fix: change model mode. (#1520)

Co-authored-by: Joel <iamjoel007@gmail.com>

README.md

@@ -37,7 +37,6 @@ https://github.com/langgenius/dify/assets/100913391/f6e658d5-31b3-4c16-a0af-9e19
 
 
 We provide the following free resources for registered Dify cloud users (sign up at [dify.ai](https://dify.ai)):
-* 600,000 free Claude model tokens to build Claude-powered apps
 * 200 free OpenAI queries to build OpenAI-based apps
 
   

README_CN.md

@@ -36,7 +36,6 @@ https://github.com/langgenius/dify/assets/100913391/f6e658d5-31b3-4c16-a0af-9e19
 
 
 我们为所有注册云端版的用户免费提供以下资源（登录 [dify.ai](https://cloud.dify.ai) 即可使用）：
-* 60 万 Tokens Claude 模型的消息调用额度，用于创建基于 Claude 模型的 AI 应用
 * 200 次 OpenAI 模型的消息调用额度，用于创建基于 OpenAI 模型的 AI 应用
 * 300 万 讯飞星火大模型 Token 的调用额度，用于创建基于讯飞星火大模型的 AI 应用
 * 100 万 MiniMax Token 的调用额度，用于创建基于 MiniMax 模型的 AI 应用

api/.env.example

@@ -18,6 +18,9 @@ SERVICE_API_URL=http://127.0.0.1:5001
 APP_API_URL=http://127.0.0.1:5001
 APP_WEB_URL=http://127.0.0.1:3000
 
+# Files URL
+FILES_URL=http://127.0.0.1:5001
+
 # celery configuration
 CELERY_BROKER_URL=redis://:difyai123456@localhost:6379/1
 
@@ -70,6 +73,14 @@ MILVUS_USER=root
 MILVUS_PASSWORD=Milvus
 MILVUS_SECURE=false
 
+# Upload configuration
+UPLOAD_FILE_SIZE_LIMIT=15
+UPLOAD_FILE_BATCH_LIMIT=5
+UPLOAD_IMAGE_FILE_SIZE_LIMIT=10
+
+# Model Configuration
+MULTIMODAL_SEND_IMAGE_FORMAT=base64
+
 # Mail configuration, support: resend
 MAIL_TYPE=
 MAIL_DEFAULT_SEND_FROM=no-reply <no-reply@dify.ai>

api/.vscode/launch.json

@@ -10,7 +10,7 @@
             "request": "launch",
             "module": "flask",
             "env": {
-                "FLASK_APP": "api/app.py",
+                "FLASK_APP": "app.py",
                 "FLASK_DEBUG": "1",
                 "GEVENT_SUPPORT": "True"
             },

api/app.py

@@ -19,7 +19,7 @@ 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_stripe
+    ext_database, ext_storage, ext_mail, ext_stripe, ext_code_based_extension
 from extensions.ext_database import db
 from extensions.ext_login import login_manager
 
@@ -79,6 +79,7 @@ def create_app(test_config=None) -> Flask:
 def initialize_extensions(app):
     # Since the application instance is now created, pass it to each Flask
     # extension instance to bind it to the Flask application instance (app)
+    ext_code_based_extension.init()
     ext_database.init_app(app)
     ext_migrate.init(app, db)
     ext_redis.init_app(app)
@@ -125,6 +126,7 @@ def register_blueprints(app):
     from controllers.service_api import bp as service_api_bp
     from controllers.web import bp as web_bp
     from controllers.console import bp as console_app_bp
+    from controllers.files import bp as files_bp
 
     CORS(service_api_bp,
          allow_headers=['Content-Type', 'Authorization', 'X-App-Code'],
@@ -154,6 +156,12 @@ def register_blueprints(app):
 
     app.register_blueprint(console_app_bp)
 
+    CORS(files_bp,
+         allow_headers=['Content-Type'],
+         methods=['GET', 'PUT', 'POST', 'DELETE', 'OPTIONS', 'PATCH']
+         )
+    app.register_blueprint(files_bp)
+
 
 # create app
 app = create_app()

api/config.py

@@ -26,6 +26,7 @@ DEFAULTS = {
     'SERVICE_API_URL': 'https://api.dify.ai',
     'APP_WEB_URL': 'https://udify.app',
     'APP_API_URL': 'https://udify.app',
+    'FILES_URL': '',
     'STORAGE_TYPE': 'local',
     'STORAGE_LOCAL_PATH': 'storage',
     'CHECK_UPDATE_URL': 'https://updates.dify.ai',
@@ -57,6 +58,9 @@ DEFAULTS = {
     'CLEAN_DAY_SETTING': 30,
     'UPLOAD_FILE_SIZE_LIMIT': 15,
     'UPLOAD_FILE_BATCH_LIMIT': 5,
+    'UPLOAD_IMAGE_FILE_SIZE_LIMIT': 10,
+    'OUTPUT_MODERATION_BUFFER_SIZE': 300,
+    'MULTIMODAL_SEND_IMAGE_FORMAT': 'base64'
 }
 
 
@@ -83,86 +87,65 @@ class Config:
     """Application configuration class."""
 
     def __init__(self):
-        # app settings
-        self.CONSOLE_API_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_API_URL')
-        self.CONSOLE_WEB_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_WEB_URL')
-        self.SERVICE_API_URL = get_env('API_URL') if get_env('API_URL') else get_env('SERVICE_API_URL')
-        self.APP_WEB_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_WEB_URL')
-        self.APP_API_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_API_URL')
-        self.CONSOLE_URL = get_env('CONSOLE_URL')
-        self.API_URL = get_env('API_URL')
-        self.APP_URL = get_env('APP_URL')
-        self.CURRENT_VERSION = "0.3.27"
+        # ------------------------
+        # General Configurations.
+        # ------------------------
+        self.CURRENT_VERSION = "0.3.30"
         self.COMMIT_SHA = get_env('COMMIT_SHA')
         self.EDITION = "SELF_HOSTED"
         self.DEPLOY_ENV = get_env('DEPLOY_ENV')
         self.TESTING = False
         self.LOG_LEVEL = get_env('LOG_LEVEL')
 
+        # The backend URL prefix of the console API.
+        # used to concatenate the login authorization callback or notion integration callback.
+        self.CONSOLE_API_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_API_URL')
+
+        # The front-end URL prefix of the console web.
+        # used to concatenate some front-end addresses and for CORS configuration use.
+        self.CONSOLE_WEB_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_WEB_URL')
+
+        # WebApp API backend Url prefix.
+        # used to declare the back-end URL for the front-end API.
+        self.APP_API_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_API_URL')
+
+        # WebApp Url prefix.
+        # used to display WebAPP API Base Url to the front-end.
+        self.APP_WEB_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_WEB_URL')
+
+        # Service API Url prefix.
+        # used to display Service API Base Url to the front-end.
+        self.SERVICE_API_URL = get_env('API_URL') if get_env('API_URL') else get_env('SERVICE_API_URL')
+
+        # File preview or download Url prefix.
+        # used to display File preview or download Url to the front-end or as Multi-model inputs;
+        # Url is signed and has expiration time.
+        self.FILES_URL = get_env('FILES_URL') if get_env('FILES_URL') else self.CONSOLE_API_URL
+
+        # Fallback Url prefix.
+        # Will be deprecated in the future.
+        self.CONSOLE_URL = get_env('CONSOLE_URL')
+        self.API_URL = get_env('API_URL')
+        self.APP_URL = get_env('APP_URL')
+
         # Your App secret key will be used for securely signing the session cookie
         # Make sure you are changing this key for your deployment with a strong key.
         # You can generate a strong key using `openssl rand -base64 42`.
         # Alternatively you can set it with `SECRET_KEY` environment variable.
         self.SECRET_KEY = get_env('SECRET_KEY')
 
-        # redis settings
-        self.REDIS_HOST = get_env('REDIS_HOST')
-        self.REDIS_PORT = get_env('REDIS_PORT')
-        self.REDIS_USERNAME = get_env('REDIS_USERNAME')
-        self.REDIS_PASSWORD = get_env('REDIS_PASSWORD')
-        self.REDIS_DB = get_env('REDIS_DB')
-        self.REDIS_USE_SSL = get_bool_env('REDIS_USE_SSL')
-
-        # storage settings
-        self.STORAGE_TYPE = get_env('STORAGE_TYPE')
-        self.STORAGE_LOCAL_PATH = get_env('STORAGE_LOCAL_PATH')
-        self.S3_ENDPOINT = get_env('S3_ENDPOINT')
-        self.S3_BUCKET_NAME = get_env('S3_BUCKET_NAME')
-        self.S3_ACCESS_KEY = get_env('S3_ACCESS_KEY')
-        self.S3_SECRET_KEY = get_env('S3_SECRET_KEY')
-        self.S3_REGION = get_env('S3_REGION')
-
-        # vector store settings, only support weaviate, qdrant
-        self.VECTOR_STORE = get_env('VECTOR_STORE')
-
-        # weaviate settings
-        self.WEAVIATE_ENDPOINT = get_env('WEAVIATE_ENDPOINT')
-        self.WEAVIATE_API_KEY = get_env('WEAVIATE_API_KEY')
-        self.WEAVIATE_GRPC_ENABLED = get_bool_env('WEAVIATE_GRPC_ENABLED')
-        self.WEAVIATE_BATCH_SIZE = int(get_env('WEAVIATE_BATCH_SIZE'))
-
-        # qdrant settings
-        self.QDRANT_URL = get_env('QDRANT_URL')
-        self.QDRANT_API_KEY = get_env('QDRANT_API_KEY')
-
-        # milvus setting
-        self.MILVUS_HOST = get_env('MILVUS_HOST')
-        self.MILVUS_PORT = get_env('MILVUS_PORT')
-        self.MILVUS_USER = get_env('MILVUS_USER')
-        self.MILVUS_PASSWORD = get_env('MILVUS_PASSWORD')
-        self.MILVUS_SECURE = get_env('MILVUS_SECURE')
-
-
         # cors settings
         self.CONSOLE_CORS_ALLOW_ORIGINS = get_cors_allow_origins(
             'CONSOLE_CORS_ALLOW_ORIGINS', self.CONSOLE_WEB_URL)
         self.WEB_API_CORS_ALLOW_ORIGINS = get_cors_allow_origins(
             'WEB_API_CORS_ALLOW_ORIGINS', '*')
 
-        # mail settings
-        self.MAIL_TYPE = get_env('MAIL_TYPE')
-        self.MAIL_DEFAULT_SEND_FROM = get_env('MAIL_DEFAULT_SEND_FROM')
-        self.RESEND_API_KEY = get_env('RESEND_API_KEY')
-
-        # sentry settings
-        self.SENTRY_DSN = get_env('SENTRY_DSN')
-        self.SENTRY_TRACES_SAMPLE_RATE = float(get_env('SENTRY_TRACES_SAMPLE_RATE'))
-        self.SENTRY_PROFILES_SAMPLE_RATE = float(get_env('SENTRY_PROFILES_SAMPLE_RATE'))
-
         # check update url
         self.CHECK_UPDATE_URL = get_env('CHECK_UPDATE_URL')
 
-        # database settings
+        # ------------------------
+        # Database Configurations.
+        # ------------------------
         db_credentials = {
             key: get_env(key) for key in
             ['DB_USERNAME', 'DB_PASSWORD', 'DB_HOST', 'DB_PORT', 'DB_DATABASE']
@@ -176,14 +159,102 @@ class Config:
 
         self.SQLALCHEMY_ECHO = get_bool_env('SQLALCHEMY_ECHO')
 
-        # celery settings
+        # ------------------------
+        # Redis Configurations.
+        # ------------------------
+        self.REDIS_HOST = get_env('REDIS_HOST')
+        self.REDIS_PORT = get_env('REDIS_PORT')
+        self.REDIS_USERNAME = get_env('REDIS_USERNAME')
+        self.REDIS_PASSWORD = get_env('REDIS_PASSWORD')
+        self.REDIS_DB = get_env('REDIS_DB')
+        self.REDIS_USE_SSL = get_bool_env('REDIS_USE_SSL')
+
+        # ------------------------
+        # Celery worker Configurations.
+        # ------------------------
         self.CELERY_BROKER_URL = get_env('CELERY_BROKER_URL')
         self.CELERY_BACKEND = get_env('CELERY_BACKEND')
         self.CELERY_RESULT_BACKEND = 'db+{}'.format(self.SQLALCHEMY_DATABASE_URI) \
             if self.CELERY_BACKEND == 'database' else self.CELERY_BROKER_URL
         self.BROKER_USE_SSL = self.CELERY_BROKER_URL.startswith('rediss://')
 
