
-      zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。
-      en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment.
-  - name: enable_search
-    type: boolean
-    default: false
-    help:
-      zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。
-      en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic.
-  - name: response_format
-    use_template: response_format
-pricing:
-  input: '0.001'
-  output: '0.002'
-  unit: '0.001'
-  currency: RMB

api/core/model_runtime/model_providers/tongyi/tongyi.yaml

@@ -11,9 +11,9 @@ background: "#EFF1FE"
 help:
   title:
     en_US: Get your API key from AliCloud
-    zh_Hans: 从阿里云百炼获取 API Key
+    zh_Hans: 从阿里云获取 API Key
   url:
-    en_US: https://bailian.console.aliyun.com/?apiKey=1#/api-key
+    en_US: https://dashscope.console.aliyun.com/api-key_management
 supported_model_types:
   - llm
   - tts

api/core/rag/datasource/retrieval_service.py

@@ -10,6 +10,7 @@ from core.rag.rerank.constants.rerank_mode import RerankMode
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
 from extensions.ext_database import db
 from models.dataset import Dataset
+from services.external_knowledge_service import ExternalDatasetService
 
 default_retrieval_model = {
     "search_method": RetrievalMethod.SEMANTIC_SEARCH.value,
@@ -22,91 +23,90 @@ default_retrieval_model = {
 
 class RetrievalService:
     @classmethod
-    def retrieve(
-        cls,
-        retrieval_method: str,
-        dataset_id: str,
-        query: str,
-        top_k: int,
-        score_threshold: Optional[float] = 0.0,
-        reranking_model: Optional[dict] = None,
-        reranking_mode: Optional[str] = "reranking_model",
-        weights: Optional[dict] = None,
-    ):
-        dataset = db.session.query(Dataset).filter(Dataset.id == dataset_id).first()
-        if not dataset or dataset.available_document_count == 0 or dataset.available_segment_count == 0:
+    def retrieve(cls, retrival_method: str, dataset_id: str, query: str,
+                 top_k: int, score_threshold: Optional[float] = .0,
+                 reranking_model: Optional[dict] = None, reranking_mode: Optional[str] = 'reranking_model',
+                 weights: Optional[dict] = None, provider: Optional[str] = None,
+                 external_retrieval_model: Optional[dict] = None):
+        dataset = db.session.query(Dataset).filter(
+            Dataset.id == dataset_id
+        ).first()
+        if not dataset:
             return []
-        all_documents = []
-        threads = []
-        exceptions = []
-        # retrieval_model source with keyword
-        if retrieval_method == "keyword_search":
-            keyword_thread = threading.Thread(
-                target=RetrievalService.keyword_search,
-                kwargs={
-                    "flask_app": current_app._get_current_object(),
-                    "dataset_id": dataset_id,
-                    "query": query,
-                    "top_k": top_k,
-                    "all_documents": all_documents,
-                    "exceptions": exceptions,
-                },
+        if provider == 'external':
+            all_documents = ExternalDatasetService.fetch_external_knowledge_retrival(
+                dataset.tenant_id,
+                dataset_id,
+                query,
+                external_retrieval_model
             )
-            threads.append(keyword_thread)
-            keyword_thread.start()
-        # retrieval_model source with semantic
-        if RetrievalMethod.is_support_semantic_search(retrieval_method):
-            embedding_thread = threading.Thread(
-                target=RetrievalService.embedding_search,
-                kwargs={
-                    "flask_app": current_app._get_current_object(),
-                    "dataset_id": dataset_id,
-                    "query": query,
-                    "top_k": top_k,
-                    "score_threshold": score_threshold,
-                    "reranking_model": reranking_model,
-                    "all_documents": all_documents,
-                    "retrieval_method": retrieval_method,
-                    "exceptions": exceptions,
-                },
-            )
-            threads.append(embedding_thread)
-            embedding_thread.start()
+        else:
+            if not dataset or dataset.available_document_count == 0 or dataset.available_segment_count == 0:
+                return []
+            all_documents = []
+            threads = []
+            exceptions = []
+            # retrieval_model source with keyword
+            if retrival_method == 'keyword_search':
+                keyword_thread = threading.Thread(target=RetrievalService.keyword_search, kwargs={
+                    'flask_app': current_app._get_current_object(),
+                    'dataset_id': dataset_id,
+                    'query': query,
+                    'top_k': top_k,
+                    'all_documents': all_documents,
+                    'exceptions': exceptions,
+                })
+                threads.append(keyword_thread)
+                keyword_thread.start()
+            # retrieval_model source with semantic
+            if RetrievalMethod.is_support_semantic_search(retrival_method):
+                embedding_thread = threading.Thread(target=RetrievalService.embedding_search, kwargs={
+                    'flask_app': current_app._get_current_object(),
+                    'dataset_id': dataset_id,
+                    'query': query,
+                    'top_k': top_k,
+                    'score_threshold': score_threshold,
+                    'reranking_model': reranking_model,
+                    'all_documents': all_documents,
+                    'retrival_method': retrival_method,
+                    'exceptions': exceptions,
+                })
+                threads.append(embedding_thread)
+                embedding_thread.start()
 
-        # retrieval source with full text
-        if RetrievalMethod.is_support_fulltext_search(retrieval_method):
-            full_text_index_thread = threading.Thread(
-                target=RetrievalService.full_text_index_search,
-                kwargs={
-                    "flask_app": current_app._get_current_object(),
-                    "dataset_id": dataset_id,
-                    "query": query,
-                    "retrieval_method": retrieval_method,
-                    "score_threshold": score_threshold,
-                    "top_k": top_k,
-                    "reranking_model": reranking_model,
-                    "all_documents": all_documents,
-                    "exceptions": exceptions,
-                },
-            )
-            threads.append(full_text_index_thread)
-            full_text_index_thread.start()
+            # retrieval source with full text
+            if RetrievalMethod.is_support_fulltext_search(retrival_method):
+                full_text_index_thread = threading.Thread(target=RetrievalService.full_text_index_search, kwargs={
+                    'flask_app': current_app._get_current_object(),
+                    'dataset_id': dataset_id,
+                    'query': query,
+                    'retrival_method': retrival_method,
+                    'score_threshold': score_threshold,
+                    'top_k': top_k,
+                    'reranking_model': reranking_model,
+                    'all_documents': all_documents,
+                    'exceptions': exceptions,
+                })
+                threads.append(full_text_index_thread)
+                full_text_index_thread.start()
 
-        for thread in threads:
-            thread.join()
+            for thread in threads:
+                thread.join()
 
-        if exceptions:
-            exception_message = ";\n".join(exceptions)
-            raise Exception(exception_message)
+            if exceptions:
+                exception_message = ';\n'.join(exceptions)
+                raise Exception(exception_message)
 
-        if retrieval_method == RetrievalMethod.HYBRID_SEARCH.value:
-            data_post_processor = DataPostProcessor(
-                str(dataset.tenant_id), reranking_mode, reranking_model, weights, False
-            )
-            all_documents = data_post_processor.invoke(
-                query=query, documents=all_documents, score_threshold=score_threshold, top_n=top_k
-            )
-        return all_documents
+            if retrival_method == RetrievalMethod.HYBRID_SEARCH.value:
+                data_post_processor = DataPostProcessor(str(dataset.tenant_id), reranking_mode,
+                                                        reranking_model, weights, False)
+                all_documents = data_post_processor.invoke(
+                    query=query,
+                    documents=all_documents,
+                    score_threshold=score_threshold,
+                    top_n=top_k
+                )
+            return all_documents
 
     @classmethod
     def keyword_search(

api/core/tools/provider/builtin/comfyui/tools/comfyui_stable_diffusion.py

@@ -290,7 +290,7 @@ class ComfyuiStableDiffusionTool(BuiltinTool):
         draw_options["6"]["inputs"]["text"] = prompt
         draw_options["7"]["inputs"]["text"] = negative_prompt
         # if the model is SD3 or FLUX series, the Latent class should be corresponding to SD3 Latent
-        if model_type in {ModelType.SD3.name, ModelType.FLUX.name}:
+        if model_type in (ModelType.SD3.name, ModelType.FLUX.name):
             draw_options["5"]["class_type"] = "EmptySD3LatentImage"
 
         if lora_list:
@@ -333,7 +333,7 @@ class ComfyuiStableDiffusionTool(BuiltinTool):
                         break
 
             return self.create_blob_message(
-                blob=image, meta={"mime_type": "image/png"}, save_as=self.VariableKey.IMAGE.value
+                blob=image, meta={"mime_type": "image/png"}, save_as=self.VARIABLE_KEY.IMAGE.value
             )
 
         except Exception as e:

api/core/workflow/graph_engine/graph_engine.py

@@ -61,9 +61,6 @@ class GraphEngineThreadPool(ThreadPoolExecutor):
 
         return super().submit(fn, *args, **kwargs)
 
