Model Runtime (#1858)

Co-authored-by: StyleZhang <jasonapring2015@outlook.com>
Co-authored-by: Garfield Dai <dai.hai@foxmail.com>
Co-authored-by: chenhe <guchenhe@gmail.com>
Co-authored-by: jyong <jyong@dify.ai>
Co-authored-by: Joel <iamjoel007@gmail.com>
Co-authored-by: Yeuoly <admin@srmxy.cn>

api/.vscode/launch.json

@@ -4,6 +4,21 @@
     // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
     "version": "0.2.0",
     "configurations": [
+        {
+            "name": "Python: Celery",
+            "type": "python",
+            "request": "launch",
+            "module": "celery",
+            "justMyCode": true,
+            "args": ["-A", "app.celery", "worker", "-P", "gevent", "-c", "1", "--loglevel", "info", "-Q", "dataset,generation,mail"],
+            "envFile": "${workspaceFolder}/.env",
+            "env": {
+                "FLASK_APP": "app.py",
+                "FLASK_DEBUG": "1",
+                "GEVENT_SUPPORT": "True"
+            },
+            "console": "integratedTerminal"
+        },
         {
             "name": "Python: Flask",
             "type": "python",

api/Dockerfile

@@ -34,9 +34,6 @@ RUN apt-get update \
 COPY --from=base /pkg /usr/local
 COPY . /app/api/
 
-RUN python -c "from transformers import GPT2TokenizerFast; GPT2TokenizerFast.from_pretrained('gpt2')"
-ENV TRANSFORMERS_OFFLINE true
-
 COPY docker/entrypoint.sh /entrypoint.sh
 RUN chmod +x /entrypoint.sh
 

api/app.py

@@ -6,9 +6,12 @@ from werkzeug.exceptions import Unauthorized
 if not os.environ.get("DEBUG") or os.environ.get("DEBUG").lower() != 'true':
     from gevent import monkey
     monkey.patch_all()
-    if os.environ.get("VECTOR_STORE") == 'milvus':
-        import grpc.experimental.gevent
-        grpc.experimental.gevent.init_gevent()
+    # if os.environ.get("VECTOR_STORE") == 'milvus':
+    import grpc.experimental.gevent
+    grpc.experimental.gevent.init_gevent()
+
+    import langchain
+    langchain.verbose = True
 
 import time
 import logging
@@ -18,9 +21,8 @@ import threading
 from flask import Flask, request, Response
 from flask_cors import CORS
 
-from core.model_providers.providers import hosted
 from extensions import ext_celery, ext_sentry, ext_redis, ext_login, ext_migrate, \
-    ext_database, ext_storage, ext_mail, ext_code_based_extension
+    ext_database, ext_storage, ext_mail, ext_code_based_extension, ext_hosting_provider
 from extensions.ext_database import db
 from extensions.ext_login import login_manager
 
@@ -79,8 +81,6 @@ def create_app(test_config=None) -> Flask:
     register_blueprints(app)
     register_commands(app)
 
-    hosted.init_app(app)
-
     return app
 
 
@@ -95,6 +95,7 @@ def initialize_extensions(app):
     ext_celery.init_app(app)
     ext_login.init_app(app)
     ext_mail.init_app(app)
+    ext_hosting_provider.init_app(app)
     ext_sentry.init_app(app)
 
 
@@ -105,13 +106,18 @@ def load_user_from_request(request_from_flask_login):
     if request.blueprint == 'console':
         # Check if the user_id contains a dot, indicating the old format
         auth_header = request.headers.get('Authorization', '')
-        if ' ' not in auth_header:
-            raise Unauthorized('Invalid Authorization header format. Expected \'Bearer <api-key>\' format.')
-        auth_scheme, auth_token = auth_header.split(None, 1)
-        auth_scheme = auth_scheme.lower()
-        if auth_scheme != 'bearer':
-            raise Unauthorized('Invalid Authorization header format. Expected \'Bearer <api-key>\' format.')
-        
+        if not auth_header:
+            auth_token = request.args.get('_token')
+            if not auth_token:
+                raise Unauthorized('Invalid Authorization token.')
+        else:
+            if ' ' not in auth_header:
+                raise Unauthorized('Invalid Authorization header format. Expected \'Bearer <api-key>\' format.')
+            auth_scheme, auth_token = auth_header.split(None, 1)
+            auth_scheme = auth_scheme.lower()
+            if auth_scheme != 'bearer':
+                raise Unauthorized('Invalid Authorization header format. Expected \'Bearer <api-key>\' format.')
+
         decoded = PassportService().verify(auth_token)
         user_id = decoded.get('user_id')
 