-        # hosted provider credentials
+        # ------------------------
+        # File Storage Configurations.
+        # ------------------------
+        self.STORAGE_TYPE = get_env('STORAGE_TYPE')
+        self.STORAGE_LOCAL_PATH = get_env('STORAGE_LOCAL_PATH')
+        self.S3_ENDPOINT = get_env('S3_ENDPOINT')
+        self.S3_BUCKET_NAME = get_env('S3_BUCKET_NAME')
+        self.S3_ACCESS_KEY = get_env('S3_ACCESS_KEY')
+        self.S3_SECRET_KEY = get_env('S3_SECRET_KEY')
+        self.S3_REGION = get_env('S3_REGION')
+
+        # ------------------------
+        # Vector Store Configurations.
+        # Currently, only support: qdrant, milvus, zilliz, weaviate
+        # ------------------------
+        self.VECTOR_STORE = get_env('VECTOR_STORE')
+
+        # qdrant settings
+        self.QDRANT_URL = get_env('QDRANT_URL')
+        self.QDRANT_API_KEY = get_env('QDRANT_API_KEY')
+
+        # milvus / zilliz setting
+        self.MILVUS_HOST = get_env('MILVUS_HOST')
+        self.MILVUS_PORT = get_env('MILVUS_PORT')
+        self.MILVUS_USER = get_env('MILVUS_USER')
+        self.MILVUS_PASSWORD = get_env('MILVUS_PASSWORD')
+        self.MILVUS_SECURE = get_env('MILVUS_SECURE')
+
+        # weaviate settings
+        self.WEAVIATE_ENDPOINT = get_env('WEAVIATE_ENDPOINT')
+        self.WEAVIATE_API_KEY = get_env('WEAVIATE_API_KEY')
+        self.WEAVIATE_GRPC_ENABLED = get_bool_env('WEAVIATE_GRPC_ENABLED')
+        self.WEAVIATE_BATCH_SIZE = int(get_env('WEAVIATE_BATCH_SIZE'))
+
+        # ------------------------
+        # Mail Configurations.
+        # ------------------------
+        self.MAIL_TYPE = get_env('MAIL_TYPE')
+        self.MAIL_DEFAULT_SEND_FROM = get_env('MAIL_DEFAULT_SEND_FROM')
+        self.RESEND_API_KEY = get_env('RESEND_API_KEY')
+
+        # ------------------------
+        # Sentry Configurations.
+        # ------------------------
+        self.SENTRY_DSN = get_env('SENTRY_DSN')
+        self.SENTRY_TRACES_SAMPLE_RATE = float(get_env('SENTRY_TRACES_SAMPLE_RATE'))
+        self.SENTRY_PROFILES_SAMPLE_RATE = float(get_env('SENTRY_PROFILES_SAMPLE_RATE'))
+
+        # ------------------------
+        # Business Configurations.
+        # ------------------------
+
+        # multi model send image format, support base64, url, default is base64
+        self.MULTIMODAL_SEND_IMAGE_FORMAT = get_env('MULTIMODAL_SEND_IMAGE_FORMAT')
+
+        # Dataset Configurations.
+        self.TENANT_DOCUMENT_COUNT = get_env('TENANT_DOCUMENT_COUNT')
+        self.CLEAN_DAY_SETTING = get_env('CLEAN_DAY_SETTING')
+
+        # File upload Configurations.
+        self.UPLOAD_FILE_SIZE_LIMIT = int(get_env('UPLOAD_FILE_SIZE_LIMIT'))
+        self.UPLOAD_FILE_BATCH_LIMIT = int(get_env('UPLOAD_FILE_BATCH_LIMIT'))
+        self.UPLOAD_IMAGE_FILE_SIZE_LIMIT = int(get_env('UPLOAD_IMAGE_FILE_SIZE_LIMIT'))
+
+        # Moderation in app Configurations.
+        self.OUTPUT_MODERATION_BUFFER_SIZE = int(get_env('OUTPUT_MODERATION_BUFFER_SIZE'))
+
+        # Notion integration setting
+        self.NOTION_CLIENT_ID = get_env('NOTION_CLIENT_ID')
+        self.NOTION_CLIENT_SECRET = get_env('NOTION_CLIENT_SECRET')
+        self.NOTION_INTEGRATION_TYPE = get_env('NOTION_INTEGRATION_TYPE')
+        self.NOTION_INTERNAL_SECRET = get_env('NOTION_INTERNAL_SECRET')
+        self.NOTION_INTEGRATION_TOKEN = get_env('NOTION_INTEGRATION_TOKEN')
+
+        # ------------------------
+        # Platform Configurations.
+        # ------------------------
         self.HOSTED_OPENAI_ENABLED = get_bool_env('HOSTED_OPENAI_ENABLED')
         self.HOSTED_OPENAI_API_KEY = get_env('HOSTED_OPENAI_API_KEY')
         self.HOSTED_OPENAI_API_BASE = get_env('HOSTED_OPENAI_API_BASE')
@@ -211,23 +282,6 @@ class Config:
         self.HOSTED_MODERATION_ENABLED = get_bool_env('HOSTED_MODERATION_ENABLED')
         self.HOSTED_MODERATION_PROVIDERS = get_env('HOSTED_MODERATION_PROVIDERS')
 
-        self.STRIPE_API_KEY = get_env('STRIPE_API_KEY')
-        self.STRIPE_WEBHOOK_SECRET = get_env('STRIPE_WEBHOOK_SECRET')
-
-        # notion import setting
-        self.NOTION_CLIENT_ID = get_env('NOTION_CLIENT_ID')
-        self.NOTION_CLIENT_SECRET = get_env('NOTION_CLIENT_SECRET')
-        self.NOTION_INTEGRATION_TYPE = get_env('NOTION_INTEGRATION_TYPE')
-        self.NOTION_INTERNAL_SECRET = get_env('NOTION_INTERNAL_SECRET')
-        self.NOTION_INTEGRATION_TOKEN = get_env('NOTION_INTEGRATION_TOKEN')
-
-        self.TENANT_DOCUMENT_COUNT = get_env('TENANT_DOCUMENT_COUNT')
-        self.CLEAN_DAY_SETTING = get_env('CLEAN_DAY_SETTING')
-
-        # uploading settings
-        self.UPLOAD_FILE_SIZE_LIMIT = int(get_env('UPLOAD_FILE_SIZE_LIMIT'))
-        self.UPLOAD_FILE_BATCH_LIMIT = int(get_env('UPLOAD_FILE_BATCH_LIMIT'))
-
 
 class CloudEditionConfig(Config):
 
@@ -242,18 +296,5 @@ class CloudEditionConfig(Config):
         self.GOOGLE_CLIENT_SECRET = get_env('GOOGLE_CLIENT_SECRET')
         self.OAUTH_REDIRECT_PATH = get_env('OAUTH_REDIRECT_PATH')
 
-
-class TestConfig(Config):
-
-    def __init__(self):
-        super().__init__()
-
-        self.EDITION = "SELF_HOSTED"
-        self.TESTING = True
-
-        db_credentials = {
-            key: get_env(key) for key in ['DB_USERNAME', 'DB_PASSWORD', 'DB_HOST', 'DB_PORT']
-        }
-
-        # use a different database for testing: dify_test
-        self.SQLALCHEMY_DATABASE_URI = f"postgresql://{db_credentials['DB_USERNAME']}:{db_credentials['DB_PASSWORD']}@{db_credentials['DB_HOST']}:{db_credentials['DB_PORT']}/dify_test"
+        self.STRIPE_API_KEY = get_env('STRIPE_API_KEY')
+        self.STRIPE_WEBHOOK_SECRET = get_env('STRIPE_WEBHOOK_SECRET')

api/controllers/console/__init__.py

@@ -6,7 +6,7 @@ bp = Blueprint('console', __name__, url_prefix='/console/api')
 api = ExternalApi(bp)
 
 # Import other controllers
-from . import setup, version, apikey, admin
+from . import extension, setup, version, apikey, admin
 
 # Import app controllers
 from .app import advanced_prompt_template, app, site, completion, model_config, statistic, conversation, message, generator, audio

api/controllers/console/app/advanced_prompt_template.py

@@ -20,7 +20,6 @@ class AdvancedPromptTemplateList(Resource):
         parser.add_argument('model_name', type=str, required=True, location='args')
         args = parser.parse_args()
 
-        service = AdvancedPromptTemplateService()
-        return service.get_prompt(args)
+        return AdvancedPromptTemplateService.get_prompt(args)
 
 api.add_resource(AdvancedPromptTemplateList, '/app/prompt-templates')

api/controllers/console/app/completion.py

@@ -40,12 +40,14 @@ class CompletionMessageApi(Resource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('model_config', type=dict, required=True, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='dev', location='json')
         args = parser.parse_args()
 
         streaming = args['response_mode'] != 'blocking'
+        args['auto_generate_name'] = False
 
         account = flask_login.current_user
 
@@ -113,6 +115,7 @@ class ChatMessageApi(Resource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('model_config', type=dict, required=True, location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
@@ -120,6 +123,7 @@ class ChatMessageApi(Resource):
         args = parser.parse_args()
 
         streaming = args['response_mode'] != 'blocking'
+        args['auto_generate_name'] = False
 
         account = flask_login.current_user
 

api/controllers/console/app/conversation.py

@@ -108,7 +108,7 @@ class CompletionConversationDetailApi(Resource):
         conversation_id = str(conversation_id)
 
         return _get_conversation(app_id, conversation_id, 'completion')
-    
+
     @setup_required
     @login_required
     @account_initialization_required
@@ -230,7 +230,7 @@ class ChatConversationDetailApi(Resource):
         conversation_id = str(conversation_id)
 
         return _get_conversation(app_id, conversation_id, 'chat')
-    
+
     @setup_required
     @login_required
     @account_initialization_required
@@ -253,8 +253,6 @@ class ChatConversationDetailApi(Resource):
         return {'result': 'success'}, 204
 
 
-
-
 api.add_resource(CompletionConversationApi, '/apps/<uuid:app_id>/completion-conversations')
 api.add_resource(CompletionConversationDetailApi, '/apps/<uuid:app_id>/completion-conversations/<uuid:conversation_id>')
 api.add_resource(ChatConversationApi, '/apps/<uuid:app_id>/chat-conversations')

api/controllers/console/app/message.py

@@ -295,8 +295,8 @@ class MessageSuggestedQuestionApi(Resource):
         try:
             questions = MessageService.get_suggested_questions_after_answer(
                 app_model=app_model,
-                user=current_user,
                 message_id=message_id,
+                user=current_user,
                 check_enabled=False
             )
         except MessageNotExistsError:

api/controllers/console/datasets/data_source.py

@@ -1,7 +1,6 @@
 import datetime
 import json
 
-from cachetools import TTLCache
 from flask import request
 from flask_login import current_user
 from libs.login import login_required
@@ -20,8 +19,6 @@ from models.source import DataSourceBinding
 from services.dataset_service import DatasetService, DocumentService
 from tasks.document_indexing_sync_task import document_indexing_sync_task
 
-cache = TTLCache(maxsize=None, ttl=30)
-
 
 class DataSourceApi(Resource):
 

api/controllers/console/datasets/file.py

@@ -1,5 +1,5 @@
-from cachetools import TTLCache
 from flask import request, current_app
+from flask_login import current_user
 
 import services
 from libs.login import login_required
@@ -15,9 +15,6 @@ from fields.file_fields import upload_config_fields, file_fields
 
 from services.file_service import FileService
 
-cache = TTLCache(maxsize=None, ttl=30)
-
-ALLOWED_EXTENSIONS = ['txt', 'markdown', 'md', 'pdf', 'html', 'htm', 'xlsx', 'docx', 'csv']
 PREVIEW_WORDS_LIMIT = 3000
 