-    def task_done_callback(self, future):
-        self.submit_count -= 1
-
     def check_is_full(self) -> None:
         print(f"submit_count: {self.submit_count}, max_submit_count: {self.max_submit_count}")
         if self.submit_count > self.max_submit_count:
@@ -429,22 +426,20 @@ class GraphEngine:
             ):
                 continue
 
-            future = self.thread_pool.submit(
-                self._run_parallel_node,
-                **{
-                    "flask_app": current_app._get_current_object(),  # type: ignore[attr-defined]
-                    "q": q,
-                    "parallel_id": parallel_id,
-                    "parallel_start_node_id": edge.target_node_id,
-                    "parent_parallel_id": in_parallel_id,
-                    "parent_parallel_start_node_id": parallel_start_node_id,
-                },
+            futures.append(
+                self.thread_pool.submit(
+                    self._run_parallel_node,
+                    **{
+                        "flask_app": current_app._get_current_object(),  # type: ignore[attr-defined]
+                        "q": q,
+                        "parallel_id": parallel_id,
+                        "parallel_start_node_id": edge.target_node_id,
+                        "parent_parallel_id": in_parallel_id,
+                        "parent_parallel_start_node_id": parallel_start_node_id,
+                    },
+                )
             )
 
-            future.add_done_callback(self.thread_pool.task_done_callback)
-
-            futures.append(future)
-
         succeeded_count = 0
         while True:
             try:

api/extensions/ext_sentry.py

@@ -5,8 +5,6 @@ from sentry_sdk.integrations.celery import CeleryIntegration
 from sentry_sdk.integrations.flask import FlaskIntegration
 from werkzeug.exceptions import HTTPException
 
-from core.model_runtime.errors.invoke import InvokeRateLimitError
-
 
 def before_send(event, hint):
     if "exc_info" in hint:
@@ -22,13 +20,7 @@ def init_app(app):
         sentry_sdk.init(
             dsn=app.config.get("SENTRY_DSN"),
             integrations=[FlaskIntegration(), CeleryIntegration()],
-            ignore_errors=[
-                HTTPException,
-                ValueError,
-                openai.APIStatusError,
-                InvokeRateLimitError,
-                parse_error.defaultErrorResponse,
-            ],
+            ignore_errors=[HTTPException, ValueError, openai.APIStatusError, parse_error.defaultErrorResponse],
             traces_sample_rate=app.config.get("SENTRY_TRACES_SAMPLE_RATE", 1.0),
             profiles_sample_rate=app.config.get("SENTRY_PROFILES_SAMPLE_RATE", 1.0),
             environment=app.config.get("DEPLOY_ENV"),

api/extensions/ext_storage.py

@@ -1,4 +1,3 @@
-import logging
 from collections.abc import Generator
 from typing import Union
 
@@ -41,56 +40,28 @@ class Storage:
             self.storage_runner = LocalStorage(app=app)
 
     def save(self, filename, data):
-        try:
-            self.storage_runner.save(filename, data)
-        except Exception as e:
-            logging.exception("Failed to save file: %s", e)
-            raise e
+        self.storage_runner.save(filename, data)
 
     def load(self, filename: str, stream: bool = False) -> Union[bytes, Generator]:
-        try:
-            if stream:
-                return self.load_stream(filename)
-            else:
-                return self.load_once(filename)
-        except Exception as e:
-            logging.exception("Failed to load file: %s", e)
-            raise e
+        if stream:
+            return self.load_stream(filename)
+        else:
+            return self.load_once(filename)
 
     def load_once(self, filename: str) -> bytes:
-        try:
-            return self.storage_runner.load_once(filename)
-        except Exception as e:
-            logging.exception("Failed to load_once file: %s", e)
-            raise e
+        return self.storage_runner.load_once(filename)
 
     def load_stream(self, filename: str) -> Generator:
-        try:
-            return self.storage_runner.load_stream(filename)
-        except Exception as e:
-            logging.exception("Failed to load_stream file: %s", e)
-            raise e
+        return self.storage_runner.load_stream(filename)
 
     def download(self, filename, target_filepath):
-        try:
-            self.storage_runner.download(filename, target_filepath)
-        except Exception as e:
-            logging.exception("Failed to download file: %s", e)
-            raise e
+        self.storage_runner.download(filename, target_filepath)
 
     def exists(self, filename):
-        try:
-            return self.storage_runner.exists(filename)
-        except Exception as e:
-            logging.exception("Failed to check file exists: %s", e)
-            raise e
+        return self.storage_runner.exists(filename)
 
     def delete(self, filename):
-        try:
-            return self.storage_runner.delete(filename)
-        except Exception as e:
-            logging.exception("Failed to delete file: %s", e)
-            raise e
+        return self.storage_runner.delete(filename)
 
 
 storage = Storage()

api/fields/external_dataset_fields.py

@@ -0,0 +1,11 @@
+from flask_restful import fields
+
+from libs.helper import TimestampField
+
+api_template_query_detail_fields = {
+    "id": fields.String,
+    "name": fields.String,
+    "setting": fields.String,
+    "created_by": fields.String,
+    "created_at": TimestampField,
+}

api/migrations/versions/2024_09_18_0659-ec3df697ebbb_external_knowledge.py

@@ -0,0 +1,74 @@
+"""external_knowledge
+
+Revision ID: ec3df697ebbb
+Revises: 675b5321501b
+Create Date: 2024-09-18 06:59:54.048478
+
+"""
+from alembic import op
+import models as models
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = 'ec3df697ebbb'
+down_revision = '675b5321501b'
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('external_api_templates',
+    sa.Column('id', models.types.StringUUID(), server_default=sa.text('uuid_generate_v4()'), nullable=False),
+    sa.Column('name', sa.String(length=255), nullable=False),
+    sa.Column('description', sa.String(length=255), nullable=False),
+    sa.Column('tenant_id', models.types.StringUUID(), nullable=False),
+    sa.Column('settings', sa.Text(), nullable=True),
+    sa.Column('created_by', models.types.StringUUID(), nullable=False),
+    sa.Column('created_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.Column('updated_by', models.types.StringUUID(), nullable=True),
+    sa.Column('updated_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.PrimaryKeyConstraint('id', name='external_api_template_pkey')
+    )
+    with op.batch_alter_table('external_api_templates', schema=None) as batch_op:
+        batch_op.create_index('external_api_templates_name_idx', ['name'], unique=False)
+        batch_op.create_index('external_api_templates_tenant_idx', ['tenant_id'], unique=False)
+
+    op.create_table('external_knowledge_bindings',
+    sa.Column('id', models.types.StringUUID(), server_default=sa.text('uuid_generate_v4()'), nullable=False),
+    sa.Column('tenant_id', models.types.StringUUID(), nullable=False),
+    sa.Column('external_api_template_id', models.types.StringUUID(), nullable=False),
+    sa.Column('dataset_id', models.types.StringUUID(), nullable=False),
+    sa.Column('external_knowledge_id', sa.Text(), nullable=False),
+    sa.Column('created_by', models.types.StringUUID(), nullable=False),
+    sa.Column('created_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.Column('updated_by', models.types.StringUUID(), nullable=True),
+    sa.Column('updated_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.PrimaryKeyConstraint('id', name='external_knowledge_bindings_pkey')
+    )
+    with op.batch_alter_table('external_knowledge_bindings', schema=None) as batch_op:
+        batch_op.create_index('external_knowledge_bindings_dataset_idx', ['dataset_id'], unique=False)
+        batch_op.create_index('external_knowledge_bindings_external_api_template_idx', ['external_api_template_id'], unique=False)
+        batch_op.create_index('external_knowledge_bindings_external_knowledge_idx', ['external_knowledge_id'], unique=False)
+        batch_op.create_index('external_knowledge_bindings_tenant_idx', ['tenant_id'], unique=False)
+
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+
+    with op.batch_alter_table('external_knowledge_bindings', schema=None) as batch_op:
+        batch_op.drop_index('external_knowledge_bindings_tenant_idx')
+        batch_op.drop_index('external_knowledge_bindings_external_knowledge_idx')
+        batch_op.drop_index('external_knowledge_bindings_external_api_template_idx')
+        batch_op.drop_index('external_knowledge_bindings_dataset_idx')
+
+    op.drop_table('external_knowledge_bindings')
+    with op.batch_alter_table('external_api_templates', schema=None) as batch_op:
+        batch_op.drop_index('external_api_templates_tenant_idx')
+        batch_op.drop_index('external_api_templates_name_idx')
+
+    op.drop_table('external_api_templates')
+    # ### end Alembic commands ###

api/models/dataset.py

@@ -37,7 +37,8 @@ class Dataset(db.Model):
         db.Index("retrieval_model_idx", "retrieval_model", postgresql_using="gin"),
     )
 