api/commands.py

@@ -12,16 +12,12 @@ import qdrant_client
 from qdrant_client.http.models import TextIndexParams, TextIndexType, TokenizerType
 from tqdm import tqdm
 from flask import current_app, Flask
-from langchain.embeddings import OpenAIEmbeddings
 from werkzeug.exceptions import NotFound
 
 from core.embedding.cached_embedding import CacheEmbedding
 from core.index.index import IndexBuilder
-from core.model_providers.model_factory import ModelFactory
-from core.model_providers.models.embedding.openai_embedding import OpenAIEmbedding
-from core.model_providers.models.entity.model_params import ModelType
-from core.model_providers.providers.hosted import hosted_model_providers
-from core.model_providers.providers.openai_provider import OpenAIProvider
+from core.model_manager import ModelManager
+from core.model_runtime.entities.model_entities import ModelType
 from libs.password import password_pattern, valid_password, hash_password
 from libs.helper import email as email_validate
 from extensions.ext_database import db
@@ -327,6 +323,8 @@ def create_qdrant_indexes():
         except NotFound:
             break
 
+        model_manager = ModelManager()
+
         page += 1
         for dataset in datasets:
             if dataset.index_struct_dict:
@@ -334,19 +332,23 @@ def create_qdrant_indexes():
                     try:
                         click.echo('Create dataset qdrant index: {}'.format(dataset.id))
                         try:
-                            embedding_model = ModelFactory.get_embedding_model(
+                            embedding_model = model_manager.get_model_instance(
                                 tenant_id=dataset.tenant_id,
-                                model_provider_name=dataset.embedding_model_provider,
-                                model_name=dataset.embedding_model
+                                provider=dataset.embedding_model_provider,
+                                model_type=ModelType.TEXT_EMBEDDING,
+                                model=dataset.embedding_model
+
                             )
                         except Exception:
                             try:
-                                embedding_model = ModelFactory.get_embedding_model(
-                                    tenant_id=dataset.tenant_id
+                                embedding_model = model_manager.get_default_model_instance(
+                                    tenant_id=dataset.tenant_id,
+                                    model_type=ModelType.TEXT_EMBEDDING,
                                 )
-                                dataset.embedding_model = embedding_model.name
-                                dataset.embedding_model_provider = embedding_model.model_provider.provider_name
+                                dataset.embedding_model = embedding_model.model
+                                dataset.embedding_model_provider = embedding_model.provider
                             except Exception:
+
                                 provider = Provider(
                                     id='provider_id',
                                     tenant_id=dataset.tenant_id,

api/config.py

@@ -87,7 +87,7 @@ class Config:
         # ------------------------
         # General Configurations.
         # ------------------------
-        self.CURRENT_VERSION = "0.3.34"
+        self.CURRENT_VERSION = "0.4.0"
         self.COMMIT_SHA = get_env('COMMIT_SHA')
         self.EDITION = "SELF_HOSTED"
         self.DEPLOY_ENV = get_env('DEPLOY_ENV')

api/controllers/console/__init__.py

@@ -18,7 +18,7 @@ from .auth import login, oauth, data_source_oauth, activate
 from .datasets import datasets, datasets_document, datasets_segments, file, hit_testing, data_source
 
 # Import workspace controllers
-from .workspace import workspace, members, providers, model_providers, account, tool_providers, models
+from .workspace import workspace, members, model_providers, account, tool_providers, models
 
 # Import explore controllers
 from .explore import installed_app, recommended_app, completion, conversation, message, parameter, saved_message, audio

api/controllers/console/app/app.py

@@ -4,6 +4,10 @@ import logging
 from datetime import datetime
 
 from flask_login import current_user
+
+from core.model_manager import ModelManager
+from core.model_runtime.entities.model_entities import ModelType
+from core.provider_manager import ProviderManager
 from libs.login import login_required
 from flask_restful import Resource, reqparse, marshal_with, abort, inputs
 from werkzeug.exceptions import Forbidden
@@ -13,9 +17,7 @@ from controllers.console import api
 from controllers.console.app.error import AppNotFoundError, ProviderNotInitializeError
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required, cloud_edition_billing_resource_check
-from core.model_providers.error import ProviderTokenNotInitError, LLMBadRequestError
-from core.model_providers.model_factory import ModelFactory
-from core.model_providers.model_provider_factory import ModelProviderFactory
+from core.errors.error import ProviderTokenNotInitError, LLMBadRequestError
 from events.app_event import app_was_created, app_was_deleted
 from fields.app_fields import app_pagination_fields, app_detail_fields, template_list_fields, \
     app_detail_fields_with_site
@@ -73,39 +75,41 @@ class AppListApi(Resource):
             raise Forbidden()
 
         try:
-            default_model = ModelFactory.get_text_generation_model(
-                tenant_id=current_user.current_tenant_id
+            provider_manager = ProviderManager()
+            default_model_entity = provider_manager.get_default_model(
+                tenant_id=current_user.current_tenant_id,
+                model_type=ModelType.LLM
             )
         except (ProviderTokenNotInitError, LLMBadRequestError):
-            default_model = None
+            default_model_entity = None
         except Exception as e:
             logging.exception(e)
-            default_model = None
+            default_model_entity = None
 
         if args['model_config'] is not None:
             # validate config
             model_config_dict = args['model_config']
 