 
@@ -30,9 +27,11 @@ class FileApi(Resource):
     def get(self):
         file_size_limit = current_app.config.get("UPLOAD_FILE_SIZE_LIMIT")
         batch_count_limit = current_app.config.get("UPLOAD_FILE_BATCH_LIMIT")
+        image_file_size_limit = current_app.config.get("UPLOAD_IMAGE_FILE_SIZE_LIMIT")
         return {
             'file_size_limit': file_size_limit,
-            'batch_count_limit': batch_count_limit
+            'batch_count_limit': batch_count_limit,
+            'image_file_size_limit': image_file_size_limit
         }, 200
 
     @setup_required
@@ -51,7 +50,7 @@ class FileApi(Resource):
         if len(request.files) > 1:
             raise TooManyFilesError()
         try:
-            upload_file = FileService.upload_file(file)
+            upload_file = FileService.upload_file(file, current_user)
         except services.errors.file.FileTooLargeError as file_too_large_error:
             raise FileTooLargeError(file_too_large_error.description)
         except services.errors.file.UnsupportedFileTypeError:

api/controllers/console/explore/completion.py

@@ -1,6 +1,7 @@
 # -*- coding:utf-8 -*-
 import json
 import logging
+from datetime import datetime
 from typing import Generator, Union
 
 from flask import Response, stream_with_context
@@ -17,6 +18,7 @@ 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 extensions.ext_database import db
 from libs.helper import uuid_value
 from services.completion_service import CompletionService
 
@@ -32,11 +34,16 @@ class CompletionApi(InstalledAppResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='explore_app', location='json')
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
+
+        installed_app.last_used_at = datetime.utcnow()
+        db.session.commit()
 
         try:
             response = CompletionService.completion(
@@ -91,12 +98,17 @@ class ChatApi(InstalledAppResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='explore_app', location='json')
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
+
+        installed_app.last_used_at = datetime.utcnow()
+        db.session.commit()
 
         try:
             response = CompletionService.completion(

api/controllers/console/explore/conversation.py

@@ -38,7 +38,8 @@ class ConversationListApi(InstalledAppResource):
                 user=current_user,
                 last_id=args['last_id'],
                 limit=args['limit'],
-                pinned=pinned
+                pinned=pinned,
+                exclude_debug_conversation=True
             )
         except LastConversationNotExistsError:
             raise NotFound("Last Conversation Not Exists.")
@@ -71,11 +72,18 @@ class ConversationRenameApi(InstalledAppResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default='False', location='json')
         args = parser.parse_args()
 
         try:
-            return ConversationService.rename(app_model, conversation_id, current_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                current_user,
+                args['name'],
+                args['auto_generate']
+            )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/console/explore/installed_app.py

@@ -39,8 +39,9 @@ class InstalledAppsListApi(Resource):
             }
             for installed_app in installed_apps
         ]
-        installed_apps.sort(key=lambda app: (-app['is_pinned'], app['last_used_at']
-                            if app['last_used_at'] is not None else datetime.min))
+        installed_apps.sort(key=lambda app: (-app['is_pinned'],
+                                             app['last_used_at'] is None,
+                                             -app['last_used_at'].timestamp() if app['last_used_at'] is not None else 0))
 
         return {'installed_apps': installed_apps}
 

api/controllers/console/explore/parameter.py

@@ -1,5 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_restful import marshal_with, fields
+from flask import current_app
 
 from controllers.console import api
 from controllers.console.explore.wraps import InstalledAppResource
@@ -19,6 +20,10 @@ class AppParameterApi(InstalledAppResource):
         'options': fields.List(fields.String)
     }
 
+    system_parameters_fields = {
+        'image_file_size_limit': fields.String
+    }
+
     parameters_fields = {
         'opening_statement': fields.String,
         'suggested_questions': fields.Raw,
@@ -27,6 +32,9 @@ class AppParameterApi(InstalledAppResource):
         'retriever_resource': fields.Raw,
         'more_like_this': fields.Raw,
         'user_input_form': fields.Raw,
+        'sensitive_word_avoidance': fields.Raw,
+        'file_upload': fields.Raw,
+        'system_parameters': fields.Nested(system_parameters_fields)
     }
 
     @marshal_with(parameters_fields)
@@ -42,7 +50,12 @@ class AppParameterApi(InstalledAppResource):
             'speech_to_text': app_model_config.speech_to_text_dict,
             'retriever_resource': app_model_config.retriever_resource_dict,
             'more_like_this': app_model_config.more_like_this_dict,
-            'user_input_form': app_model_config.user_input_form_list
+            'user_input_form': app_model_config.user_input_form_list,
+            'sensitive_word_avoidance': app_model_config.sensitive_word_avoidance_dict,
+            'file_upload': app_model_config.file_upload_dict,
+            'system_parameters': {
+                'image_file_size_limit': current_app.config.get('UPLOAD_IMAGE_FILE_SIZE_LIMIT')
+            }
         }
 
 

api/controllers/console/explore/saved_message.py

@@ -9,6 +9,7 @@ from controllers.console.explore.wraps import InstalledAppResource
 from libs.helper import uuid_value, TimestampField
 from services.errors.message import MessageNotExistsError
 from services.saved_message_service import SavedMessageService
+from fields.conversation_fields import message_file_fields
 
 feedback_fields = {
     'rating': fields.String
@@ -19,6 +20,7 @@ message_fields = {
     'inputs': fields.Raw,
     'query': fields.String,
     'answer': fields.String,
+    'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
     'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
     'created_at': TimestampField
 }

api/controllers/console/extension.py

@@ -0,0 +1,114 @@
+from flask_restful import Resource, reqparse, marshal_with
+from flask_login import current_user
+
+from controllers.console import api
+from controllers.console.setup import setup_required
+from controllers.console.wraps import account_initialization_required
+from libs.login import login_required
+from models.api_based_extension import APIBasedExtension
+from fields.api_based_extension_fields import api_based_extension_fields
+from services.code_based_extension_service import CodeBasedExtensionService
+from services.api_based_extension_service import APIBasedExtensionService
+
+
+class CodeBasedExtensionAPI(Resource):
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument('module', type=str, required=True, location='args')
+        args = parser.parse_args()
+
+        return {
+            'module': args['module'],
+            'data': CodeBasedExtensionService.get_code_based_extension(args['module'])
+        }
+
+
+class APIBasedExtensionAPI(Resource):
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def get(self):
+        tenant_id = current_user.current_tenant_id
+        return APIBasedExtensionService.get_all_by_tenant_id(tenant_id)
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def post(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('api_endpoint', type=str, required=True, location='json')
+        parser.add_argument('api_key', type=str, required=True, location='json')
+        args = parser.parse_args()
+
+        extension_data = APIBasedExtension(
+            tenant_id=current_user.current_tenant_id,
+            name=args['name'],
+            api_endpoint=args['api_endpoint'],
+            api_key=args['api_key']
+        )
+
+        return APIBasedExtensionService.save(extension_data)
+
+
+class APIBasedExtensionDetailAPI(Resource):
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def get(self, id):
+        api_based_extension_id = str(id)
+        tenant_id = current_user.current_tenant_id
+
+        return APIBasedExtensionService.get_with_tenant_id(tenant_id, api_based_extension_id)
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def post(self, id):
+        api_based_extension_id = str(id)
+        tenant_id = current_user.current_tenant_id
+
+        extension_data_from_db = APIBasedExtensionService.get_with_tenant_id(tenant_id, api_based_extension_id)
+
+        parser = reqparse.RequestParser()
+        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('api_endpoint', type=str, required=True, location='json')
+        parser.add_argument('api_key', type=str, required=True, location='json')
+        args = parser.parse_args()
+
+        extension_data_from_db.name = args['name']
+        extension_data_from_db.api_endpoint = args['api_endpoint']
+
+        if args['api_key'] != '[__HIDDEN__]':
+            extension_data_from_db.api_key = args['api_key']
+
+        return APIBasedExtensionService.save(extension_data_from_db)
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def delete(self, id):
+        api_based_extension_id = str(id)
+        tenant_id = current_user.current_tenant_id
+
+        extension_data_from_db = APIBasedExtensionService.get_with_tenant_id(tenant_id, api_based_extension_id)
+
+        APIBasedExtensionService.delete(extension_data_from_db)
+
+        return {'result': 'success'}
+
+
+api.add_resource(CodeBasedExtensionAPI, '/code-based-extension')
+
+api.add_resource(APIBasedExtensionAPI, '/api-based-extension')
+api.add_resource(APIBasedExtensionDetailAPI, '/api-based-extension/<uuid:id>')

api/controllers/console/universal_chat/chat.py

@@ -25,6 +25,7 @@ class UniversalChatApi(UniversalChatResource):
 
         parser = reqparse.RequestParser()
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('provider', type=str, required=True, location='json')
         parser.add_argument('model', type=str, required=True, location='json')
@@ -60,6 +61,8 @@ class UniversalChatApi(UniversalChatResource):
         del args['model']
         del args['tools']
 
+        args['auto_generate_name'] = False
+
         try:
             response = CompletionService.completion(
                 app_model=app_model,

api/controllers/console/universal_chat/conversation.py

@@ -65,11 +65,18 @@ class UniversalChatConversationRenameApi(UniversalChatResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default='False', location='json')
         args = parser.parse_args()
 
         try:
-            return ConversationService.rename(app_model, conversation_id, current_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                current_user,
+                args['name'],
+                args['auto_generate']
+            )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/files/__init__.py

@@ -0,0 +1,10 @@
+# -*- coding:utf-8 -*-
+from flask import Blueprint
+
+from libs.external_api import ExternalApi
+
+bp = Blueprint('files', __name__)
+api = ExternalApi(bp)
+
+
+from . import image_preview

api/controllers/files/image_preview.py

@@ -0,0 +1,40 @@
+from flask import request, Response
+from flask_restful import Resource
+
+import services
+from controllers.files import api
+from libs.exception import BaseHTTPException
+from services.file_service import FileService
+
+
+class ImagePreviewApi(Resource):
+    def get(self, file_id):
+        file_id = str(file_id)
+
+        timestamp = request.args.get('timestamp')
+        nonce = request.args.get('nonce')
+        sign = request.args.get('sign')
+
+        if not timestamp or not nonce or not sign:
+            return {'content': 'Invalid request.'}, 400
+
+        try:
+            generator, mimetype = FileService.get_image_preview(
+                file_id,
+                timestamp,
+                nonce,
+                sign
+            )
+        except services.errors.file.UnsupportedFileTypeError:
+            raise UnsupportedFileTypeError()
+
+        return Response(generator, mimetype=mimetype)
+
+
+api.add_resource(ImagePreviewApi, '/files/<uuid:file_id>/image-preview')
+
+
+class UnsupportedFileTypeError(BaseHTTPException):
+    error_code = 'unsupported_file_type'
+    description = "File type not allowed."
+    code = 415

api/controllers/service_api/__init__.py

@@ -7,6 +7,6 @@ bp = Blueprint('service_api', __name__, url_prefix='/v1')
 api = ExternalApi(bp)
 
 
-from .app import completion, app, conversation, message, audio
+from .app import completion, app, conversation, message, audio, file
 
 from .dataset import document, segment, dataset

api/controllers/service_api/app/app.py

@@ -1,5 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_restful import fields, marshal_with
+from flask import current_app
 
 from controllers.service_api import api
 from controllers.service_api.wraps import AppApiResource
@@ -20,6 +21,10 @@ class AppParameterApi(AppApiResource):
         'options': fields.List(fields.String)
     }
 
+    system_parameters_fields = {
+        'image_file_size_limit': fields.String
+    }
+
     parameters_fields = {
         'opening_statement': fields.String,
         'suggested_questions': fields.Raw,
@@ -28,6 +33,9 @@ class AppParameterApi(AppApiResource):
         'retriever_resource': fields.Raw,
         'more_like_this': fields.Raw,
         'user_input_form': fields.Raw,
+        'sensitive_word_avoidance': fields.Raw,
+        'file_upload': fields.Raw,
+        'system_parameters': fields.Nested(system_parameters_fields)
     }
 