-    INDEXING_TECHNIQUE_LIST = ["high_quality", "economy", None]
+    INDEXING_TECHNIQUE_LIST = ['high_quality', 'economy', None]
+    PROVIDER_LIST = ['vendor', 'external', None]
 
     id = db.Column(StringUUID, server_default=db.text("uuid_generate_v4()"))
     tenant_id = db.Column(StringUUID, nullable=False)
@@ -687,3 +688,67 @@ class DatasetPermission(db.Model):
     tenant_id = db.Column(StringUUID, nullable=False)
     has_permission = db.Column(db.Boolean, nullable=False, server_default=db.text("true"))
     created_at = db.Column(db.DateTime, nullable=False, server_default=db.text("CURRENT_TIMESTAMP(0)"))
+
+
+class ExternalApiTemplates(db.Model):
+    __tablename__ = 'external_api_templates'
+    __table_args__ = (
+        db.PrimaryKeyConstraint('id', name='external_api_template_pkey'),
+        db.Index('external_api_templates_tenant_idx', 'tenant_id'),
+        db.Index('external_api_templates_name_idx', 'name'),
+    )
+
+    id = db.Column(StringUUID, nullable=False,
+                   server_default=db.text('uuid_generate_v4()'))
+    name = db.Column(db.String(255), nullable=False)
+    description = db.Column(db.String(255), nullable=False)
+    tenant_id = db.Column(StringUUID, nullable=False)
+    settings = db.Column(db.Text, nullable=True)
+    created_by = db.Column(StringUUID, nullable=False)
+    created_at = db.Column(db.DateTime, nullable=False,
+                           server_default=db.text('CURRENT_TIMESTAMP(0)'))
+    updated_by = db.Column(StringUUID, nullable=True)
+    updated_at = db.Column(db.DateTime, nullable=False,
+                           server_default=db.text('CURRENT_TIMESTAMP(0)'))
+
+    def to_dict(self):
+        return {
+            'id': self.id,
+            'tenant_id': self.tenant_id,
+            'name': self.name,
+            'description': self.description,
+            'settings': self.settings_dict,
+            'created_by': self.created_by,
+            'created_at': self.created_at.isoformat(),
+        }
+
+    @property
+    def settings_dict(self):
+        try:
+            return json.loads(self.settings) if self.settings else None
+        except JSONDecodeError:
+            return None
+
+
+class ExternalKnowledgeBindings(db.Model):
+    __tablename__ = 'external_knowledge_bindings'
+    __table_args__ = (
+        db.PrimaryKeyConstraint('id', name='external_knowledge_bindings_pkey'),
+        db.Index('external_knowledge_bindings_tenant_idx', 'tenant_id'),
+        db.Index('external_knowledge_bindings_dataset_idx', 'dataset_id'),
+        db.Index('external_knowledge_bindings_external_knowledge_idx', 'external_knowledge_id'),
+        db.Index('external_knowledge_bindings_external_api_template_idx', 'external_api_template_id'),
+    )
+
+    id = db.Column(StringUUID, nullable=False,
+                   server_default=db.text('uuid_generate_v4()'))
+    tenant_id = db.Column(StringUUID, nullable=False)
+    external_api_template_id = db.Column(StringUUID, nullable=False)
+    dataset_id = db.Column(StringUUID, nullable=False)
+    external_knowledge_id = db.Column(db.Text, nullable=False)
+    created_by = db.Column(StringUUID, nullable=False)
+    created_at = db.Column(db.DateTime, nullable=False,
+                           server_default=db.text('CURRENT_TIMESTAMP(0)'))
+    updated_by = db.Column(StringUUID, nullable=True)
+    updated_at = db.Column(db.DateTime, nullable=False,
+                           server_default=db.text('CURRENT_TIMESTAMP(0)'))

api/poetry.lock

@@ -2296,18 +2296,18 @@ files = [
 
 [[package]]
 name = "duckduckgo-search"
-version = "6.2.12"
+version = "6.2.11"
 description = "Search for words, documents, images, news, maps and text translation using the DuckDuckGo.com search engine."
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "duckduckgo_search-6.2.12-py3-none-any.whl", hash = "sha256:0d379c1f845b632a41553efb13d571788f19ad289229e641a27b5710d92097a6"},
-    {file = "duckduckgo_search-6.2.12.tar.gz", hash = "sha256:04f9f1459763668d268344c7a32d943173d0e060dad53a5c2df4b4d3ca9a74cf"},
+    {file = "duckduckgo_search-6.2.11-py3-none-any.whl", hash = "sha256:6fb7069b79e8928f487001de6859034ade19201bdcd257ec198802430e374bfe"},
+    {file = "duckduckgo_search-6.2.11.tar.gz", hash = "sha256:6b6ef1b552c5e67f23e252025d2504caf6f9fc14f70e86c6dd512200f386c673"},
 ]
 
 [package.dependencies]
 click = ">=8.1.7"
-primp = ">=0.6.2"
+primp = ">=0.6.1"
 
 [package.extras]
 dev = ["mypy (>=1.11.1)", "pytest (>=8.3.1)", "pytest-asyncio (>=0.23.8)", "ruff (>=0.6.1)"]
@@ -6356,19 +6356,19 @@ dill = ["dill (>=0.3.8)"]
 
 [[package]]
 name = "primp"
-version = "0.6.2"
+version = "0.6.1"
 description = "HTTP client that can impersonate web browsers, mimicking their headers and `TLS/JA3/JA4/HTTP2` fingerprints"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "primp-0.6.2-cp38-abi3-macosx_10_12_x86_64.whl", hash = "sha256:4a35d441462a55d9a9525bf170e2ffd2fcb3db6039b23e802859fa22c18cdd51"},
-    {file = "primp-0.6.2-cp38-abi3-macosx_11_0_arm64.whl", hash = "sha256:f67ccade95bdbca3cf9b96b93aa53f9617d85ddbf988da4e9c523aa785fd2d54"},
-    {file = "primp-0.6.2-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8074b93befaf36567e4cf3d4a1a8cd6ab9cc6e4dd4ff710650678daa405aee71"},
-    {file = "primp-0.6.2-cp38-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:7d3e2a3f8c6262e9b883651b79c4ff2b7677a76f47293a139f541c9ea333ce3b"},
-    {file = "primp-0.6.2-cp38-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:a460ea389371c6d04839b4b50b5805d99da8ebe281a2e8b534d27377c6d44f0e"},
-    {file = "primp-0.6.2-cp38-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:5b6b27e89d3c05c811aff0e4fde7a36d6957b15b3112f4ce28b6b99e8ca1e725"},
-    {file = "primp-0.6.2-cp38-abi3-win_amd64.whl", hash = "sha256:1006a40a85f88a4c5222094813a1ebc01f85a63e9a33d2c443288c0720bed321"},
-    {file = "primp-0.6.2.tar.gz", hash = "sha256:5a96a6b65195a8a989157e67d23bd171c49be238654e02bdf1b1fda36cbcc068"},
+    {file = "primp-0.6.1-cp38-abi3-macosx_10_12_x86_64.whl", hash = "sha256:60cfe95e0bdf154b0f9036d38acaddc9aef02d6723ed125839b01449672d3946"},
+    {file = "primp-0.6.1-cp38-abi3-macosx_11_0_arm64.whl", hash = "sha256:e1e92433ecf32639f9e800bc3a5d58b03792bdec99421b7fb06500e2fae63c85"},
+    {file = "primp-0.6.1-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6e02353f13f07fb5a6f91df9e2f4d8ec9f41312de95088744dce1c9729a3865d"},
+    {file = "primp-0.6.1-cp38-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:c5a2ccfdf488b17be225a529a31e2b22724b2e22fba8e1ae168a222f857c2dc0"},
+    {file = "primp-0.6.1-cp38-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:f335c2ace907800a23bbb7bc6e15acc7fff659b86a2d5858817f6ed79cea07cf"},
+    {file = "primp-0.6.1-cp38-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:5dc15bd9d47ded7bc356fcb5d8321972dcbeba18e7d3b7250e12bb7365447b2b"},
+    {file = "primp-0.6.1-cp38-abi3-win_amd64.whl", hash = "sha256:eebf0412ebba4089547b16b97b765d83f69f1433d811bb02b02cdcdbca20f672"},
+    {file = "primp-0.6.1.tar.gz", hash = "sha256:64b3c12e3d463a887518811c46f3ec37cca02e6af1ddf1287e548342de436301"},
 ]
 