             # get model provider
-            model_provider = ModelProviderFactory.get_preferred_model_provider(
-                current_user.current_tenant_id,
-                model_config_dict["model"]["provider"]
+            model_manager = ModelManager()
+            model_instance = model_manager.get_default_model_instance(
+                tenant_id=current_user.current_tenant_id,
+                model_type=ModelType.LLM
             )
 
-            if not model_provider:
-                if not default_model:
-                    raise ProviderNotInitializeError(
-                        f"No Default System Reasoning Model available. Please configure "
-                        f"in the Settings -> Model Provider.")
-                else:
-                    model_config_dict["model"]["provider"] = default_model.model_provider.provider_name
-                    model_config_dict["model"]["name"] = default_model.name
+            if not model_instance:
+                raise ProviderNotInitializeError(
+                    f"No Default System Reasoning Model available. Please configure "
+                    f"in the Settings -> Model Provider.")
+            else:
+                model_config_dict["model"]["provider"] = model_instance.provider
+                model_config_dict["model"]["name"] = model_instance.model
 
             model_configuration = AppModelConfigService.validate_configuration(
                 tenant_id=current_user.current_tenant_id,
                 account=current_user,
                 config=model_config_dict,
-                mode=args['mode']
+                app_mode=args['mode']
             )
 
             app = App(
@@ -129,21 +133,27 @@ class AppListApi(Resource):
             app_model_config = AppModelConfig(**model_config_template['model_config'])
 
             # get model provider
-            model_provider = ModelProviderFactory.get_preferred_model_provider(
-                current_user.current_tenant_id,
-                app_model_config.model_dict["provider"]
-            )
+            model_manager = ModelManager()
 
-            if not model_provider:
-                if not default_model:
-                    raise ProviderNotInitializeError(
-                        f"No Default System Reasoning Model available. Please configure "
-                        f"in the Settings -> Model Provider.")
-                else:
-                    model_dict = app_model_config.model_dict
-                    model_dict['provider'] = default_model.model_provider.provider_name
-                    model_dict['name'] = default_model.name
-                    app_model_config.model = json.dumps(model_dict)
+            try:
+                model_instance = model_manager.get_default_model_instance(
+                    tenant_id=current_user.current_tenant_id,
+                    model_type=ModelType.LLM
+                )
+            except ProviderTokenNotInitError:
+                raise ProviderNotInitializeError(
+                    f"No Default System Reasoning Model available. Please configure "
+                    f"in the Settings -> Model Provider.")
+
+            if not model_instance:
+                raise ProviderNotInitializeError(
+                    f"No Default System Reasoning Model available. Please configure "
+                    f"in the Settings -> Model Provider.")
+            else:
+                model_dict = app_model_config.model_dict
+                model_dict['provider'] = model_instance.provider
+                model_dict['name'] = model_instance.model
+                app_model_config.model = json.dumps(model_dict)
 
         app.name = args['name']
         app.mode = args['mode']

api/controllers/console/app/audio.py

@@ -2,6 +2,8 @@
 import logging
 
 from flask import request
+
+from core.model_runtime.errors.invoke import InvokeError
 from libs.login import login_required
 from werkzeug.exceptions import InternalServerError
 
@@ -14,8 +16,7 @@ from controllers.console.app.error import AppUnavailableError, \
     UnsupportedAudioTypeError, ProviderNotSupportSpeechToTextError
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
-from core.model_providers.error import LLMBadRequestError, LLMAPIUnavailableError, LLMAuthorizationError, LLMAPIConnectionError, \
-    LLMRateLimitError, ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
 from flask_restful import Resource
 from services.audio_service import AudioService
 from services.errors.audio import NoAudioUploadedServiceError, AudioTooLargeServiceError, \
@@ -56,8 +57,7 @@ class ChatMessageAudioApi(Resource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e

api/controllers/console/app/completion.py

@@ -5,6 +5,10 @@ from typing import Generator, Union
 
 import flask_login
 from flask import Response, stream_with_context
+
+from core.application_queue_manager import ApplicationQueueManager
+from core.entities.application_entities import InvokeFrom
+from core.model_runtime.errors.invoke import InvokeError
 from libs.login import login_required
 from werkzeug.exceptions import InternalServerError, NotFound
 