     @marshal_with(parameters_fields)
@@ -42,7 +50,12 @@ class AppParameterApi(AppApiResource):
             'speech_to_text': app_model_config.speech_to_text_dict,
             'retriever_resource': app_model_config.retriever_resource_dict,
             'more_like_this': app_model_config.more_like_this_dict,
-            'user_input_form': app_model_config.user_input_form_list
+            'user_input_form': app_model_config.user_input_form_list,
+            'sensitive_word_avoidance': app_model_config.sensitive_word_avoidance_dict,
+            'file_upload': app_model_config.file_upload_dict,
+            'system_parameters': {
+                'image_file_size_limit': current_app.config.get('UPLOAD_IMAGE_FILE_SIZE_LIMIT')
+            }
         }
 
 

api/controllers/service_api/app/completion.py

@@ -28,6 +28,7 @@ class CompletionApi(AppApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('user', type=str, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='dev', location='json')
@@ -39,13 +40,15 @@ class CompletionApi(AppApiResource):
         if end_user is None and args['user'] is not None:
             end_user = create_or_update_end_user_for_user_id(app_model, args['user'])
 
+        args['auto_generate_name'] = False
+
         try:
             response = CompletionService.completion(
                 app_model=app_model,
                 user=end_user,
                 args=args,
                 from_source='api',
-                streaming=streaming
+                streaming=streaming,
             )
 
             return compact_response(response)
@@ -90,10 +93,12 @@ class ChatApi(AppApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('user', type=str, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='dev', location='json')
+        parser.add_argument('auto_generate_name', type=bool, required=False, default='True', location='json')
 
         args = parser.parse_args()
 
@@ -183,4 +188,3 @@ api.add_resource(CompletionApi, '/completion-messages')
 api.add_resource(CompletionStopApi, '/completion-messages/<string:task_id>/stop')
 api.add_resource(ChatApi, '/chat-messages')
 api.add_resource(ChatStopApi, '/chat-messages/<string:task_id>/stop')
-

api/controllers/service_api/app/conversation.py

@@ -54,6 +54,7 @@ class ConversationDetailApi(AppApiResource):
             raise NotFound("Conversation Not Exists.")
         return {"result": "success"}, 204
 
+
 class ConversationRenameApi(AppApiResource):
 
     @marshal_with(simple_conversation_fields)
@@ -64,15 +65,22 @@ class ConversationRenameApi(AppApiResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
         parser.add_argument('user', type=str, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default='False', location='json')
         args = parser.parse_args()
 
         if end_user is None and args['user'] is not None:
             end_user = create_or_update_end_user_for_user_id(app_model, args['user'])
 
         try:
-            return ConversationService.rename(app_model, conversation_id, end_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                end_user,
+                args['name'],
+                args['auto_generate']
+            )
         except services.errors.conversation.ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/service_api/app/error.py

@@ -75,3 +75,26 @@ class ProviderNotSupportSpeechToTextError(BaseHTTPException):
     description = "Provider not support speech to text."
     code = 400
 
+
+class NoFileUploadedError(BaseHTTPException):
+    error_code = 'no_file_uploaded'
+    description = "Please upload your file."
+    code = 400
+
+
+class TooManyFilesError(BaseHTTPException):
+    error_code = 'too_many_files'
+    description = "Only one file is allowed."
+    code = 400
+
+
+class FileTooLargeError(BaseHTTPException):
+    error_code = 'file_too_large'
+    description = "File size exceeded. {message}"
+    code = 413
+
+
+class UnsupportedFileTypeError(BaseHTTPException):
+    error_code = 'unsupported_file_type'
+    description = "File type not allowed."
+    code = 415

api/controllers/service_api/app/file.py

@@ -0,0 +1,42 @@
+from flask import request
+from flask_restful import marshal_with
+
+from controllers.service_api import api
+from controllers.service_api.wraps import AppApiResource
+from controllers.service_api.app import create_or_update_end_user_for_user_id
+from controllers.service_api.app.error import NoFileUploadedError, TooManyFilesError, FileTooLargeError, \
+    UnsupportedFileTypeError
+import services
+from services.file_service import FileService
+from fields.file_fields import file_fields
+
+
+class FileApi(AppApiResource):
+
+    @marshal_with(file_fields)
+    def post(self, app_model, end_user):
+
+        file = request.files['file']
+        user_args = request.form.get('user')
+
+        if end_user is None and user_args is not None:
+            end_user = create_or_update_end_user_for_user_id(app_model, user_args)
+
+        # check file
+        if 'file' not in request.files:
+            raise NoFileUploadedError()
+
+        if len(request.files) > 1:
+            raise TooManyFilesError()
+
+        try:
+            upload_file = FileService.upload_file(file, end_user)
+        except services.errors.file.FileTooLargeError as file_too_large_error:
+            raise FileTooLargeError(file_too_large_error.description)
+        except services.errors.file.UnsupportedFileTypeError:
+            raise UnsupportedFileTypeError()
+
+        return upload_file, 201
+
+
+api.add_resource(FileApi, '/files/upload')

api/controllers/service_api/app/message.py

@@ -10,7 +10,9 @@ from controllers.service_api.app.error import NotChatAppError
 from controllers.service_api.wraps import AppApiResource
 from libs.helper import TimestampField, uuid_value
 from services.message_service import MessageService
-
+from extensions.ext_database import db
+from models.model import Message, EndUser
+from fields.conversation_fields import message_file_fields
 
 class MessageListApi(AppApiResource):
     feedback_fields = {
@@ -41,6 +43,7 @@ class MessageListApi(AppApiResource):
         'inputs': fields.Raw,
         'query': fields.String,
         'answer': fields.String,
+        'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
         'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
         'retriever_resources': fields.List(fields.Nested(retriever_resource_fields)),
         'created_at': TimestampField
@@ -96,5 +99,38 @@ class MessageFeedbackApi(AppApiResource):
         return {'result': 'success'}
 
 
+class MessageSuggestedApi(AppApiResource):
+    def get(self, app_model, end_user, message_id):
+        message_id = str(message_id)
+        if app_model.mode != 'chat':
+            raise NotChatAppError()
+        try:
+            message = db.session.query(Message).filter(
+                Message.id == message_id,
+                Message.app_id == app_model.id,
+            ).first()
+
+            if end_user is None and message.from_end_user_id is not None:
+                user = db.session.query(EndUser) \
+                    .filter(
+                        EndUser.tenant_id == app_model.tenant_id,
+                        EndUser.id == message.from_end_user_id,
+                        EndUser.type == 'service_api'
+                    ).first()
+            else:
+                user = end_user
+            questions = MessageService.get_suggested_questions_after_answer(
+                app_model=app_model,
+                user=user,
+                message_id=message_id,
+                check_enabled=False
+            )
+        except services.errors.message.MessageNotExistsError:
+            raise NotFound("Message Not Exists.")
+
+        return {'result': 'success', 'data': questions}
+
+
 api.add_resource(MessageListApi, '/messages')
 api.add_resource(MessageFeedbackApi, '/messages/<uuid:message_id>/feedbacks')
+api.add_resource(MessageSuggestedApi, '/messages/<uuid:message_id>/suggested')

api/controllers/service_api/dataset/document.py

@@ -2,6 +2,7 @@ import json
 
 from flask import request
 from flask_restful import reqparse, marshal
+from flask_login import current_user
 from sqlalchemy import desc
 from werkzeug.exceptions import NotFound
 
@@ -173,7 +174,7 @@ class DocumentAddByFileApi(DatasetApiResource):
         if len(request.files) > 1:
             raise TooManyFilesError()
 
-        upload_file = FileService.upload_file(file)
+        upload_file = FileService.upload_file(file, current_user)
         data_source = {
             'type': 'upload_file',
             'info_list': {
@@ -235,7 +236,7 @@ class DocumentUpdateByFileApi(DatasetApiResource):
             if len(request.files) > 1:
                 raise TooManyFilesError()
 
-            upload_file = FileService.upload_file(file)
+            upload_file = FileService.upload_file(file, current_user)
             data_source = {
                 'type': 'upload_file',
                 'info_list': {

api/controllers/web/__init__.py

@@ -7,4 +7,4 @@ bp = Blueprint('web', __name__, url_prefix='/api')
 api = ExternalApi(bp)
 
 
-from . import completion, app, conversation, message, site, saved_message, audio, passport
+from . import completion, app, conversation, message, site, saved_message, audio, passport, file

api/controllers/web/app.py

@@ -1,5 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_restful import marshal_with, fields
+from flask import current_app
 
 from controllers.web import api
 from controllers.web.wraps import WebApiResource
@@ -19,6 +20,10 @@ class AppParameterApi(WebApiResource):
         'options': fields.List(fields.String)
     }
 
+    system_parameters_fields = {
+        'image_file_size_limit': fields.String
+    }
+
     parameters_fields = {
         'opening_statement': fields.String,
         'suggested_questions': fields.Raw,
@@ -27,6 +32,9 @@ class AppParameterApi(WebApiResource):
         'retriever_resource': fields.Raw,
         'more_like_this': fields.Raw,
         'user_input_form': fields.Raw,
+        'sensitive_word_avoidance': fields.Raw,
+        'file_upload': fields.Raw,
+        'system_parameters': fields.Nested(system_parameters_fields)
     }
 
     @marshal_with(parameters_fields)
@@ -41,7 +49,12 @@ class AppParameterApi(WebApiResource):
             'speech_to_text': app_model_config.speech_to_text_dict,
             'retriever_resource': app_model_config.retriever_resource_dict,
             'more_like_this': app_model_config.more_like_this_dict,
-            'user_input_form': app_model_config.user_input_form_list
+            'user_input_form': app_model_config.user_input_form_list,
+            'sensitive_word_avoidance': app_model_config.sensitive_word_avoidance_dict,
+            'file_upload': app_model_config.file_upload_dict,
+            'system_parameters': {
+                'image_file_size_limit': current_app.config.get('UPLOAD_IMAGE_FILE_SIZE_LIMIT')
+            }
         }
 
 

api/controllers/web/completion.py

@@ -30,12 +30,14 @@ class CompletionApi(WebApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='web_app', location='json')
 
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
 
         try:
             response = CompletionService.completion(
@@ -88,6 +90,7 @@ class ChatApi(WebApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='web_app', location='json')
@@ -95,6 +98,7 @@ class ChatApi(WebApiResource):
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
 
         try:
             response = CompletionService.completion(
@@ -139,7 +143,7 @@ class ChatStopApi(WebApiResource):
         return {'result': 'success'}, 200
 
 
-def compact_response(response: Union[dict | Generator]) -> Response:
+def compact_response(response: Union[dict, Generator]) -> Response:
     if isinstance(response, dict):
         return Response(response=json.dumps(response), status=200, mimetype='application/json')
     else:

api/controllers/web/conversation.py

@@ -67,11 +67,18 @@ class ConversationRenameApi(WebApiResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default='False', location='json')
         args = parser.parse_args()
 
         try:
-            return ConversationService.rename(app_model, conversation_id, end_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                end_user,
+                args['name'],
+                args['auto_generate']
+            )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/web/error.py