 [package.extras]

api/schedule/clean_unused_messages_task.py

@@ -0,0 +1,92 @@
+import datetime
+import time
+
+import click
+from sqlalchemy import func
+from werkzeug.exceptions import NotFound
+
+import app
+from configs import dify_config
+from core.rag.index_processor.index_processor_factory import IndexProcessorFactory
+from extensions.ext_database import db
+from models.dataset import Dataset, DatasetQuery, Document
+
+
+@app.celery.task(queue="dataset")
+def clean_unused_message_task():
+    click.echo(click.style("Start clean unused messages .", fg="green"))
+    clean_days = int(dify_config.CLEAN_DAY_SETTING)
+    start_at = time.perf_counter()
+    thirty_days_ago = datetime.datetime.now() - datetime.timedelta(days=clean_days)
+    page = 1
+    while True:
+        try:
+            # Subquery for counting new documents
+            document_subquery_new = (
+                db.session.query(Document.dataset_id, func.count(Document.id).label("document_count"))
+                .filter(
+                    Document.indexing_status == "completed",
+                    Document.enabled == True,
+                    Document.archived == False,
+                    Document.updated_at > thirty_days_ago,
+                )
+                .group_by(Document.dataset_id)
+                .subquery()
+            )
+
+            # Subquery for counting old documents
+            document_subquery_old = (
+                db.session.query(Document.dataset_id, func.count(Document.id).label("document_count"))
+                .filter(
+                    Document.indexing_status == "completed",
+                    Document.enabled == True,
+                    Document.archived == False,
+                    Document.updated_at < thirty_days_ago,
+                )
+                .group_by(Document.dataset_id)
+                .subquery()
+            )
+
+            # Main query with join and filter
+            datasets = (
+                db.session.query(Dataset)
+                .outerjoin(document_subquery_new, Dataset.id == document_subquery_new.c.dataset_id)
+                .outerjoin(document_subquery_old, Dataset.id == document_subquery_old.c.dataset_id)
+                .filter(
+                    Dataset.created_at < thirty_days_ago,
+                    func.coalesce(document_subquery_new.c.document_count, 0) == 0,
+                    func.coalesce(document_subquery_old.c.document_count, 0) > 0,
+                )
+                .order_by(Dataset.created_at.desc())
+                .paginate(page=page, per_page=50)
+            )
+
+        except NotFound:
+            break
+        if datasets.items is None or len(datasets.items) == 0:
+            break
+        page += 1
+        for dataset in datasets:
+            dataset_query = (
+                db.session.query(DatasetQuery)
+                .filter(DatasetQuery.created_at > thirty_days_ago, DatasetQuery.dataset_id == dataset.id)
+                .all()
+            )
+            if not dataset_query or len(dataset_query) == 0:
+                try:
+                    # remove index
+                    index_processor = IndexProcessorFactory(dataset.doc_form).init_index_processor()
+                    index_processor.clean(dataset, None)
+
+                    # update document
+                    update_params = {Document.enabled: False}
+
+                    Document.query.filter_by(dataset_id=dataset.id).update(update_params)
+                    db.session.commit()
+                    click.echo(click.style("Cleaned unused dataset {} from db success!".format(dataset.id), fg="green"))
+                except Exception as e:
+                    click.echo(
+                        click.style("clean dataset index error: {} {}".format(e.__class__.__name__, str(e)), fg="red")
+                    )
+    end_at = time.perf_counter()
+    click.echo(click.style("Cleaned unused dataset from db success latency: {}".format(end_at - start_at), fg="green"))

api/services/dataset_service.py

@@ -32,6 +32,7 @@ from models.dataset import (
     DatasetQuery,
     Document,
     DocumentSegment,
+    ExternalKnowledgeBindings,
 )
 from models.model import UploadFile
 from models.source import DataSourceOauthBinding
@@ -39,6 +40,7 @@ from services.errors.account import NoPermissionError
 from services.errors.dataset import DatasetNameDuplicateError
 from services.errors.document import DocumentIndexingError
 from services.errors.file import FileNotExistsError
+from services.external_knowledge_service import ExternalDatasetService
 from services.feature_service import FeatureModel, FeatureService
 from services.tag_service import TagService
 from services.vector_service import VectorService
@@ -137,7 +139,14 @@ class DatasetService:
 
     @staticmethod
     def create_empty_dataset(
-        tenant_id: str, name: str, indexing_technique: Optional[str], account: Account, permission: Optional[str] = None
+        tenant_id: str,
+        name: str,
+        indexing_technique: Optional[str],
+        account: Account,
+        permission: Optional[str] = None,
+        provider: str = "vendor",
+        external_api_template_id: Optional[str] = None,
+        external_knowledge_id: Optional[str] = None,
     ):
         # check if dataset name already exists
         if Dataset.query.filter_by(name=name, tenant_id=tenant_id).first():
@@ -156,7 +165,23 @@ class DatasetService:
         dataset.embedding_model_provider = embedding_model.provider if embedding_model else None
         dataset.embedding_model = embedding_model.model if embedding_model else None
         dataset.permission = permission or DatasetPermissionEnum.ONLY_ME
+        dataset.provider = provider
         db.session.add(dataset)
+        db.session.flush()
+
+        if provider == "external" and external_api_template_id:
+            external_api_template = ExternalDatasetService.get_api_template(external_api_template_id)
+            if not external_api_template:
+                raise ValueError("External API template not found.")
+            external_knowledge_binding = ExternalKnowledgeBindings(
+                tenant_id=tenant_id,
+                dataset_id=dataset.id,
+                external_api_template_id=external_api_template_id,
+                external_knowledge_id=external_knowledge_id,
+                created_by=account.id,
+            )
+            db.session.add(external_knowledge_binding)
+
         db.session.commit()
         return dataset
 

api/services/entities/external_knowledge_entities/external_knowledge_entities.py

@@ -0,0 +1,28 @@
+from typing import Literal, Optional, Union
+
+from pydantic import BaseModel
+
+
+class AuthorizationConfig(BaseModel):
+    type: Literal[None, "basic", "bearer", "custom"]
+    api_key: Union[None, str] = None
+    header: Union[None, str] = None
+
+
+class Authorization(BaseModel):
+    type: Literal["no-auth", "api-key"]
+    config: Optional[AuthorizationConfig] = None
+
+
+class ProcessStatusSetting(BaseModel):
+    request_method: str
+    url: str
+
+
+class ApiTemplateSetting(BaseModel):
+    method: str
+    url: str
+    request_method: str
+    api_token: str
+    headers: Optional[dict] = None
+    params: Optional[dict] = None