@@ -16,9 +20,7 @@ from controllers.console.app.error import ConversationCompletedError, AppUnavail
     ProviderModelCurrentlyNotSupportError
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
-from core.conversation_message_task import PubHandler
-from core.model_providers.error import LLMBadRequestError, LLMAPIUnavailableError, LLMAuthorizationError, LLMAPIConnectionError, \
-    LLMRateLimitError, ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
 from libs.helper import uuid_value
 from flask_restful import Resource, reqparse
 
@@ -56,7 +58,7 @@ class CompletionMessageApi(Resource):
                 app_model=app_model,
                 user=account,
                 args=args,
-                from_source='console',
+                invoke_from=InvokeFrom.DEBUGGER,
                 streaming=streaming,
                 is_model_config_override=True
             )
@@ -75,8 +77,7 @@ class CompletionMessageApi(Resource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e
@@ -97,7 +98,7 @@ class CompletionMessageStopApi(Resource):
 
         account = flask_login.current_user
 
-        PubHandler.stop(account, task_id)
+        ApplicationQueueManager.set_stop_flag(task_id, InvokeFrom.DEBUGGER, account.id)
 
         return {'result': 'success'}, 200
 
@@ -132,7 +133,7 @@ class ChatMessageApi(Resource):
                 app_model=app_model,
                 user=account,
                 args=args,
-                from_source='console',
+                invoke_from=InvokeFrom.DEBUGGER,
                 streaming=streaming,
                 is_model_config_override=True
             )
@@ -151,8 +152,7 @@ class ChatMessageApi(Resource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e
@@ -182,9 +182,8 @@ def compact_response(response: Union[dict, Generator]) -> Response:
                 yield "data: " + json.dumps(api.handle_error(ProviderQuotaExceededError()).get_json()) + "\n\n"
             except ModelCurrentlyNotSupportError:
                 yield "data: " + json.dumps(api.handle_error(ProviderModelCurrentlyNotSupportError()).get_json()) + "\n\n"
-            except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                    LLMRateLimitError, LLMAuthorizationError) as e:
-                yield "data: " + json.dumps(api.handle_error(CompletionRequestError(str(e))).get_json()) + "\n\n"
+            except InvokeError as e:
+                yield "data: " + json.dumps(api.handle_error(CompletionRequestError(e.description)).get_json()) + "\n\n"
             except ValueError as e:
                 yield "data: " + json.dumps(api.handle_error(e).get_json()) + "\n\n"
             except Exception:
@@ -207,7 +206,7 @@ class ChatMessageStopApi(Resource):
 
         account = flask_login.current_user
 
-        PubHandler.stop(account, task_id)
+        ApplicationQueueManager.set_stop_flag(task_id, InvokeFrom.DEBUGGER, account.id)
 
         return {'result': 'success'}, 200
 

api/controllers/console/app/generator.py

@@ -1,4 +1,6 @@
 from flask_login import current_user
+
+from core.model_runtime.errors.invoke import InvokeError
 from libs.login import login_required
 from flask_restful import Resource, reqparse
 
@@ -8,8 +10,7 @@ from controllers.console.app.error import ProviderNotInitializeError, ProviderQu
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
 from core.generator.llm_generator import LLMGenerator
-from core.model_providers.error import ProviderTokenNotInitError, QuotaExceededError, LLMBadRequestError, LLMAPIConnectionError, \
-    LLMAPIUnavailableError, LLMRateLimitError, LLMAuthorizationError, ModelCurrentlyNotSupportError
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
 
 
 class RuleGenerateApi(Resource):
@@ -36,8 +37,7 @@ class RuleGenerateApi(Resource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
 
         return rules

api/controllers/console/app/message.py

@@ -14,8 +14,9 @@ from controllers.console.app.error import CompletionRequestError, ProviderNotIni
     AppMoreLikeThisDisabledError, ProviderQuotaExceededError, ProviderModelCurrentlyNotSupportError
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required, cloud_edition_billing_resource_check
-from core.model_providers.error import LLMRateLimitError, LLMBadRequestError, LLMAuthorizationError, LLMAPIConnectionError, \
-    ProviderTokenNotInitError, LLMAPIUnavailableError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.entities.application_entities import InvokeFrom
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.model_runtime.errors.invoke import InvokeError
 from libs.login import login_required
 from fields.conversation_fields import message_detail_fields, annotation_fields
 from libs.helper import uuid_value
@@ -208,7 +209,13 @@ class MessageMoreLikeThisApi(Resource):
         app_model = _get_app(app_id, 'completion')
 