@@ -85,4 +85,28 @@ class UnsupportedAudioTypeError(BaseHTTPException):
 class ProviderNotSupportSpeechToTextError(BaseHTTPException):
     error_code = 'provider_not_support_speech_to_text'
     description = "Provider not support speech to text."
-    code = 400
+    code = 400
+
+
+class NoFileUploadedError(BaseHTTPException):
+    error_code = 'no_file_uploaded'
+    description = "Please upload your file."
+    code = 400
+
+
+class TooManyFilesError(BaseHTTPException):
+    error_code = 'too_many_files'
+    description = "Only one file is allowed."
+    code = 400
+
+
+class FileTooLargeError(BaseHTTPException):
+    error_code = 'file_too_large'
+    description = "File size exceeded. {message}"
+    code = 413
+
+
+class UnsupportedFileTypeError(BaseHTTPException):
+    error_code = 'unsupported_file_type'
+    description = "File type not allowed."
+    code = 415

api/controllers/web/file.py

@@ -0,0 +1,36 @@
+from flask import request
+from flask_restful import marshal_with
+
+from controllers.web import api
+from controllers.web.wraps import WebApiResource
+from controllers.web.error import NoFileUploadedError, TooManyFilesError, FileTooLargeError, \
+    UnsupportedFileTypeError
+import services
+from services.file_service import FileService
+from fields.file_fields import file_fields
+
+
+class FileApi(WebApiResource):
+
+    @marshal_with(file_fields)
+    def post(self, app_model, end_user):
+        # get file from request
+        file = request.files['file']
+
+        # check file
+        if 'file' not in request.files:
+            raise NoFileUploadedError()
+
+        if len(request.files) > 1:
+            raise TooManyFilesError()
+        try:
+            upload_file = FileService.upload_file(file, end_user)
+        except services.errors.file.FileTooLargeError as file_too_large_error:
+            raise FileTooLargeError(file_too_large_error.description)
+        except services.errors.file.UnsupportedFileTypeError:
+            raise UnsupportedFileTypeError()
+
+        return upload_file, 201
+
+
+api.add_resource(FileApi, '/files/upload')

api/controllers/web/message.py

@@ -22,6 +22,7 @@ from services.errors.app import MoreLikeThisDisabledError
 from services.errors.conversation import ConversationNotExistsError
 from services.errors.message import MessageNotExistsError, SuggestedQuestionsAfterAnswerDisabledError
 from services.message_service import MessageService
+from fields.conversation_fields import message_file_fields
 
 
 class MessageListApi(WebApiResource):
@@ -54,6 +55,7 @@ class MessageListApi(WebApiResource):
         'inputs': fields.Raw,
         'query': fields.String,
         'answer': fields.String,
+        'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
         'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
         'retriever_resources': fields.List(fields.Nested(retriever_resource_fields)),
         'created_at': TimestampField

api/controllers/web/saved_message.py

@@ -8,6 +8,8 @@ from controllers.web.wraps import WebApiResource
 from libs.helper import uuid_value, TimestampField
 from services.errors.message import MessageNotExistsError
 from services.saved_message_service import SavedMessageService
+from fields.conversation_fields import message_file_fields
+
 
 feedback_fields = {
     'rating': fields.String
@@ -18,6 +20,7 @@ message_fields = {
     'inputs': fields.Raw,
     'query': fields.String,
     'answer': fields.String,
+    'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
     'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
     'created_at': TimestampField
 }

api/core/__init__.py

@@ -0,0 +1 @@
+import core.moderation.base

api/core/agent/agent/multi_dataset_router_agent.py

@@ -76,7 +76,7 @@ class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
             agent_decision = self.real_plan(intermediate_steps, callbacks, **kwargs)
             if isinstance(agent_decision, AgentAction):
                 tool_inputs = agent_decision.tool_input
-                if isinstance(tool_inputs, dict) and 'query' in tool_inputs:
+                if isinstance(tool_inputs, dict) and 'query' in tool_inputs and 'chat_history' not in kwargs:
                     tool_inputs['query'] = kwargs['input']
                     agent_decision.tool_input = tool_inputs
             else:

api/core/agent/agent/structed_multi_dataset_router_agent.py

@@ -1,7 +1,7 @@
 import re
 from typing import List, Tuple, Any, Union, Sequence, Optional, cast
 
-from langchain import BasePromptTemplate
+from langchain import BasePromptTemplate, PromptTemplate
 from langchain.agents import StructuredChatAgent, AgentOutputParser, Agent
 from langchain.agents.structured_chat.base import HUMAN_MESSAGE_TEMPLATE
 from langchain.callbacks.base import BaseCallbackManager
@@ -12,6 +12,7 @@ from langchain.tools import BaseTool
 from langchain.agents.structured_chat.prompt import PREFIX, SUFFIX
 
 from core.chain.llm_chain import LLMChain
+from core.model_providers.models.entity.model_params import ModelMode
 from core.model_providers.models.llm.base import BaseLLM
 from core.tool.dataset_retriever_tool import DatasetRetrieverTool
 
@@ -92,6 +93,10 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
             rst = tool.run(tool_input={'query': kwargs['input']})
             return AgentFinish(return_values={"output": rst}, log=rst)
 
+        if intermediate_steps:
+            _, observation = intermediate_steps[-1]
+            return AgentFinish(return_values={"output": observation}, log=observation)
+
         full_inputs = self.get_full_inputs(intermediate_steps, **kwargs)
 
         try:
@@ -107,6 +112,8 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
                 if isinstance(tool_inputs, dict) and 'query' in tool_inputs:
                     tool_inputs['query'] = kwargs['input']
                     agent_decision.tool_input = tool_inputs
+                elif isinstance(tool_inputs, str):
+                    agent_decision.tool_input = kwargs['input']
             else:
                 agent_decision.return_values['output'] = ''
             return agent_decision
@@ -143,6 +150,61 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
         ]
         return ChatPromptTemplate(input_variables=input_variables, messages=messages)
 
+    @classmethod
+    def create_completion_prompt(
+            cls,
+            tools: Sequence[BaseTool],
+            prefix: str = PREFIX,
+            format_instructions: str = FORMAT_INSTRUCTIONS,
+            input_variables: Optional[List[str]] = None,
+    ) -> PromptTemplate:
+        """Create prompt in the style of the zero shot agent.
+
+        Args:
+            tools: List of tools the agent will have access to, used to format the
+                prompt.
+            prefix: String to put before the list of tools.
+            input_variables: List of input variables the final prompt will expect.
+
+        Returns:
+            A PromptTemplate with the template assembled from the pieces here.
+        """
+        suffix = """Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.
+Question: {input}
+Thought: {agent_scratchpad}
+"""
+
+        tool_strings = "\n".join([f"{tool.name}: {tool.description}" for tool in tools])
+        tool_names = ", ".join([tool.name for tool in tools])
+        format_instructions = format_instructions.format(tool_names=tool_names)
+        template = "\n\n".join([prefix, tool_strings, format_instructions, suffix])
+        if input_variables is None:
+            input_variables = ["input", "agent_scratchpad"]
+        return PromptTemplate(template=template, input_variables=input_variables)
+
+    def _construct_scratchpad(
+        self, intermediate_steps: List[Tuple[AgentAction, str]]
+    ) -> str:
+        agent_scratchpad = ""
+        for action, observation in intermediate_steps:
+            agent_scratchpad += action.log
+            agent_scratchpad += f"\n{self.observation_prefix}{observation}\n{self.llm_prefix}"
+
+        if not isinstance(agent_scratchpad, str):
+            raise ValueError("agent_scratchpad should be of type string.")
+        if agent_scratchpad:
+            llm_chain = cast(LLMChain, self.llm_chain)
+            if llm_chain.model_instance.model_mode == ModelMode.CHAT:
+                return (
+                    f"This was your previous work "
+                    f"(but I haven't seen any of it! I only see what "
+                    f"you return as final answer):\n{agent_scratchpad}"
+                )
+            else:
+                return agent_scratchpad
+        else:
+            return agent_scratchpad
+
     @classmethod
     def from_llm_and_tools(
             cls,
@@ -160,15 +222,23 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
     ) -> Agent:
         """Construct an agent from an LLM and tools."""
         cls._validate_tools(tools)
-        prompt = cls.create_prompt(
-            tools,
-            prefix=prefix,
-            suffix=suffix,
-            human_message_template=human_message_template,
-            format_instructions=format_instructions,
-            input_variables=input_variables,
-            memory_prompts=memory_prompts,
-        )
+        if model_instance.model_mode == ModelMode.CHAT:
+            prompt = cls.create_prompt(
+                tools,
+                prefix=prefix,
+                suffix=suffix,
+                human_message_template=human_message_template,
+                format_instructions=format_instructions,
+                input_variables=input_variables,
+                memory_prompts=memory_prompts,
+            )
+        else:
+            prompt = cls.create_completion_prompt(
+                tools,
+                prefix=prefix,
+                format_instructions=format_instructions,
+                input_variables=input_variables
+            )
         llm_chain = LLMChain(
             model_instance=model_instance,
             prompt=prompt,

api/core/agent/agent/structured_chat.py

@@ -1,7 +1,7 @@
 import re
-from typing import List, Tuple, Any, Union, Sequence, Optional
+from typing import List, Tuple, Any, Union, Sequence, Optional, cast
 
-from langchain import BasePromptTemplate
+from langchain import BasePromptTemplate, PromptTemplate
 from langchain.agents import StructuredChatAgent, AgentOutputParser, Agent
 from langchain.agents.structured_chat.base import HUMAN_MESSAGE_TEMPLATE
 from langchain.callbacks.base import BaseCallbackManager
@@ -15,6 +15,7 @@ from langchain.agents.structured_chat.prompt import PREFIX, SUFFIX
 
 from core.agent.agent.calc_token_mixin import CalcTokenMixin, ExceededLLMTokensLimitError
 from core.chain.llm_chain import LLMChain
+from core.model_providers.models.entity.model_params import ModelMode
 from core.model_providers.models.llm.base import BaseLLM
 
 FORMAT_INSTRUCTIONS = """Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).
@@ -184,6 +185,61 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
         ]
         return ChatPromptTemplate(input_variables=input_variables, messages=messages)
 
+    @classmethod
+    def create_completion_prompt(
+            cls,
+            tools: Sequence[BaseTool],
+            prefix: str = PREFIX,
+            format_instructions: str = FORMAT_INSTRUCTIONS,
+            input_variables: Optional[List[str]] = None,
+    ) -> PromptTemplate:
+        """Create prompt in the style of the zero shot agent.
+
+        Args:
+            tools: List of tools the agent will have access to, used to format the
+                prompt.
+            prefix: String to put before the list of tools.
+            input_variables: List of input variables the final prompt will expect.
+
+        Returns:
+            A PromptTemplate with the template assembled from the pieces here.
+        """
+        suffix = """Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.
+Question: {input}
+Thought: {agent_scratchpad}
+"""
+
+        tool_strings = "\n".join([f"{tool.name}: {tool.description}" for tool in tools])
+        tool_names = ", ".join([tool.name for tool in tools])
+        format_instructions = format_instructions.format(tool_names=tool_names)
+        template = "\n\n".join([prefix, tool_strings, format_instructions, suffix])
+        if input_variables is None:
+            input_variables = ["input", "agent_scratchpad"]
+        return PromptTemplate(template=template, input_variables=input_variables)
+
+    def _construct_scratchpad(
+        self, intermediate_steps: List[Tuple[AgentAction, str]]
+    ) -> str:
+        agent_scratchpad = ""
+        for action, observation in intermediate_steps:
+            agent_scratchpad += action.log
+            agent_scratchpad += f"\n{self.observation_prefix}{observation}\n{self.llm_prefix}"
+
+        if not isinstance(agent_scratchpad, str):
+            raise ValueError("agent_scratchpad should be of type string.")
+        if agent_scratchpad:
+            llm_chain = cast(LLMChain, self.llm_chain)
+            if llm_chain.model_instance.model_mode == ModelMode.CHAT:
+                return (
+                    f"This was your previous work "
+                    f"(but I haven't seen any of it! I only see what "
+                    f"you return as final answer):\n{agent_scratchpad}"
+                )
+            else:
+                return agent_scratchpad
+        else:
+            return agent_scratchpad
+
     @classmethod
     def from_llm_and_tools(
             cls,
@@ -201,15 +257,23 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
     ) -> Agent:
         """Construct an agent from an LLM and tools."""
         cls._validate_tools(tools)
-        prompt = cls.create_prompt(
-            tools,
-            prefix=prefix,
-            suffix=suffix,
-            human_message_template=human_message_template,
-            format_instructions=format_instructions,
-            input_variables=input_variables,
-            memory_prompts=memory_prompts,
-        )
+        if model_instance.model_mode == ModelMode.CHAT:
+            prompt = cls.create_prompt(
+                tools,
+                prefix=prefix,
+                suffix=suffix,
+                human_message_template=human_message_template,
+                format_instructions=format_instructions,
+                input_variables=input_variables,
+                memory_prompts=memory_prompts,
+            )
+        else:
+            prompt = cls.create_completion_prompt(
+                tools,
+                prefix=prefix,
+                format_instructions=format_instructions,
+                input_variables=input_variables,
+            )
         llm_chain = LLMChain(
             model_instance=model_instance,
             prompt=prompt,

api/core/callback_handler/llm_callback_handler.py

@@ -1,13 +1,26 @@
 import logging
-from typing import Any, Dict, List, Union
+import threading
+import time
+from typing import Any, Dict, List, Union, Optional
 