api/services/external_knowledge_service.py

@@ -0,0 +1,308 @@
+import json
+import random
+import time
+from copy import deepcopy
+from datetime import datetime, timezone
+from typing import Any, Optional, Union
+
+import httpx
+
+from core.helper import ssrf_proxy
+from extensions.ext_database import db
+from models.dataset import (
+    Dataset,
+    Document,
+    ExternalApiTemplates,
+    ExternalKnowledgeBindings,
+)
+from models.model import UploadFile
+from services.entities.external_knowledge_entities.external_knowledge_entities import ApiTemplateSetting, Authorization
+from services.errors.dataset import DatasetNameDuplicateError
+# from tasks.external_document_indexing_task import external_document_indexing_task
+
+
+class ExternalDatasetService:
+    @staticmethod
+    def get_external_api_templates(page, per_page, tenant_id, search=None) -> tuple[list[ExternalApiTemplates], int]:
+        query = ExternalApiTemplates.query.filter(ExternalApiTemplates.tenant_id == tenant_id).order_by(
+            ExternalApiTemplates.created_at.desc()
+        )
+        if search:
+            query = query.filter(ExternalApiTemplates.name.ilike(f"%{search}%"))
+
+        api_templates = query.paginate(page=page, per_page=per_page, max_per_page=100, error_out=False)
+
+        return api_templates.items, api_templates.total
+
+    @classmethod
+    def validate_api_list(cls, api_settings: dict):
+        if not api_settings:
+            raise ValueError("api list is empty")
+        if "endpoint" not in api_settings and not api_settings["endpoint"]:
+            raise ValueError("endpoint is required")
+        if "api_key" not in api_settings and not api_settings["api_key"]:
+            raise ValueError("api_key is required")
+
+    @staticmethod
+    def create_api_template(tenant_id: str, user_id: str, args: dict) -> ExternalApiTemplates:
+        api_template = ExternalApiTemplates(
+            tenant_id=tenant_id,
+            created_by=user_id,
+            updated_by=user_id,
+            name=args.get("name"),
+            description=args.get("description", ""),
+            settings=json.dumps(args.get("settings"), ensure_ascii=False),
+        )
+
+        db.session.add(api_template)
+        db.session.commit()
+        return api_template
+
+    @staticmethod
+    def get_api_template(api_template_id: str) -> ExternalApiTemplates:
+        return ExternalApiTemplates.query.filter_by(id=api_template_id).first()
+
+    @staticmethod
+    def update_api_template(tenant_id, user_id, api_template_id, args) -> ExternalApiTemplates:
+        api_template = ExternalApiTemplates.query.filter_by(id=api_template_id, tenant_id=tenant_id).first()
+        if api_template is None:
+            raise ValueError("api template not found")
+
+        api_template.name = args.get("name")
+        api_template.description = args.get("description", "")
+        api_template.settings = json.dumps(args.get("settings"), ensure_ascii=False)
+        api_template.updated_by = user_id
+        api_template.updated_at = datetime.now(timezone.utc).replace(tzinfo=None)
+        db.session.commit()
+
+        return api_template
+
+    @staticmethod
+    def delete_api_template(tenant_id: str, api_template_id: str):
+        api_template = ExternalApiTemplates.query.filter_by(id=api_template_id, tenant_id=tenant_id).first()
+        if api_template is None:
+            raise ValueError("api template not found")
+
+        db.session.delete(api_template)
+        db.session.commit()
+
+    @staticmethod
+    def external_api_template_use_check(api_template_id: str) -> bool:
+        count = ExternalKnowledgeBindings.query.filter_by(external_api_template_id=api_template_id).count()
+        if count > 0:
+            return True
+        return False
+
+    @staticmethod
+    def get_external_knowledge_binding_with_dataset_id(tenant_id: str, dataset_id: str) -> ExternalKnowledgeBindings:
+        external_knowledge_binding = ExternalKnowledgeBindings.query.filter_by(
+            dataset_id=dataset_id, tenant_id=tenant_id
+        ).first()
+        if not external_knowledge_binding:
+            raise ValueError("external knowledge binding not found")
+        return external_knowledge_binding
+
+    @staticmethod
+    def document_create_args_validate(tenant_id: str, api_template_id: str, process_parameter: dict):
+        api_template = ExternalApiTemplates.query.filter_by(id=api_template_id, tenant_id=tenant_id).first()
+        if api_template is None:
+            raise ValueError("api template not found")
+        settings = json.loads(api_template.settings)
+        for settings in settings:
+            custom_parameters = settings.get("document_process_setting")
+            if custom_parameters:
+                for parameter in custom_parameters:
+                    if parameter.get("required", False) and not process_parameter.get(parameter.get("name")):
+                        raise ValueError(f'{parameter.get("name")} is required')
+
+    @staticmethod
+    def init_external_dataset(tenant_id: str, user_id: str, args: dict, created_from: str = "web"):
+        api_template_id = args.get("api_template_id")
+
+        data_source = args.get("data_source")
+        if data_source is None:
+            raise ValueError("data source is required")
+
+        process_parameter = args.get("process_parameter")
+        api_template = ExternalApiTemplates.query.filter_by(id=api_template_id, tenant_id=tenant_id).first()
+        if api_template is None:
+            raise ValueError("api template not found")
+
+        dataset = Dataset(
+            tenant_id=tenant_id,
+            name=args.get("name"),
+            description=args.get("description", ""),
+            provider="external",
+            created_by=user_id,
+        )
+
+        db.session.add(dataset)
+        db.session.flush()
+
+        document = Document.query.filter_by(dataset_id=dataset.id).order_by(Document.position.desc()).first()
+
+        position = document.position + 1 if document else 1
+
+        batch = time.strftime("%Y%m%d%H%M%S") + str(random.randint(100000, 999999))
+        document_ids = []
+        if data_source["type"] == "upload_file":
+            upload_file_list = data_source["info_list"]["file_info_list"]["file_ids"]
+            for file_id in upload_file_list:
+                file = (
+                    db.session.query(UploadFile)
+                    .filter(UploadFile.tenant_id == dataset.tenant_id, UploadFile.id == file_id)
+                    .first()
+                )
+                if file:
+                    data_source_info = {
+                        "upload_file_id": file_id,
+                    }
+                    document = Document(
+                        tenant_id=dataset.tenant_id,
+                        dataset_id=dataset.id,
+                        position=position,
+                        data_source_type=data_source["type"],
+                        data_source_info=json.dumps(data_source_info),
+                        batch=batch,
+                        name=file.name,
+                        created_from=created_from,
+                        created_by=user_id,
+                    )
+                    position += 1
+                    db.session.add(document)
+                    db.session.flush()
+                    document_ids.append(document.id)
+        db.session.commit()
+        #external_document_indexing_task.delay(dataset.id, api_template_id, data_source, process_parameter)
+
+        return dataset
+
+    @staticmethod
+    def process_external_api(settings: ApiTemplateSetting, files: Union[None, dict[str, Any]]) -> httpx.Response:
+        """
+        do http request depending on api bundle
+        """
+
+        kwargs = {
+            "url": settings.url,
+            "headers": settings.headers,
+            "follow_redirects": True,
+        }
+
+        response = getattr(ssrf_proxy, settings.request_method)(data=settings.params, files=files, **kwargs)
+
+        return response
+
+    @staticmethod
+    def assembling_headers(authorization: Authorization, headers: Optional[dict] = None) -> dict[str, Any]:
+        authorization = deepcopy(authorization)
+        if headers:
+            headers = deepcopy(headers)
+        else:
+            headers = {}
+        if authorization.type == "api-key":
+            if authorization.config is None:
+                raise ValueError("authorization config is required")
+
+            if authorization.config.api_key is None:
+                raise ValueError("api_key is required")
+
+            if not authorization.config.header:
+                authorization.config.header = "Authorization"
+
+            if authorization.config.type == "bearer":
+                headers[authorization.config.header] = f"Bearer {authorization.config.api_key}"
+            elif authorization.config.type == "basic":
+                headers[authorization.config.header] = f"Basic {authorization.config.api_key}"
+            elif authorization.config.type == "custom":
+                headers[authorization.config.header] = authorization.config.api_key
+
+        return headers
+
+    @staticmethod
+    def get_api_template_settings(settings: dict) -> ApiTemplateSetting:
+        return ApiTemplateSetting.parse_obj(settings)
+
+    @staticmethod
+    def create_external_dataset(tenant_id: str, user_id: str, args: dict) -> Dataset:
+        # check if dataset name already exists
+        if Dataset.query.filter_by(name=args.get("name"), tenant_id=tenant_id).first():
+            raise DatasetNameDuplicateError(f"Dataset with name {args.get('name')} already exists.")
+        api_template = ExternalApiTemplates.query.filter_by(
+            id=args.get("external_api_template_id"), tenant_id=tenant_id
+        ).first()
+
+        if api_template is None:
+            raise ValueError("api template not found")
+
+        dataset = Dataset(
+            tenant_id=tenant_id,
+            name=args.get("name"),
+            description=args.get("description", ""),
+            provider="external",
+            created_by=user_id,
+        )
+
+        db.session.add(dataset)
+        db.session.flush()
+
+        external_knowledge_binding = ExternalKnowledgeBindings(
+            tenant_id=tenant_id,
+            dataset_id=dataset.id,
+            external_api_template_id=args.get("external_api_template_id"),
+            external_knowledge_id=args.get("external_knowledge_id"),
+            created_by=user_id,
+        )
+        db.session.add(external_knowledge_binding)
+
+        db.session.commit()
+
+        return dataset
+
+    @staticmethod
+    def fetch_external_knowledge_retrival(
+        tenant_id: str, dataset_id: str, query: str, external_retrival_parameters: dict
+    ):
+        external_knowledge_binding = ExternalKnowledgeBindings.query.filter_by(
+            dataset_id=dataset_id, tenant_id=tenant_id
+        ).first()
+        if not external_knowledge_binding:
+            raise ValueError("external knowledge binding not found")
+
+        external_api_template = ExternalApiTemplates.query.filter_by(
+            id=external_knowledge_binding.external_api_template_id
+        ).first()
+        if not external_api_template:
+            raise ValueError("external api template not found")
+
+        settings = json.loads(external_api_template.settings)
+        headers = {}
+        if settings.get("api_token"):
+            headers["Authorization"] = f"Bearer {settings.get('api_token')}"
+
+        external_retrival_parameters["query"] = query
+
+        api_template_setting = {
+            "url": f"{settings.get('endpoint')}/dify/external-knowledge/retrival-documents",
+            "request_method": "post",
+            "headers": settings.get("headers"),
+            "params": external_retrival_parameters,
+        }
+        response = ExternalDatasetService.process_external_api(ApiTemplateSetting(**api_template_setting), None)
+
+
+    @staticmethod
+    def test_external_knowledge_retrival(
+        top_k: int, score_threshold: float
+    ):
+        api_template_setting = {
+            "url": f"{settings.get('endpoint')}/dify/external-knowledge/retrival-documents",
+            "request_method": "post",
+            "headers": settings.get("headers"),
+            "params": {
+                "top_k": top_k,
+                "score_threshold": score_threshold,
+            },
+        }
+        response = ExternalDatasetService.process_external_api(ApiTemplateSetting(**api_template_setting), None)
+        return response.json()