         try:
-            response = CompletionService.generate_more_like_this(app_model, current_user, message_id, streaming)
+            response = CompletionService.generate_more_like_this(
+                app_model=app_model,
+                user=current_user,
+                message_id=message_id,
+                invoke_from=InvokeFrom.DEBUGGER,
+                streaming=streaming
+            )
             return compact_response(response)
         except MessageNotExistsError:
             raise NotFound("Message Not Exists.")
@@ -220,8 +227,7 @@ class MessageMoreLikeThisApi(Resource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e
@@ -249,8 +255,7 @@ def compact_response(response: Union[dict, Generator]) -> Response:
             except ModelCurrentlyNotSupportError:
                 yield "data: " + json.dumps(
                     api.handle_error(ProviderModelCurrentlyNotSupportError()).get_json()) + "\n\n"
-            except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                    LLMRateLimitError, LLMAuthorizationError) as e:
+            except InvokeError as e:
                 yield "data: " + json.dumps(api.handle_error(CompletionRequestError(str(e))).get_json()) + "\n\n"
             except ValueError as e:
                 yield "data: " + json.dumps(api.handle_error(e).get_json()) + "\n\n"
@@ -290,8 +295,7 @@ class MessageSuggestedQuestionApi(Resource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except Exception:
             logging.exception("internal server error.")

api/controllers/console/app/model_config.py

@@ -31,7 +31,7 @@ class ModelConfigResource(Resource):
             tenant_id=current_user.current_tenant_id,
             account=current_user,
             config=request.json,
-            mode=app.mode
+            app_mode=app.mode
         )
 
         new_app_model_config = AppModelConfig(

api/controllers/console/datasets/datasets.py

@@ -4,6 +4,8 @@ from flask import request, current_app
 from flask_login import current_user
 
 from controllers.console.apikey import api_key_list, api_key_fields
+from core.model_runtime.entities.model_entities import ModelType
+from core.provider_manager import ProviderManager
 from libs.login import login_required
 from flask_restful import Resource, reqparse, marshal, marshal_with
 from werkzeug.exceptions import NotFound, Forbidden
@@ -14,8 +16,7 @@ from controllers.console.datasets.error import DatasetNameDuplicateError
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
 from core.indexing_runner import IndexingRunner
-from core.model_providers.error import LLMBadRequestError, ProviderTokenNotInitError
-from core.model_providers.models.entity.model_params import ModelType
+from core.errors.error import LLMBadRequestError, ProviderTokenNotInitError
 from fields.app_fields import related_app_list
 from fields.dataset_fields import dataset_detail_fields, dataset_query_detail_fields
 from fields.document_fields import document_status_fields
@@ -23,7 +24,6 @@ from extensions.ext_database import db
 from models.dataset import DocumentSegment, Document
 from models.model import UploadFile, ApiToken
 from services.dataset_service import DatasetService, DocumentService
-from services.provider_service import ProviderService
 
 
 def _validate_name(name):
@@ -55,16 +55,20 @@ class DatasetListApi(Resource):
                                                           current_user.current_tenant_id, current_user)
 
         # check embedding setting
-        provider_service = ProviderService()
-        valid_model_list = provider_service.get_valid_model_list(current_user.current_tenant_id,
-                                                                 ModelType.EMBEDDINGS.value)
-        # if len(valid_model_list) == 0:
-        #     raise ProviderNotInitializeError(
-        #         f"No Embedding Model available. Please configure a valid provider "
-        #         f"in the Settings -> Model Provider.")
+        provider_manager = ProviderManager()
+        configurations = provider_manager.get_configurations(
+            tenant_id=current_user.current_tenant_id
+        )
+
+        embedding_models = configurations.get_models(
+            model_type=ModelType.TEXT_EMBEDDING,
+            only_active=True
+        )
+
         model_names = []
-        for valid_model in valid_model_list:
-            model_names.append(f"{valid_model['model_name']}:{valid_model['model_provider']['provider_name']}")
+        for embedding_model in embedding_models:
+            model_names.append(f"{embedding_model.model}:{embedding_model.provider.provider}")
+
         data = marshal(datasets, dataset_detail_fields)
         for item in data:
             if item['indexing_technique'] == 'high_quality':
@@ -75,6 +79,7 @@ class DatasetListApi(Resource):
                     item['embedding_available'] = False
             else:
                 item['embedding_available'] = True
+
         response = {
             'data': data,
             'has_more': len(datasets) == limit,
@@ -130,13 +135,20 @@ class DatasetApi(Resource):
             raise Forbidden(str(e))
         data = marshal(dataset, dataset_detail_fields)
         # check embedding setting
-        provider_service = ProviderService()
-        # get valid model list
-        valid_model_list = provider_service.get_valid_model_list(current_user.current_tenant_id,
-                                                                 ModelType.EMBEDDINGS.value)
+        provider_manager = ProviderManager()
+        configurations = provider_manager.get_configurations(
+            tenant_id=current_user.current_tenant_id
+        )
+
+        embedding_models = configurations.get_models(
+            model_type=ModelType.TEXT_EMBEDDING,
+            only_active=True
+        )
+
         model_names = []
-        for valid_model in valid_model_list:
-            model_names.append(f"{valid_model['model_name']}:{valid_model['model_provider']['provider_name']}")
+        for embedding_model in embedding_models:
+            model_names.append(f"{embedding_model.model}:{embedding_model.provider.provider}")
+
         if data['indexing_technique'] == 'high_quality':
             item_model = f"{data['embedding_model']}:{data['embedding_model_provider']}"
             if item_model in model_names:

api/controllers/console/datasets/datasets_document.py

@@ -2,8 +2,12 @@
 from datetime import datetime
 from typing import List
 
-from flask import request, current_app
+from flask import request
 from flask_login import current_user
+
+from core.model_manager import ModelManager
+from core.model_runtime.entities.model_entities import ModelType
+from core.model_runtime.errors.invoke import InvokeAuthorizationError
 from libs.login import login_required
 from flask_restful import Resource, fields, marshal, marshal_with, reqparse
 from sqlalchemy import desc, asc
@@ -18,9 +22,8 @@ from controllers.console.datasets.error import DocumentAlreadyFinishedError, Inv
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required, cloud_edition_billing_resource_check
 from core.indexing_runner import IndexingRunner
-from core.model_providers.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError, \
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError, \
     LLMBadRequestError
-from core.model_providers.model_factory import ModelFactory
 from extensions.ext_redis import redis_client
 from fields.document_fields import document_with_segments_fields, document_fields, \
     dataset_and_document_fields, document_status_fields
@@ -272,10 +275,12 @@ class DatasetInitApi(Resource):
         args = parser.parse_args()
         if args['indexing_technique'] == 'high_quality':
             try:
-                ModelFactory.get_embedding_model(
-                    tenant_id=current_user.current_tenant_id
+                model_manager = ModelManager()
+                model_manager.get_default_model_instance(
+                    tenant_id=current_user.current_tenant_id,
+                    model_type=ModelType.TEXT_EMBEDDING
                 )
-            except LLMBadRequestError:
+            except InvokeAuthorizationError:
                 raise ProviderNotInitializeError(
                     f"No Embedding Model available. Please configure a valid provider "
                     f"in the Settings -> Model Provider.")

api/controllers/console/datasets/datasets_segments.py

@@ -12,8 +12,9 @@ from controllers.console.app.error import ProviderNotInitializeError
 from controllers.console.datasets.error import InvalidActionError, NoFileUploadedError, TooManyFilesError
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required, cloud_edition_billing_resource_check
-from core.model_providers.error import LLMBadRequestError, ProviderTokenNotInitError
-from core.model_providers.model_factory import ModelFactory
+from core.errors.error import LLMBadRequestError, ProviderTokenNotInitError
+from core.model_manager import ModelManager
+from core.model_runtime.entities.model_entities import ModelType
 from libs.login import login_required
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
@@ -133,10 +134,12 @@ class DatasetDocumentSegmentApi(Resource):
         if dataset.indexing_technique == 'high_quality':
             # check embedding model setting
             try:
-                ModelFactory.get_embedding_model(
+                model_manager = ModelManager()
+                model_manager.get_model_instance(
                     tenant_id=current_user.current_tenant_id,
-                    model_provider_name=dataset.embedding_model_provider,
-                    model_name=dataset.embedding_model
+                    provider=dataset.embedding_model_provider,
+                    model_type=ModelType.TEXT_EMBEDDING,
+                    model=dataset.embedding_model
                 )
             except LLMBadRequestError:
                 raise ProviderNotInitializeError(
@@ -219,10 +222,12 @@ class DatasetDocumentSegmentAddApi(Resource):
         # check embedding model setting
         if dataset.indexing_technique == 'high_quality':
             try:
-                ModelFactory.get_embedding_model(
+                model_manager = ModelManager()
+                model_manager.get_model_instance(
                     tenant_id=current_user.current_tenant_id,
-                    model_provider_name=dataset.embedding_model_provider,
-                    model_name=dataset.embedding_model
+                    provider=dataset.embedding_model_provider,
+                    model_type=ModelType.TEXT_EMBEDDING,
+                    model=dataset.embedding_model
                 )
             except LLMBadRequestError:
                 raise ProviderNotInitializeError(
@@ -269,10 +274,12 @@ class DatasetDocumentSegmentUpdateApi(Resource):
         if dataset.indexing_technique == 'high_quality':
             # check embedding model setting
             try:
-                ModelFactory.get_embedding_model(
+                model_manager = ModelManager()
+                model_manager.get_model_instance(
                     tenant_id=current_user.current_tenant_id,
-                    model_provider_name=dataset.embedding_model_provider,
-                    model_name=dataset.embedding_model
+                    provider=dataset.embedding_model_provider,
+                    model_type=ModelType.TEXT_EMBEDDING,
+                    model=dataset.embedding_model
                 )
             except LLMBadRequestError:
                 raise ProviderNotInitializeError(