+from flask import Flask, current_app
 from langchain.callbacks.base import BaseCallbackHandler
 from langchain.schema import LLMResult, BaseMessage
+from pydantic import BaseModel
 
 from core.callback_handler.entity.llm_message import LLMMessage
-from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException
-from core.model_providers.models.entity.message import to_prompt_messages, PromptMessage
+from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException, \
+    ConversationTaskInterruptException
+from core.model_providers.models.entity.message import to_prompt_messages, PromptMessage, LCHumanMessageWithFiles, \
+    ImagePromptMessageFile
 from core.model_providers.models.llm.base import BaseLLM
+from core.moderation.base import ModerationOutputsResult, ModerationAction
+from core.moderation.factory import ModerationFactory
+
+
+class ModerationRule(BaseModel):
+    type: str
+    config: Dict[str, Any]
 
 
 class LLMCallbackHandler(BaseCallbackHandler):
@@ -20,6 +33,24 @@ class LLMCallbackHandler(BaseCallbackHandler):
         self.start_at = None
         self.conversation_message_task = conversation_message_task
 
+        self.output_moderation_handler = None
+        self.init_output_moderation()
+
+    def init_output_moderation(self):
+        app_model_config = self.conversation_message_task.app_model_config
+        sensitive_word_avoidance_dict = app_model_config.sensitive_word_avoidance_dict
+
+        if sensitive_word_avoidance_dict and sensitive_word_avoidance_dict.get("enabled"):
+            self.output_moderation_handler = OutputModerationHandler(
+                tenant_id=self.conversation_message_task.tenant_id,
+                app_id=self.conversation_message_task.app.id,
+                rule=ModerationRule(
+                    type=sensitive_word_avoidance_dict.get("type"),
+                    config=sensitive_word_avoidance_dict.get("config")
+                ),
+                on_message_replace_func=self.conversation_message_task.on_message_replace
+            )
+
     @property
     def always_verbose(self) -> bool:
         """Whether to call verbose callbacks even if verbose is False."""
@@ -42,7 +73,12 @@ class LLMCallbackHandler(BaseCallbackHandler):
 
             real_prompts.append({
                 "role": role,
-                "text": message.content
+                "text": message.content,
+                "files": [{
+                    "type": file.type.value,
+                    "data": file.data[:10] + '...[TRUNCATED]...' + file.data[-10:],
+                    "detail": file.detail.value if isinstance(file, ImagePromptMessageFile) else None,
+                } for file in (message.files if isinstance(message, LCHumanMessageWithFiles) else [])]
             })
 
         self.llm_message.prompt = real_prompts
@@ -59,10 +95,19 @@ class LLMCallbackHandler(BaseCallbackHandler):
         self.llm_message.prompt_tokens = self.model_instance.get_num_tokens([PromptMessage(content=prompts[0])])
 
     def on_llm_end(self, response: LLMResult, **kwargs: Any) -> None:
-        if not self.conversation_message_task.streaming:
-            self.conversation_message_task.append_message_text(response.generations[0][0].text)
+        if self.output_moderation_handler:
+            self.output_moderation_handler.stop_thread()
+
+            self.llm_message.completion = self.output_moderation_handler.moderation_completion(
+                completion=response.generations[0][0].text,
+                public_event=True if self.conversation_message_task.streaming else False
+            )
+        else:
             self.llm_message.completion = response.generations[0][0].text
 
+        if not self.conversation_message_task.streaming:
+            self.conversation_message_task.append_message_text(self.llm_message.completion)
+
         if response.llm_output and 'token_usage' in response.llm_output:
             if 'prompt_tokens' in response.llm_output['token_usage']:
                 self.llm_message.prompt_tokens = response.llm_output['token_usage']['prompt_tokens']
@@ -79,23 +124,161 @@ class LLMCallbackHandler(BaseCallbackHandler):
         self.conversation_message_task.save_message(self.llm_message)
 
     def on_llm_new_token(self, token: str, **kwargs: Any) -> None:
-        try:
-            self.conversation_message_task.append_message_text(token)
-        except ConversationTaskStoppedException as ex:
+        if self.output_moderation_handler and self.output_moderation_handler.should_direct_output():
+            # stop subscribe new token when output moderation should direct output
+            ex = ConversationTaskInterruptException()
             self.on_llm_error(error=ex)
             raise ex
 
-        self.llm_message.completion += token
+        try:
+            self.conversation_message_task.append_message_text(token)
+            self.llm_message.completion += token
+
+            if self.output_moderation_handler:
+                self.output_moderation_handler.append_new_token(token)
+        except ConversationTaskStoppedException as ex:
+            self.on_llm_error(error=ex)
+            raise ex
 
     def on_llm_error(
             self, error: Union[Exception, KeyboardInterrupt], **kwargs: Any
     ) -> None:
         """Do nothing."""
+        if self.output_moderation_handler:
+            self.output_moderation_handler.stop_thread()
+
         if isinstance(error, ConversationTaskStoppedException):
             if self.conversation_message_task.streaming:
                 self.llm_message.completion_tokens = self.model_instance.get_num_tokens(
                     [PromptMessage(content=self.llm_message.completion)]
                 )
                 self.conversation_message_task.save_message(llm_message=self.llm_message, by_stopped=True)
+        if isinstance(error, ConversationTaskInterruptException):
+            self.llm_message.completion = self.output_moderation_handler.get_final_output()
+            self.llm_message.completion_tokens = self.model_instance.get_num_tokens(
+                [PromptMessage(content=self.llm_message.completion)]
+            )
+            self.conversation_message_task.save_message(llm_message=self.llm_message)
         else:
             logging.debug("on_llm_error: %s", error)
+
+
+class OutputModerationHandler(BaseModel):
+    DEFAULT_BUFFER_SIZE: int = 300
+
+    tenant_id: str
+    app_id: str
+
+    rule: ModerationRule
+    on_message_replace_func: Any
+
+    thread: Optional[threading.Thread] = None
+    thread_running: bool = True
+    buffer: str = ''
+    is_final_chunk: bool = False
+    final_output: Optional[str] = None
+
+    class Config:
+        arbitrary_types_allowed = True
+
+    def should_direct_output(self):
+        return self.final_output is not None
+
+    def get_final_output(self):
+        return self.final_output
+
+    def append_new_token(self, token: str):
+        self.buffer += token
+
+        if not self.thread:
+            self.thread = self.start_thread()
+
+    def moderation_completion(self, completion: str, public_event: bool = False) -> str:
+        self.buffer = completion
+        self.is_final_chunk = True
+
+        result = self.moderation(
+            tenant_id=self.tenant_id,
+            app_id=self.app_id,
+            moderation_buffer=completion
+        )
+
+        if not result or not result.flagged:
+            return completion
+
+        if result.action == ModerationAction.DIRECT_OUTPUT:
+            final_output = result.preset_response
+        else:
+            final_output = result.text
+
+        if public_event:
+            self.on_message_replace_func(final_output)
+
+        return final_output
+
+    def start_thread(self) -> threading.Thread:
+        buffer_size = int(current_app.config.get('MODERATION_BUFFER_SIZE', self.DEFAULT_BUFFER_SIZE))
+        thread = threading.Thread(target=self.worker, kwargs={
+            'flask_app': current_app._get_current_object(),
+            'buffer_size': buffer_size if buffer_size > 0 else self.DEFAULT_BUFFER_SIZE
+        })
+
+        thread.start()
+
+        return thread
+
+    def stop_thread(self):
+        if self.thread and self.thread.is_alive():
+            self.thread_running = False
+
+    def worker(self, flask_app: Flask, buffer_size: int):
+        with flask_app.app_context():
+            current_length = 0
+            while self.thread_running:
+                moderation_buffer = self.buffer
+                buffer_length = len(moderation_buffer)
+                if not self.is_final_chunk:
+                    chunk_length = buffer_length - current_length
+                    if 0 <= chunk_length < buffer_size:
+                        time.sleep(1)
+                        continue
+
+                current_length = buffer_length
+
+                result = self.moderation(
+                    tenant_id=self.tenant_id,
+                    app_id=self.app_id,
+                    moderation_buffer=moderation_buffer
+                )
+
+                if not result or not result.flagged:
+                    continue
+
+                if result.action == ModerationAction.DIRECT_OUTPUT:
+                    final_output = result.preset_response
+                    self.final_output = final_output
+                else:
+                    final_output = result.text + self.buffer[len(moderation_buffer):]
+
+                # trigger replace event
+                if self.thread_running:
+                    self.on_message_replace_func(final_output)
+
+                if result.action == ModerationAction.DIRECT_OUTPUT:
+                    break
+
+    def moderation(self, tenant_id: str, app_id: str, moderation_buffer: str) -> Optional[ModerationOutputsResult]:
+        try:
+            moderation_factory = ModerationFactory(
+                name=self.rule.type,
+                app_id=app_id,
+                tenant_id=tenant_id,
+                config=self.rule.config
+            )
+
+            result: ModerationOutputsResult = moderation_factory.moderation_for_outputs(moderation_buffer)
+            return result
+        except Exception as e:
+            logging.error("Moderation Output error: %s", e)
+
+        return None

api/core/chain/sensitive_word_avoidance_chain.py

@@ -1,92 +0,0 @@
-import enum
-import logging
-from typing import List, Dict, Optional, Any
-
-from langchain.callbacks.manager import CallbackManagerForChainRun
-from langchain.chains.base import Chain
-from pydantic import BaseModel
-
-from core.model_providers.error import LLMBadRequestError
-from core.model_providers.model_factory import ModelFactory
-from core.model_providers.models.llm.base import BaseLLM
-from core.model_providers.models.moderation import openai_moderation
-
-
-class SensitiveWordAvoidanceRule(BaseModel):
-    class Type(enum.Enum):
-        MODERATION = "moderation"
-        KEYWORDS = "keywords"
-
-    type: Type
-    canned_response: str = 'Your content violates our usage policy. Please revise and try again.'
-    extra_params: dict = {}
-
-
-class SensitiveWordAvoidanceChain(Chain):
-    input_key: str = "input"  #: :meta private:
-    output_key: str = "output"  #: :meta private:
-
-    model_instance: BaseLLM
-    sensitive_word_avoidance_rule: SensitiveWordAvoidanceRule
-
-    @property
-    def _chain_type(self) -> str:
-        return "sensitive_word_avoidance_chain"
-
-    @property
-    def input_keys(self) -> List[str]:
-        """Expect input key.
-
-        :meta private:
-        """
-        return [self.input_key]
-
-    @property
-    def output_keys(self) -> List[str]:
-        """Return output key.
-
-        :meta private:
-        """
-        return [self.output_key]
-
-    def _check_sensitive_word(self, text: str) -> bool:
-        for word in self.sensitive_word_avoidance_rule.extra_params.get('sensitive_words', []):
-            if word in text:
-                return False
-        return True
-
-    def _check_moderation(self, text: str) -> bool:
-        moderation_model_instance = ModelFactory.get_moderation_model(
-            tenant_id=self.model_instance.model_provider.provider.tenant_id,
-            model_provider_name='openai',
-            model_name=openai_moderation.DEFAULT_MODEL
-        )
-
-        try:
-            return moderation_model_instance.run(text=text)
-        except Exception as ex:
-            logging.exception(ex)
-            raise LLMBadRequestError('Rate limit exceeded, please try again later.')
-
-    def _call(
-            self,
-            inputs: Dict[str, Any],
-            run_manager: Optional[CallbackManagerForChainRun] = None,
-    ) -> Dict[str, Any]:
-        text = inputs[self.input_key]
-
-        if self.sensitive_word_avoidance_rule.type == SensitiveWordAvoidanceRule.Type.KEYWORDS:
-            result = self._check_sensitive_word(text)
-        else:
-            result = self._check_moderation(text)
-
-        if not result:
-            raise SensitiveWordAvoidanceError(self.sensitive_word_avoidance_rule.canned_response)
-
-        return {self.output_key: text}
-
-
-class SensitiveWordAvoidanceError(Exception):
-    def __init__(self, message):
-        super().__init__(message)
-        self.message = message

api/core/completion.py

@@ -1,29 +1,39 @@
+import concurrent
+import json
 import logging
-from typing import Optional, List, Union
+from concurrent.futures import ThreadPoolExecutor
+from typing import Optional, List, Union, Tuple
 