api/services/hit_testing_service.py

@@ -19,7 +19,15 @@ default_retrieval_model = {
 
 class HitTestingService:
     @classmethod
-    def retrieve(cls, dataset: Dataset, query: str, account: Account, retrieval_model: dict, limit: int = 10) -> dict:
+    def retrieve(
+        cls,
+        dataset: Dataset,
+        query: str,
+        account: Account,
+        retrieval_model: dict,
+        external_retrieval_model: dict,
+        limit: int = 10,
+    ) -> dict:
         if dataset.available_document_count == 0 or dataset.available_segment_count == 0:
             return {
                 "query": {
@@ -48,6 +56,8 @@ class HitTestingService:
             else None,
             reranking_mode=retrieval_model.get("reranking_mode") or "reranking_model",
             weights=retrieval_model.get("weights", None),
+            provider=dataset.provider,
+            external_retrieval_model=external_retrieval_model,
         )
 
         end = time.perf_counter()

api/tasks/external_document_indexing_task.py

@@ -0,0 +1,85 @@
+import json
+import logging
+import time
+
+import click
+from celery import shared_task
+
+from core.indexing_runner import DocumentIsPausedException
+from extensions.ext_database import db
+from extensions.ext_storage import storage
+from models.dataset import Dataset, ExternalApiTemplates
+from models.model import UploadFile
+from services.external_knowledge_service import ExternalDatasetService
+
+
+@shared_task(queue="dataset")
+def external_document_indexing_task(dataset_id: str, api_template_id: str, data_source: dict, process_parameter: dict):
+    """
+    Async process document
+    :param dataset_id:
+    :param api_template_id:
+    :param data_source:
+    :param process_parameter:
+    Usage: external_document_indexing_task.delay(dataset_id, document_id)
+    """
+    start_at = time.perf_counter()
+
+    dataset = db.session.query(Dataset).filter(Dataset.id == dataset_id).first()
+    if not dataset:
+        logging.info(
+            click.style("Processed external dataset: {} failed, dataset not exit.".format(dataset_id), fg="red")
+        )
+        return
+
+    # get external api template
+    api_template = (
+        db.session.query(ExternalApiTemplates)
+        .filter(ExternalApiTemplates.id == api_template_id, ExternalApiTemplates.tenant_id == dataset.tenant_id)
+        .first()
+    )
+
+    if not api_template:
+        logging.info(
+            click.style(
+                "Processed external dataset: {} failed, api template: {} not exit.".format(dataset_id, api_template_id),
+                fg="red",
+            )
+        )
+        return
+    files = {}
+    if data_source["type"] == "upload_file":
+        upload_file_list = data_source["info_list"]["file_info_list"]["file_ids"]
+        for file_id in upload_file_list:
+            file = (
+                db.session.query(UploadFile)
+                .filter(UploadFile.tenant_id == dataset.tenant_id, UploadFile.id == file_id)
+                .first()
+            )
+            if file:
+                files[file.id] = (file.name, storage.load_once(file.key), file.mime_type)
+    try:
+        settings = ExternalDatasetService.get_api_template_settings(json.loads(api_template.settings))
+        # assemble headers
+        headers = ExternalDatasetService.assembling_headers(settings.authorization, settings.headers)
+
+        # do http request
+        response = ExternalDatasetService.process_external_api(settings, headers, process_parameter, files)
+        job_id = response.json().get("job_id")
+        if job_id:
+            # save job_id to dataset
+            dataset.job_id = job_id
+            db.session.commit()
+
+        end_at = time.perf_counter()
+        logging.info(
+            click.style(
+                "Processed external dataset: {} successful, latency: {}".format(dataset.id, end_at - start_at),
+                fg="green",
+            )
+        )
+    except DocumentIsPausedException as ex:
+        logging.info(click.style(str(ex), fg="yellow"))
+
+    except Exception:
+        pass

docker-legacy/docker-compose.yaml

@@ -2,7 +2,7 @@ version: '3'
 services:
   # API service
   api:
-    image: langgenius/dify-api:0.8.3
+    image: langgenius/dify-api:0.8.2
     restart: always
     environment:
       # Startup mode, 'api' starts the API server.
@@ -227,7 +227,7 @@ services:
   # worker service
   # The Celery worker for processing the queue.
   worker:
-    image: langgenius/dify-api:0.8.3
+    image: langgenius/dify-api:0.8.2
     restart: always
     environment:
       CONSOLE_WEB_URL: ''
@@ -396,7 +396,7 @@ services:
 
   # Frontend web application.
   web:
-    image: langgenius/dify-web:0.8.3
+    image: langgenius/dify-web:0.8.2
     restart: always
     environment:
       # The base URL of console application api server, refers to the Console base URL of WEB service if console domain is

docker/docker-compose.yaml

@@ -208,7 +208,7 @@ x-shared-env: &shared-api-worker-env
 services:
   # API service
   api:
-    image: langgenius/dify-api:0.8.3
+    image: langgenius/dify-api:0.8.2
     restart: always
     environment:
       # Use the shared environment variables.
@@ -228,7 +228,7 @@ services:
   # worker service
   # The Celery worker for processing the queue.
   worker:
-    image: langgenius/dify-api:0.8.3
+    image: langgenius/dify-api:0.8.2
     restart: always
     environment:
       # Use the shared environment variables.
@@ -247,7 +247,7 @@ services:
 
   # Frontend web application.
   web:
-    image: langgenius/dify-web:0.8.3
+    image: langgenius/dify-web:0.8.2
     restart: always
     environment:
       CONSOLE_API_URL: ${CONSOLE_API_URL:-}

docker/unstructured.yaml

@@ -0,0 +1,24 @@
+
+  # unstructured .
+  # (if used, you need to set ETL_TYPE to Unstructured in the api & worker service.)
+  unstructured:
+    image: downloads.unstructured.io/unstructured-io/unstructured-api:latest
+    profiles:
+      - unstructured
+    restart: always
+    volumes:
+      - ./volumes/unstructured:/app/data
+
+networks:
+  # create a network between sandbox, api and ssrf_proxy, and can not access outside.
+  ssrf_proxy_network:
+    driver: bridge
+    internal: true
+  milvus:
+    driver: bridge
+  opensearch-net:
+    driver: bridge
+    internal: true
+
+volumes:
+  oradata:

web/.husky/pre-commit

@@ -51,32 +51,5 @@ if $web_modified; then
     echo "Running ESLint on web module"
     cd ./web || exit 1
     npx lint-staged
-
-    echo "Running unit tests check"
-    modified_files=$(git diff --cached --name-only -- utils | grep -v '\.spec\.ts$' || true)
-
-    if [ -n "$modified_files" ]; then
-        for file in $modified_files; do
-            test_file="${file%.*}.spec.ts"
-            echo "Checking for test file: $test_file"
-
-            # check if the test file exists
-            if [ -f "../$test_file" ]; then
-                echo "Detected changes in $file, running corresponding unit tests..."
-                npm run test "../$test_file"
-
-                if [ $? -ne 0 ]; then
-                    echo "Unit tests failed. Please fix the errors before committing."
-                    exit 1
-                fi
-                echo "Unit tests for $file passed."
-            else
-                echo "Warning: $file does not have a corresponding test file."
-            fi
-
-        done
-        echo "All unit tests for modified web/utils files have passed."
-    fi
-
     cd ../
 fi

web/README.md

@@ -18,10 +18,6 @@ yarn install --frozen-lockfile
 
 Then, configure the environment variables. Create a file named `.env.local` in the current directory and copy the contents from `.env.example`. Modify the values of these environment variables according to your requirements:
 