api/controllers/console/datasets/hit_testing.py

@@ -12,7 +12,7 @@ from controllers.console.app.error import ProviderNotInitializeError, ProviderQu
 from controllers.console.datasets.error import HighQualityDatasetOnlyError, DatasetNotInitializedError
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
-from core.model_providers.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError, \
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError, \
     LLMBadRequestError
 from fields.hit_testing_fields import hit_testing_record_fields
 from services.dataset_service import DatasetService

api/controllers/console/explore/audio.py

@@ -11,8 +11,8 @@ from controllers.console.app.error import AppUnavailableError, ProviderNotInitia
     NoAudioUploadedError, AudioTooLargeError, \
     UnsupportedAudioTypeError, ProviderNotSupportSpeechToTextError
 from controllers.console.explore.wraps import InstalledAppResource
-from core.model_providers.error import LLMBadRequestError, LLMAPIUnavailableError, LLMAuthorizationError, LLMAPIConnectionError, \
-    LLMRateLimitError, ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.model_runtime.errors.invoke import InvokeError
 from services.audio_service import AudioService
 from services.errors.audio import NoAudioUploadedServiceError, AudioTooLargeServiceError, \
     UnsupportedAudioTypeServiceError, ProviderNotSupportSpeechToTextServiceError
@@ -53,8 +53,7 @@ class ChatAudioApi(InstalledAppResource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e

api/controllers/console/explore/completion.py

@@ -15,9 +15,10 @@ from controllers.console.app.error import ConversationCompletedError, AppUnavail
     ProviderQuotaExceededError, ProviderModelCurrentlyNotSupportError, CompletionRequestError
 from controllers.console.explore.error import NotCompletionAppError, NotChatAppError
 from controllers.console.explore.wraps import InstalledAppResource
-from core.conversation_message_task import PubHandler
-from core.model_providers.error import LLMBadRequestError, LLMAPIUnavailableError, LLMAuthorizationError, LLMAPIConnectionError, \
-    LLMRateLimitError, ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.application_queue_manager import ApplicationQueueManager
+from core.entities.application_entities import InvokeFrom
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.model_runtime.errors.invoke import InvokeError
 from extensions.ext_database import db
 from libs.helper import uuid_value
 from services.completion_service import CompletionService
@@ -50,7 +51,7 @@ class CompletionApi(InstalledAppResource):
                 app_model=app_model,
                 user=current_user,
                 args=args,
-                from_source='console',
+                invoke_from=InvokeFrom.EXPLORE,
                 streaming=streaming
             )
 
@@ -68,8 +69,7 @@ class CompletionApi(InstalledAppResource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e
@@ -84,7 +84,7 @@ class CompletionStopApi(InstalledAppResource):
         if app_model.mode != 'completion':
             raise NotCompletionAppError()
 
-        PubHandler.stop(current_user, task_id)
+        ApplicationQueueManager.set_stop_flag(task_id, InvokeFrom.EXPLORE, current_user.id)
 
         return {'result': 'success'}, 200
 
@@ -115,7 +115,7 @@ class ChatApi(InstalledAppResource):
                 app_model=app_model,
                 user=current_user,
                 args=args,
-                from_source='console',
+                invoke_from=InvokeFrom.EXPLORE,
                 streaming=streaming
             )
 
@@ -133,8 +133,7 @@ class ChatApi(InstalledAppResource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e
@@ -149,7 +148,7 @@ class ChatStopApi(InstalledAppResource):
         if app_model.mode != 'chat':
             raise NotChatAppError()
 