+from flask import current_app, Flask
 from requests.exceptions import ChunkedEncodingError
 
 from core.agent.agent_executor import AgentExecuteResult, PlanningStrategy
 from core.callback_handler.main_chain_gather_callback_handler import MainChainGatherCallbackHandler
 from core.callback_handler.llm_callback_handler import LLMCallbackHandler
-from core.chain.sensitive_word_avoidance_chain import SensitiveWordAvoidanceError
-from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException
+from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException, \
+    ConversationTaskInterruptException
+from core.external_data_tool.factory import ExternalDataToolFactory
+from core.file.file_obj import FileObj
 from core.model_providers.error import LLMBadRequestError
 from core.memory.read_only_conversation_token_db_buffer_shared_memory import \
     ReadOnlyConversationTokenDBBufferSharedMemory
 from core.model_providers.model_factory import ModelFactory
-from core.model_providers.models.entity.message import PromptMessage
+from core.model_providers.models.entity.message import PromptMessage, PromptMessageFile
 from core.model_providers.models.llm.base import BaseLLM
 from core.orchestrator_rule_parser import OrchestratorRuleParser
 from core.prompt.prompt_template import PromptTemplateParser
+from core.prompt.prompt_transform import PromptTransform
 from models.model import App, AppModelConfig, Account, Conversation, EndUser
+from core.moderation.base import ModerationException, ModerationAction
+from core.moderation.factory import ModerationFactory
 
 
 class Completion:
     @classmethod
     def generate(cls, task_id: str, app: App, app_model_config: AppModelConfig, query: str, inputs: dict,
-                 user: Union[Account, EndUser], conversation: Optional[Conversation], streaming: bool,
-                 is_override: bool = False, retriever_from: str = 'dev'):
+                 files: List[FileObj], user: Union[Account, EndUser], conversation: Optional[Conversation],
+                 streaming: bool, is_override: bool = False, retriever_from: str = 'dev',
+                 auto_generate_name: bool = True):
         """
         errors: ProviderTokenNotInitError
         """
@@ -56,16 +66,21 @@ class Completion:
             is_override=is_override,
             inputs=inputs,
             query=query,
+            files=files,
             streaming=streaming,
-            model_instance=final_model_instance
+            model_instance=final_model_instance,
+            auto_generate_name=auto_generate_name
         )
 
+        prompt_message_files = [file.prompt_message_file for file in files]
+
         rest_tokens_for_context_and_memory = cls.get_validate_rest_tokens(
             mode=app.mode,
             model_instance=final_model_instance,
             app_model_config=app_model_config,
             query=query,
-            inputs=inputs
+            inputs=inputs,
+            files=prompt_message_files
         )
 
         # init orchestrator rule parser
@@ -75,26 +90,36 @@ class Completion:
         )
 
         try:
-            # parse sensitive_word_avoidance_chain
             chain_callback = MainChainGatherCallbackHandler(conversation_message_task)
-            sensitive_word_avoidance_chain = orchestrator_rule_parser.to_sensitive_word_avoidance_chain(
-                final_model_instance, [chain_callback])
-            if sensitive_word_avoidance_chain:
-                try:
-                    query = sensitive_word_avoidance_chain.run(query)
-                except SensitiveWordAvoidanceError as ex:
-                    cls.run_final_llm(
-                        model_instance=final_model_instance,
-                        mode=app.mode,
-                        app_model_config=app_model_config,
-                        query=query,
-                        inputs=inputs,
-                        agent_execute_result=None,
-                        conversation_message_task=conversation_message_task,
-                        memory=memory,
-                        fake_response=ex.message
-                    )
-                    return
+
+            try:
+                # process sensitive_word_avoidance
+                inputs, query = cls.moderation_for_inputs(app.id, app.tenant_id, app_model_config, inputs, query)
+            except ModerationException as e:
+                cls.run_final_llm(
+                    model_instance=final_model_instance,
+                    mode=app.mode,
+                    app_model_config=app_model_config,
+                    query=query,
+                    inputs=inputs,
+                    files=prompt_message_files,
+                    agent_execute_result=None,
+                    conversation_message_task=conversation_message_task,
+                    memory=memory,
+                    fake_response=str(e)
+                )
+                return
+
+            # fill in variable inputs from external data tools if exists
+            external_data_tools = app_model_config.external_data_tools_list
+            if external_data_tools:
+                inputs = cls.fill_in_inputs_from_external_data_tools(
+                    tenant_id=app.tenant_id,
+                    app_id=app.id,
+                    external_data_tools=external_data_tools,
+                    inputs=inputs,
+                    query=query
+                )
 
             # get agent executor
             agent_executor = orchestrator_rule_parser.to_agent_executor(
@@ -129,51 +154,147 @@ class Completion:
                 app_model_config=app_model_config,
                 query=query,
                 inputs=inputs,
+                files=prompt_message_files,
                 agent_execute_result=agent_execute_result,
                 conversation_message_task=conversation_message_task,
                 memory=memory,
                 fake_response=fake_response
             )
-        except ConversationTaskStoppedException:
+        except (ConversationTaskInterruptException, ConversationTaskStoppedException):
             return
         except ChunkedEncodingError as e:
             # Interrupt by LLM (like OpenAI), handle it.
             logging.warning(f'ChunkedEncodingError: {e}')
             conversation_message_task.end()
             return
-        
+
+    @classmethod
+    def moderation_for_inputs(cls, app_id: str, tenant_id: str, app_model_config: AppModelConfig, inputs: dict, query: str):
+        if not app_model_config.sensitive_word_avoidance_dict['enabled']:
+            return inputs, query
+
+        type = app_model_config.sensitive_word_avoidance_dict['type']
+
+        moderation = ModerationFactory(type, app_id, tenant_id, app_model_config.sensitive_word_avoidance_dict['config'])
+        moderation_result = moderation.moderation_for_inputs(inputs, query)
+
+        if not moderation_result.flagged:
+            return inputs, query
+
+        if moderation_result.action == ModerationAction.DIRECT_OUTPUT:
+            raise ModerationException(moderation_result.preset_response)
+        elif moderation_result.action == ModerationAction.OVERRIDED:
+            inputs = moderation_result.inputs
+            query = moderation_result.query
+
+        return inputs, query
+
+    @classmethod
+    def fill_in_inputs_from_external_data_tools(cls, tenant_id: str, app_id: str, external_data_tools: list[dict],
+                                                inputs: dict, query: str) -> dict:
+        """
+        Fill in variable inputs from external data tools if exists.
+
+        :param tenant_id: workspace id
+        :param app_id: app id
+        :param external_data_tools: external data tools configs
+        :param inputs: the inputs
+        :param query: the query
+        :return: the filled inputs
+        """
+        # Group tools by type and config
+        grouped_tools = {}
+        for tool in external_data_tools:
+            if not tool.get("enabled"):
+                continue
+
+            tool_key = (tool.get("type"), json.dumps(tool.get("config"), sort_keys=True))
+            grouped_tools.setdefault(tool_key, []).append(tool)
+
+        results = {}
+        with ThreadPoolExecutor() as executor:
+            futures = {}
+            for tool in external_data_tools:
+                if not tool.get("enabled"):
+                    continue
+
+                future = executor.submit(
+                    cls.query_external_data_tool, current_app._get_current_object(), tenant_id, app_id, tool,
+                    inputs, query
+                )
+
+                futures[future] = tool
+
+            for future in concurrent.futures.as_completed(futures):
+                tool_variable, result = future.result()
+                results[tool_variable] = result
+
+        inputs.update(results)
+        return inputs
+
+    @classmethod
+    def query_external_data_tool(cls, flask_app: Flask, tenant_id: str, app_id: str, external_data_tool: dict,
+                                 inputs: dict, query: str) -> Tuple[Optional[str], Optional[str]]:
+        with flask_app.app_context():
+            tool_variable = external_data_tool.get("variable")
+            tool_type = external_data_tool.get("type")
+            tool_config = external_data_tool.get("config")
+
+            external_data_tool_factory = ExternalDataToolFactory(
+                name=tool_type,
+                tenant_id=tenant_id,
+                app_id=app_id,
+                variable=tool_variable,
+                config=tool_config
+            )
+
+            # query external data tool
+            result = external_data_tool_factory.query(
+                inputs=inputs,
+                query=query
+            )
+
+            return tool_variable, result
+
     @classmethod
     def get_query_for_agent(cls, app: App, app_model_config: AppModelConfig, query: str, inputs: dict) -> str:
         if app.mode != 'completion':
             return query
-        
+
         return inputs.get(app_model_config.dataset_query_variable, "")
 
     @classmethod
     def run_final_llm(cls, model_instance: BaseLLM, mode: str, app_model_config: AppModelConfig, query: str,
                       inputs: dict,
+                      files: List[PromptMessageFile],
                       agent_execute_result: Optional[AgentExecuteResult],
                       conversation_message_task: ConversationMessageTask,
                       memory: Optional[ReadOnlyConversationTokenDBBufferSharedMemory],
                       fake_response: Optional[str]):
+        prompt_transform = PromptTransform()
+
         # get llm prompt
         if app_model_config.prompt_type == 'simple':
-            prompt_messages, stop_words = model_instance.get_prompt(
-                mode=mode,
+            prompt_messages, stop_words = prompt_transform.get_prompt(
+                app_mode=mode,
                 pre_prompt=app_model_config.pre_prompt,
                 inputs=inputs,
                 query=query,
+                files=files,
                 context=agent_execute_result.output if agent_execute_result else None,
-                memory=memory
+                memory=memory,
+                model_instance=model_instance
             )
         else:
-            prompt_messages = model_instance.get_advanced_prompt(
+            prompt_messages = prompt_transform.get_advanced_prompt(
                 app_mode=mode,
                 app_model_config=app_model_config,
                 inputs=inputs,
                 query=query,
+                files=files,
                 context=agent_execute_result.output if agent_execute_result else None,
-                memory=memory
+                memory=memory,
+                model_instance=model_instance
             )
 
             model_config = app_model_config.model_dict
@@ -228,7 +349,7 @@ class Completion:
 
     @classmethod
     def get_validate_rest_tokens(cls, mode: str, model_instance: BaseLLM, app_model_config: AppModelConfig,
-                                 query: str, inputs: dict) -> int:
+                                 query: str, inputs: dict, files: List[PromptMessageFile]) -> int:
         model_limited_tokens = model_instance.model_rules.max_tokens.max
         max_tokens = model_instance.get_model_kwargs().max_tokens
 
@@ -238,15 +359,31 @@ class Completion:
         if max_tokens is None:
             max_tokens = 0
 