-```bash
-cp .env.example .env.local
-```
-
 ```
 # For production release, change this to PRODUCTION
 NEXT_PUBLIC_DEPLOY_ENV=DEVELOPMENT
@@ -82,7 +78,7 @@ If your IDE is VSCode, rename `web/.vscode/settings.example.json` to `web/.vscod
 
 We start to use [Jest](https://jestjs.io/) and [React Testing Library](https://testing-library.com/docs/react-testing-library/intro/) for Unit Testing.
 
-You can create a test file with a suffix of `.spec` beside the file that to be tested. For example, if you want to test a file named `util.ts`. The test file name should be `util.spec.ts`.
+You can create a test file with a suffix of `.spec` beside the file that to be tested. For example, if you want to test a file named `util.ts`. The test file name should be `util.spec.ts`. 
 
 Run test:
 

web/app/(commonLayout)/app/(appDetailLayout)/[appId]/layout.tsx

@@ -109,11 +109,6 @@ const AppDetailLayout: FC<IAppDetailLayoutProps> = (props) => {
     setAppDetail()
     fetchAppDetail({ url: '/apps', id: appId }).then((res) => {
       // redirection
-      const canIEditApp = isCurrentWorkspaceEditor
-      if (!canIEditApp && (pathname.endsWith('configuration') || pathname.endsWith('workflow') || pathname.endsWith('logs'))) {
-        router.replace(`/app/${appId}/overview`)
-        return
-      }
       if ((res.mode === 'workflow' || res.mode === 'advanced-chat') && (pathname).endsWith('configuration')) {
         router.replace(`/app/${appId}/workflow`)
       }
@@ -123,7 +118,7 @@ const AppDetailLayout: FC<IAppDetailLayoutProps> = (props) => {
       else {
         setAppDetail({ ...res, enable_sso: false })
         setNavigation(getNavigations(appId, isCurrentWorkspaceEditor, res.mode))
-        if (systemFeatures.enable_web_sso_switch_component && canIEditApp) {
+        if (systemFeatures.enable_web_sso_switch_component) {
           fetchAppSSO({ appId }).then((ssoRes) => {
             setAppDetail({ ...res, enable_sso: ssoRes.enabled })
           })
@@ -133,7 +128,7 @@ const AppDetailLayout: FC<IAppDetailLayoutProps> = (props) => {
       if (e.status === 404)
         router.replace('/apps')
     })
-  }, [appId, isCurrentWorkspaceEditor, systemFeatures, getNavigations, pathname, router, setAppDetail])
+  }, [appId, isCurrentWorkspaceEditor, systemFeatures])
 
   useUnmount(() => {
     setAppDetail()

web/app/components/app/overview/settings/index.tsx

@@ -16,7 +16,7 @@ import type { AppIconType, AppSSO, Language } from '@/types/app'
 import { useToastContext } from '@/app/components/base/toast'
 import { languages } from '@/i18n/language'
 import Tooltip from '@/app/components/base/tooltip'
-import AppContext, { useAppContext } from '@/context/app-context'
+import AppContext from '@/context/app-context'
 import type { AppIconSelection } from '@/app/components/base/app-icon-picker'
 import AppIconPicker from '@/app/components/base/app-icon-picker'
 
@@ -57,7 +57,6 @@ const SettingsModal: FC<ISettingsModalProps> = ({
   onSave,
 }) => {
   const systemFeatures = useContextSelector(AppContext, state => state.systemFeatures)
-  const { isCurrentWorkspaceEditor } = useAppContext()
   const { notify } = useToastContext()
   const [isShowMore, setIsShowMore] = useState(false)
   const {
@@ -266,7 +265,7 @@ const SettingsModal: FC<ISettingsModalProps> = ({
               }
               asChild={false}
             >
-              <Switch disabled={!systemFeatures.sso_enforced_for_web || !isCurrentWorkspaceEditor} defaultValue={systemFeatures.sso_enforced_for_web && inputInfo.enable_sso} onChange={v => setInputInfo({ ...inputInfo, enable_sso: v })}></Switch>
+              <Switch disabled={!systemFeatures.sso_enforced_for_web} defaultValue={systemFeatures.sso_enforced_for_web && inputInfo.enable_sso} onChange={v => setInputInfo({ ...inputInfo, enable_sso: v })}></Switch>
             </Tooltip>
           </div>
           <p className='body-xs-regular text-gray-500'>{t(`${prefixSettings}.sso.description`)}</p>

web/app/components/datasets/create/step-two/escape.ts

@@ -1,18 +0,0 @@
-function escape(input: string): string {
-  if (!input || typeof input !== 'string')
-    return ''
-
-  const res = input
-    .replaceAll('\\', '\\\\')
-    .replaceAll('\0', '\\0')
-    .replaceAll('\b', '\\b')
-    .replaceAll('\f', '\\f')
-    .replaceAll('\n', '\\n')
-    .replaceAll('\r', '\\r')
-    .replaceAll('\t', '\\t')
-    .replaceAll('\v', '\\v')
-    .replaceAll('\'', '\\\'')
-  return res
-}
-
-export default escape

web/app/components/datasets/create/step-two/index.tsx

@@ -1,5 +1,5 @@
 'use client'
-import React, { useCallback, useEffect, useLayoutEffect, useRef, useState } from 'react'
+import React, { useEffect, useLayoutEffect, useRef, useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import { useContext } from 'use-context-selector'
 import { useBoolean } from 'ahooks'
@@ -13,8 +13,6 @@ import { groupBy } from 'lodash-es'
 import PreviewItem, { PreviewType } from './preview-item'
 import LanguageSelect from './language-select'
 import s from './index.module.css'
-import unescape from './unescape'
-import escape from './escape'
 import cn from '@/utils/classnames'
 import type { CrawlOptions, CrawlResultItem, CreateDocumentReq, CustomFile, FileIndexingEstimateResponse, FullDocumentDetail, IndexingEstimateParams, NotionInfo, PreProcessingRule, ProcessRule, Rules, createDocumentResponse } from '@/models/datasets'
 import {
@@ -80,8 +78,6 @@ enum IndexingType {
   ECONOMICAL = 'economy',
 }
 
-const DEFAULT_SEGMENT_IDENTIFIER = '\\n\\n'
-
 const StepTwo = ({
   isSetting,
   documentDetail,
@@ -114,11 +110,8 @@ const StepTwo = ({
   const previewScrollRef = useRef<HTMLDivElement>(null)
   const [previewScrolled, setPreviewScrolled] = useState(false)
   const [segmentationType, setSegmentationType] = useState<SegmentType>(SegmentType.AUTO)
-  const [segmentIdentifier, doSetSegmentIdentifier] = useState(DEFAULT_SEGMENT_IDENTIFIER)
-  const setSegmentIdentifier = useCallback((value: string) => {
-    doSetSegmentIdentifier(value ? escape(value) : DEFAULT_SEGMENT_IDENTIFIER)
-  }, [])
-  const [max, setMax] = useState(4000) // default chunk length
+  const [segmentIdentifier, setSegmentIdentifier] = useState('\\n')
+  const [max, setMax] = useState(5000) // default chunk length
   const [overlap, setOverlap] = useState(50)
   const [rules, setRules] = useState<PreProcessingRule[]>([])
   const [defaultConfig, setDefaultConfig] = useState<Rules>()
@@ -190,7 +183,7 @@ const StepTwo = ({
   }
   const resetRules = () => {
     if (defaultConfig) {
-      setSegmentIdentifier(defaultConfig.segmentation.separator)
+      setSegmentIdentifier((defaultConfig.segmentation.separator === '\n' ? '\\n' : defaultConfig.segmentation.separator) || '\\n')
       setMax(defaultConfig.segmentation.max_tokens)
       setOverlap(defaultConfig.segmentation.chunk_overlap)
       setRules(defaultConfig.pre_processing_rules)
@@ -224,7 +217,7 @@ const StepTwo = ({
       const ruleObj = {
         pre_processing_rules: rules,
         segmentation: {
-          separator: unescape(segmentIdentifier),
+          separator: segmentIdentifier === '\\n' ? '\n' : segmentIdentifier,
           max_tokens: max,
           chunk_overlap: overlap,
         },
@@ -401,7 +394,7 @@ const StepTwo = ({
     try {
       const res = await fetchDefaultProcessRule({ url: '/datasets/process-rule' })
       const separator = res.rules.segmentation.separator
-      setSegmentIdentifier(separator)
+      setSegmentIdentifier((separator === '\n' ? '\\n' : separator) || '\\n')
       setMax(res.rules.segmentation.max_tokens)
       setOverlap(res.rules.segmentation.chunk_overlap)
       setRules(res.rules.pre_processing_rules)
@@ -418,7 +411,7 @@ const StepTwo = ({
       const separator = rules.segmentation.separator
       const max = rules.segmentation.max_tokens
       const overlap = rules.segmentation.chunk_overlap