-        PubHandler.stop(current_user, task_id)
+        ApplicationQueueManager.set_stop_flag(task_id, InvokeFrom.EXPLORE, current_user.id)
 
         return {'result': 'success'}, 200
 
@@ -175,8 +174,7 @@ def compact_response(response: Union[dict, Generator]) -> Response:
                 yield "data: " + json.dumps(api.handle_error(ProviderQuotaExceededError()).get_json()) + "\n\n"
             except ModelCurrentlyNotSupportError:
                 yield "data: " + json.dumps(api.handle_error(ProviderModelCurrentlyNotSupportError()).get_json()) + "\n\n"
-            except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                    LLMRateLimitError, LLMAuthorizationError) as e:
+            except InvokeError as e:
                 yield "data: " + json.dumps(api.handle_error(CompletionRequestError(str(e))).get_json()) + "\n\n"
             except ValueError as e:
                 yield "data: " + json.dumps(api.handle_error(e).get_json()) + "\n\n"

api/controllers/console/explore/message.py

@@ -5,7 +5,7 @@ from typing import Generator, Union
 
 from flask import stream_with_context, Response
 from flask_login import current_user
-from flask_restful import reqparse, fields, marshal_with
+from flask_restful import reqparse, marshal_with
 from flask_restful.inputs import int_range
 from werkzeug.exceptions import NotFound, InternalServerError
 
@@ -13,12 +13,14 @@ import services
 from controllers.console import api
 from controllers.console.app.error import AppMoreLikeThisDisabledError, ProviderNotInitializeError, \
     ProviderQuotaExceededError, ProviderModelCurrentlyNotSupportError, CompletionRequestError
-from controllers.console.explore.error import NotCompletionAppError, AppSuggestedQuestionsAfterAnswerDisabledError
+from controllers.console.explore.error import NotCompletionAppError, AppSuggestedQuestionsAfterAnswerDisabledError, \
+    NotChatAppError
 from controllers.console.explore.wraps import InstalledAppResource
-from core.model_providers.error import LLMRateLimitError, LLMBadRequestError, LLMAuthorizationError, LLMAPIConnectionError, \
-    ProviderTokenNotInitError, LLMAPIUnavailableError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.entities.application_entities import InvokeFrom
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.model_runtime.errors.invoke import InvokeError
 from fields.message_fields import message_infinite_scroll_pagination_fields
-from libs.helper import uuid_value, TimestampField
+from libs.helper import uuid_value
 from services.completion_service import CompletionService
 from services.errors.app import MoreLikeThisDisabledError
 from services.errors.conversation import ConversationNotExistsError
@@ -83,7 +85,13 @@ class MessageMoreLikeThisApi(InstalledAppResource):
         streaming = args['response_mode'] == 'streaming'
 
         try:
-            response = CompletionService.generate_more_like_this(app_model, current_user, message_id, streaming)
+            response = CompletionService.generate_more_like_this(
+                app_model=app_model,
+                user=current_user,
+                message_id=message_id,
+                invoke_from=InvokeFrom.EXPLORE,
+                streaming=streaming
+            )
             return compact_response(response)
         except MessageNotExistsError:
             raise NotFound("Message Not Exists.")
@@ -95,8 +103,7 @@ class MessageMoreLikeThisApi(InstalledAppResource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e
@@ -123,8 +130,7 @@ def compact_response(response: Union[dict, Generator]) -> Response:
                 yield "data: " + json.dumps(api.handle_error(ProviderQuotaExceededError()).get_json()) + "\n\n"
             except ModelCurrentlyNotSupportError:
                 yield "data: " + json.dumps(api.handle_error(ProviderModelCurrentlyNotSupportError()).get_json()) + "\n\n"
-            except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                    LLMRateLimitError, LLMAuthorizationError) as e:
+            except InvokeError as e:
                 yield "data: " + json.dumps(api.handle_error(CompletionRequestError(str(e))).get_json()) + "\n\n"
             except ValueError as e:
                 yield "data: " + json.dumps(api.handle_error(e).get_json()) + "\n\n"
@@ -162,8 +168,7 @@ class MessageSuggestedQuestionApi(InstalledAppResource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except Exception:
             logging.exception("internal server error.")

api/controllers/console/universal_chat/audio.py

@@ -11,8 +11,8 @@ from controllers.console.app.error import AppUnavailableError, ProviderNotInitia
     NoAudioUploadedError, AudioTooLargeError, \
     UnsupportedAudioTypeError, ProviderNotSupportSpeechToTextError
 from controllers.console.universal_chat.wraps import UniversalChatResource
-from core.model_providers.error import LLMBadRequestError, LLMAPIUnavailableError, LLMAuthorizationError, LLMAPIConnectionError, \
-    LLMRateLimitError, ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.errors.error import ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from core.model_runtime.errors.invoke import InvokeError
 from services.audio_service import AudioService
 from services.errors.audio import NoAudioUploadedServiceError, AudioTooLargeServiceError, \
     UnsupportedAudioTypeServiceError, ProviderNotSupportSpeechToTextServiceError
@@ -53,8 +53,7 @@ class UniversalChatAudioApi(UniversalChatResource):
             raise ProviderQuotaExceededError()
         except ModelCurrentlyNotSupportError:
             raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
+        except InvokeError as e:
             raise CompletionRequestError(str(e))
         except ValueError as e:
             raise e