-      setSegmentIdentifier(separator)
+      setSegmentIdentifier((separator === '\n' ? '\\n' : separator) || '\\n')
       setMax(max)
       setOverlap(overlap)
       setRules(rules.pre_processing_rules)
@@ -623,22 +616,12 @@ const StepTwo = ({
                 <div className={s.typeFormBody}>
                   <div className={s.formRow}>
                     <div className='w-full'>
-                      <div className={s.label}>
-                        {t('datasetCreation.stepTwo.separator')}
-                        <Tooltip
-                          popupContent={
-                            <div className='max-w-[200px]'>
-                              {t('datasetCreation.stepTwo.separatorTip')}
-                            </div>
-                          }
-                        />
-                      </div>
+                      <div className={s.label}>{t('datasetCreation.stepTwo.separator')}</div>
                       <input
                         type="text"
                         className={s.input}
-                        placeholder={t('datasetCreation.stepTwo.separatorPlaceholder') || ''}
-                        value={segmentIdentifier}
-                        onChange={e => doSetSegmentIdentifier(e.target.value)}
+                        placeholder={t('datasetCreation.stepTwo.separatorPlaceholder') || ''} value={segmentIdentifier}
+                        onChange={e => setSegmentIdentifier(e.target.value)}
                       />
                     </div>
                   </div>

web/app/components/datasets/create/step-two/unescape.ts

@@ -1,54 +0,0 @@
-// https://github.com/iamakulov/unescape-js/blob/master/src/index.js
-
-/**
- * \\ - matches the backslash which indicates the beginning of an escape sequence
- * (
- *   u\{([0-9A-Fa-f]+)\} - first alternative; matches the variable-length hexadecimal escape sequence (\u{ABCD0})
- * |
- *   u([0-9A-Fa-f]{4}) - second alternative; matches the 4-digit hexadecimal escape sequence (\uABCD)
- * |
- *   x([0-9A-Fa-f]{2}) - third alternative; matches the 2-digit hexadecimal escape sequence (\xA5)
- * |
- *   ([1-7][0-7]{0,2}|[0-7]{2,3}) - fourth alternative; matches the up-to-3-digit octal escape sequence (\5 or \512)
- * |
- *   (['"tbrnfv0\\]) - fifth alternative; matches the special escape characters (\t, \n and so on)
- * |
- *   \U([0-9A-Fa-f]+) - sixth alternative; matches the 8-digit hexadecimal escape sequence used by python (\U0001F3B5)
- * )
- */
-const jsEscapeRegex = /\\(u\{([0-9A-Fa-f]+)\}|u([0-9A-Fa-f]{4})|x([0-9A-Fa-f]{2})|([1-7][0-7]{0,2}|[0-7]{2,3})|(['"tbrnfv0\\]))|\\U([0-9A-Fa-f]{8})/g
-
-const usualEscapeSequences: Record<string, string> = {
-  '0': '\0',
-  'b': '\b',
-  'f': '\f',
-  'n': '\n',
-  'r': '\r',
-  't': '\t',
-  'v': '\v',
-  '\'': '\'',
-  '"': '"',
-  '\\': '\\',
-}
-
-const fromHex = (str: string) => String.fromCodePoint(parseInt(str, 16))
-const fromOct = (str: string) => String.fromCodePoint(parseInt(str, 8))
-
-const unescape = (str: string) => {
-  return str.replace(jsEscapeRegex, (_, __, varHex, longHex, shortHex, octal, specialCharacter, python) => {
-    if (varHex !== undefined)
-      return fromHex(varHex)
-    else if (longHex !== undefined)
-      return fromHex(longHex)
-    else if (shortHex !== undefined)
-      return fromHex(shortHex)
-    else if (octal !== undefined)
-      return fromOct(octal)
-    else if (python !== undefined)
-      return fromHex(python)
-    else
-      return usualEscapeSequences[specialCharacter]
-  })
-}
-
-export default unescape

web/i18n/en-US/dataset-creation.ts

@@ -87,8 +87,7 @@ const translation = {
     custom: 'Custom',
     customDescription: 'Customize chunks rules, chunks length, and preprocessing rules, etc.',
     separator: 'Delimiter',
-    separatorTip: 'A delimiter is the character used to separate text. \\n\\n and \\n are commonly used delimiters for separating paragraphs and lines. Combined with commas (\\n\\n,\\n), paragraphs will be segmented by lines when exceeding the maximum chunk length. You can also use special delimiters defined by yourself (e.g. ***).',
-    separatorPlaceholder: '\\n\\n for separating paragraphs; \\n for separating lines',
+    separatorPlaceholder: 'For example, newline (\\\\n) or special separator (such as "***")',
     maxLength: 'Maximum chunk length',
     overlap: 'Chunk overlap',
     overlapTip: 'Setting the chunk overlap can maintain the semantic relevance between them, enhancing the retrieve effect. It is recommended to set 10%-25% of the maximum chunk size.',

web/i18n/zh-Hans/dataset-creation.ts

@@ -87,8 +87,7 @@ const translation = {
     custom: '自定义',
     customDescription: '自定义分段规则、分段长度以及预处理规则等参数',
     separator: '分段标识符',
-    separatorTip: '分隔符是用于分隔文本的字符。\\n\\n 和 \\n 是常用于分隔段落和行的分隔符。用逗号连接分隔符（\\n\\n,\\n），当段落超过最大块长度时，会按行进行分割。你也可以使用自定义的特殊分隔符（例如 ***）。',
-    separatorPlaceholder: '\\n\\n 用于分段；\\n 用于分行',
+    separatorPlaceholder: '例如换行符（\n）或特定的分隔符（如 "***"）',
     maxLength: '分段最大长度',
     overlap: '分段重叠长度',
     overlapTip: '设置分段之间的重叠长度可以保留分段之间的语义关系，提升召回效果。建议设置为最大分段长度的10%-25%',

web/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dify-web",
-  "version": "0.8.3",
+  "version": "0.8.2",
   "private": true,
   "engines": {
     "node": ">=18.17.0"
@@ -37,7 +37,6 @@
     "@remixicon/react": "^4.2.0",
     "@sentry/react": "^7.54.0",
     "@sentry/utils": "^7.54.0",
-    "@svgdotjs/svg.js": "^3.2.4",
     "@tailwindcss/line-clamp": "^0.4.4",
     "@tailwindcss/typography": "^0.5.9",
     "ahooks": "^3.7.5",
@@ -45,6 +44,7 @@
     "classnames": "^2.3.2",
     "copy-to-clipboard": "^3.3.3",
     "crypto-js": "^4.2.0",
+    "@svgdotjs/svg.js": "^3.2.4",
     "dayjs": "^1.11.7",
     "echarts": "^5.4.1",
     "echarts-for-react": "^3.0.2",

web/utils/format.spec.ts

@@ -1,61 +0,0 @@
-import { formatFileSize, formatNumber, formatTime } from './format'
-describe('formatNumber', () => {
-  test('should correctly format integers', () => {
-    expect(formatNumber(1234567)).toBe('1,234,567')
-  })
-  test('should correctly format decimals', () => {
-    expect(formatNumber(1234567.89)).toBe('1,234,567.89')
-  })
-  test('should correctly handle string input', () => {
-    expect(formatNumber('1234567')).toBe('1,234,567')
-  })
-  test('should correctly handle zero', () => {
-    expect(formatNumber(0)).toBe(0)
-  })
-  test('should correctly handle negative numbers', () => {
-    expect(formatNumber(-1234567)).toBe('-1,234,567')
-  })
-  test('should correctly handle empty input', () => {
-    expect(formatNumber('')).toBe('')
-  })
-})
-describe('formatFileSize', () => {
-  test('should return the input if it is falsy', () => {
-    expect(formatFileSize(0)).toBe(0)
-  })
-  test('should format bytes correctly', () => {
-    expect(formatFileSize(500)).toBe('500.00B')
-  })
-  test('should format kilobytes correctly', () => {
-    expect(formatFileSize(1500)).toBe('1.46KB')
-  })
-  test('should format megabytes correctly', () => {
-    expect(formatFileSize(1500000)).toBe('1.43MB')
-  })
-  test('should format gigabytes correctly', () => {
-    expect(formatFileSize(1500000000)).toBe('1.40GB')
-  })
-  test('should format terabytes correctly', () => {
-    expect(formatFileSize(1500000000000)).toBe('1.36TB')
-  })
-  test('should format petabytes correctly', () => {
-    expect(formatFileSize(1500000000000000)).toBe('1.33PB')
-  })
-})
-describe('formatTime', () => {
-  test('should return the input if it is falsy', () => {
-    expect(formatTime(0)).toBe(0)
-  })
-  test('should format seconds correctly', () => {
-    expect(formatTime(30)).toBe('30.00 sec')
-  })
-  test('should format minutes correctly', () => {
-    expect(formatTime(90)).toBe('1.50 min')
-  })
-  test('should format hours correctly', () => {
-    expect(formatTime(3600)).toBe('1.00 h')
-  })
-  test('should handle large numbers', () => {
-    expect(formatTime(7200)).toBe('2.00 h')
-  })
-})