+        prompt_transform = PromptTransform()
+
         # get prompt without memory and context
-        prompt_messages, _ = model_instance.get_prompt(
-            mode=mode,
-            pre_prompt=app_model_config.pre_prompt,
-            inputs=inputs,
-            query=query,
-            context=None,
-            memory=None
-        )
+        if app_model_config.prompt_type == 'simple':
+            prompt_messages, _ = prompt_transform.get_prompt(
+                app_mode=mode,
+                pre_prompt=app_model_config.pre_prompt,
+                inputs=inputs,
+                query=query,
+                files=files,
+                context=None,
+                memory=None,
+                model_instance=model_instance
+            )
+        else:
+            prompt_messages = prompt_transform.get_advanced_prompt(
+                app_mode=mode,
+                app_model_config=app_model_config,
+                inputs=inputs,
+                query=query,
+                files=files,
+                context=None,
+                memory=None,
+                model_instance=model_instance
+            )
 
         prompt_tokens = model_instance.get_num_tokens(prompt_messages)
         rest_tokens = model_limited_tokens - max_tokens - prompt_tokens

api/core/conversation_message_task.py

@@ -6,8 +6,9 @@ from core.callback_handler.entity.agent_loop import AgentLoop
 from core.callback_handler.entity.dataset_query import DatasetQueryObj
 from core.callback_handler.entity.llm_message import LLMMessage
 from core.callback_handler.entity.chain_result import ChainResult
+from core.file.file_obj import FileObj
 from core.model_providers.model_factory import ModelFactory
-from core.model_providers.models.entity.message import to_prompt_messages, MessageType
+from core.model_providers.models.entity.message import to_prompt_messages, MessageType, PromptMessageFile
 from core.model_providers.models.llm.base import BaseLLM
 from core.prompt.prompt_builder import PromptBuilder
 from core.prompt.prompt_template import PromptTemplateParser
@@ -16,13 +17,14 @@ from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from models.dataset import DatasetQuery
 from models.model import AppModelConfig, Conversation, Account, Message, EndUser, App, MessageAgentThought, \
-    MessageChain, DatasetRetrieverResource
+    MessageChain, DatasetRetrieverResource, MessageFile
 
 
 class ConversationMessageTask:
     def __init__(self, task_id: str, app: App, app_model_config: AppModelConfig, user: Account,
-                 inputs: dict, query: str, streaming: bool, model_instance: BaseLLM,
-                 conversation: Optional[Conversation] = None, is_override: bool = False):
+                 inputs: dict, query: str, files: List[FileObj], streaming: bool,
+                 model_instance: BaseLLM, conversation: Optional[Conversation] = None, is_override: bool = False,
+                 auto_generate_name: bool = True):
         self.start_at = time.perf_counter()
 
         self.task_id = task_id
@@ -35,6 +37,7 @@ class ConversationMessageTask:
         self.user = user
         self.inputs = inputs
         self.query = query
+        self.files = files
         self.streaming = streaming
 
         self.conversation = conversation
@@ -45,6 +48,7 @@ class ConversationMessageTask:
         self.message = None
 
         self.retriever_resource = None
+        self.auto_generate_name = auto_generate_name
 
         self.model_dict = self.app_model_config.model_dict
         self.provider_name = self.model_dict.get('provider')
@@ -100,7 +104,7 @@ class ConversationMessageTask:
                 model_id=self.model_name,
                 override_model_configs=json.dumps(override_model_configs) if override_model_configs else None,
                 mode=self.mode,
-                name='',
+                name='New conversation',
                 inputs=self.inputs,
                 introduction=introduction,
                 system_instruction=system_instruction,
@@ -142,6 +146,19 @@ class ConversationMessageTask:
         db.session.add(self.message)
         db.session.commit()
 
+        for file in self.files:
+            message_file = MessageFile(
+                message_id=self.message.id,
+                type=file.type.value,
+                transfer_method=file.transfer_method.value,
+                url=file.url,
+                upload_file_id=file.upload_file_id,
+                created_by_role=('account' if isinstance(self.user, Account) else 'end_user'),
+                created_by=self.user.id
+            )
+            db.session.add(message_file)
+            db.session.commit()
+
     def append_message_text(self, text: str):
         if text is not None:
             self._pub_handler.pub_text(text)
@@ -176,7 +193,8 @@ class ConversationMessageTask:
         message_was_created.send(
             self.message,
             conversation=self.conversation,
-            is_first_message=self.is_new_conversation
+            is_first_message=self.is_new_conversation,
+            auto_generate_name=self.auto_generate_name
         )
 
         if not by_stopped:
@@ -290,6 +308,10 @@ class ConversationMessageTask:
                 db.session.commit()
             self.retriever_resource = resource
 
+    def on_message_replace(self, text: str):
+        if text is not None:
+            self._pub_handler.pub_message_replace(text)
+
     def message_end(self):
         self._pub_handler.pub_message_end(self.retriever_resource)
 
@@ -342,6 +364,24 @@ class PubHandler:
             self.pub_end()
             raise ConversationTaskStoppedException()
 
+    def pub_message_replace(self, text: str):
+        content = {
+            'event': 'message_replace',
+            'data': {
+                'task_id': self._task_id,
+                'message_id': str(self._message.id),
+                'text': text,
+                'mode': self._conversation.mode,
+                'conversation_id': str(self._conversation.id)
+            }
+        }
+
+        redis_client.publish(self._channel, json.dumps(content))
+
+        if self._is_stopped():
+            self.pub_end()
+            raise ConversationTaskStoppedException()
+
     def pub_chain(self, message_chain: MessageChain):
         if self._chain_pub:
             content = {
@@ -443,3 +483,7 @@ class PubHandler:
 
 class ConversationTaskStoppedException(Exception):
     pass
+
+
+class ConversationTaskInterruptException(Exception):
+    pass

api/core/extension/api_based_extension_requestor.py

@@ -0,0 +1,62 @@
+import os
+
+import requests
+
+from models.api_based_extension import APIBasedExtensionPoint
+
+
+class APIBasedExtensionRequestor:
+    timeout: (int, int) = (5, 60)
+    """timeout for request connect and read"""
+
+    def __init__(self, api_endpoint: str, api_key: str) -> None:
+        self.api_endpoint = api_endpoint
+        self.api_key = api_key
+
+    def request(self, point: APIBasedExtensionPoint, params: dict) -> dict:
+        """
+        Request the api.
+
+        :param point: the api point
+        :param params: the request params
+        :return: the response json
+        """
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": "Bearer {}".format(self.api_key)
+        }
+
+        url = self.api_endpoint
+
+        try:
+            # proxy support for security
+            proxies = None
+            if os.environ.get("API_BASED_EXTENSION_HTTP_PROXY") and os.environ.get("API_BASED_EXTENSION_HTTPS_PROXY"):
+                proxies = {
+                    'http': os.environ.get("API_BASED_EXTENSION_HTTP_PROXY"),
+                    'https': os.environ.get("API_BASED_EXTENSION_HTTPS_PROXY"),
+                }
+
+            response = requests.request(
+                method='POST',
+                url=url,
+                json={
+                    'point': point.value,
+                    'params': params
+                },
+                headers=headers,
+                timeout=self.timeout,
+                proxies=proxies
+            )
+        except requests.exceptions.Timeout:
+            raise ValueError("request timeout")
+        except requests.exceptions.ConnectionError:
+            raise ValueError("request connection error")
+
+        if response.status_code != 200:
+            raise ValueError("request error, status_code: {}, content: {}".format(
+                response.status_code,
+                response.text[:100]
+            ))
+
+        return response.json()

api/core/extension/extensible.py

@@ -0,0 +1,111 @@
+import enum
+import importlib.util
+import json
+import logging
+import os
+from collections import OrderedDict
+from typing import Any, Optional
+
+from pydantic import BaseModel
+
+
+class ExtensionModule(enum.Enum):
+    MODERATION = 'moderation'
+    EXTERNAL_DATA_TOOL = 'external_data_tool'
+
+
+class ModuleExtension(BaseModel):
+    extension_class: Any
+    name: str
+    label: Optional[dict] = None
+    form_schema: Optional[list] = None
+    builtin: bool = True
+    position: Optional[int] = None
+
+
+class Extensible:
+    module: ExtensionModule
+
+    name: str
+    tenant_id: str
+    config: Optional[dict] = None
+
+    def __init__(self, tenant_id: str, config: Optional[dict] = None) -> None:
+        self.tenant_id = tenant_id
+        self.config = config
+
+    @classmethod
+    def scan_extensions(cls):
+        extensions = {}
+
+        # get the path of the current class
+        current_path = os.path.abspath(cls.__module__.replace(".", os.path.sep) + '.py')
+        current_dir_path = os.path.dirname(current_path)
+
+        # traverse subdirectories
+        for subdir_name in os.listdir(current_dir_path):
+            if subdir_name.startswith('__'):
+                continue
+
+            subdir_path = os.path.join(current_dir_path, subdir_name)
+            extension_name = subdir_name
+            if os.path.isdir(subdir_path):
+                file_names = os.listdir(subdir_path)
+
+                # is builtin extension, builtin extension
+                # in the front-end page and business logic, there are special treatments.
+                builtin = False
+                position = None
+                if '__builtin__' in file_names:
+                    builtin = True
+
+                    builtin_file_path = os.path.join(subdir_path, '__builtin__')
+                    if os.path.exists(builtin_file_path):
+                        with open(builtin_file_path, 'r') as f:
+                            position = int(f.read().strip())
+
+                if (extension_name + '.py') not in file_names:
+                    logging.warning(f"Missing {extension_name}.py file in {subdir_path}, Skip.")
+                    continue
+
+                # Dynamic loading {subdir_name}.py file and find the subclass of Extensible
+                py_path = os.path.join(subdir_path, extension_name + '.py')
+                spec = importlib.util.spec_from_file_location(extension_name, py_path)
+                mod = importlib.util.module_from_spec(spec)
+                spec.loader.exec_module(mod)
+
+                extension_class = None
+                for name, obj in vars(mod).items():
+                    if isinstance(obj, type) and issubclass(obj, cls) and obj != cls:
+                        extension_class = obj
+                        break
+
+                if not extension_class:
+                    logging.warning(f"Missing subclass of {cls.__name__} in {py_path}, Skip.")
+                    continue
+
+                json_data = {}
+                if not builtin:
+                    if 'schema.json' not in file_names:
+                        logging.warning(f"Missing schema.json file in {subdir_path}, Skip.")
+                        continue
+
+                    json_path = os.path.join(subdir_path, 'schema.json')
+                    json_data = {}
+                    if os.path.exists(json_path):
+                        with open(json_path, 'r') as f:
+                            json_data = json.load(f)
+
+                extensions[extension_name] = ModuleExtension(
+                    extension_class=extension_class,
+                    name=extension_name,
+                    label=json_data.get('label'),
+                    form_schema=json_data.get('form_schema'),
+                    builtin=builtin,
+                    position=position
+                )
+
+        sorted_items = sorted(extensions.items(), key=lambda x: (x[1].position is None, x[1].position))
+        sorted_extensions = OrderedDict(sorted_items)
+
+        return sorted_extensions

api/core/extension/extension.py

@@ -0,0 +1,47 @@
+from core.extension.extensible import ModuleExtension, ExtensionModule
+from core.external_data_tool.base import ExternalDataTool
+from core.moderation.base import Moderation
+
+
+class Extension:
+    __module_extensions: dict[str, dict[str, ModuleExtension]] = {}
+
+    module_classes = {
+        ExtensionModule.MODERATION: Moderation,
+        ExtensionModule.EXTERNAL_DATA_TOOL: ExternalDataTool
+    }
+
+    def init(self):
+        for module, module_class in self.module_classes.items():
+            self.__module_extensions[module.value] = module_class.scan_extensions()
+
+    def module_extensions(self, module: str) -> list[ModuleExtension]:
+        module_extensions = self.__module_extensions.get(module)
+
+        if not module_extensions:
+            raise ValueError(f"Extension Module {module} not found")
+
+        return list(module_extensions.values())
+
+    def module_extension(self, module: ExtensionModule, extension_name: str) -> ModuleExtension:
+        module_extensions = self.__module_extensions.get(module.value)
+
+        if not module_extensions:
+            raise ValueError(f"Extension Module {module} not found")
+
+        module_extension = module_extensions.get(extension_name)
+
+        if not module_extension:
+            raise ValueError(f"Extension {extension_name} not found")
+
+        return module_extension
+
+    def extension_class(self, module: ExtensionModule, extension_name: str) -> type:
+        module_extension = self.module_extension(module, extension_name)
+        return module_extension.extension_class
+
+    def validate_form_schema(self, module: ExtensionModule, extension_name: str, config: dict) -> None:
+        module_extension = self.module_extension(module, extension_name)
+        form_schema = module_extension.form_schema
+
+        # TODO validate form_schema

api/core/external_data_tool/api/__builtin__

@@ -0,0 +1 @@
+1

api/core/external_data_tool/api/api.py

@@ -0,0 +1,92 @@
+from typing import Optional
+
+from core.extension.api_based_extension_requestor import APIBasedExtensionRequestor
+from core.external_data_tool.base import ExternalDataTool
+from core.helper import encrypter
+from extensions.ext_database import db
+from models.api_based_extension import APIBasedExtension, APIBasedExtensionPoint
+
+
+class ApiExternalDataTool(ExternalDataTool):
+    """
+    The api external data tool.
+    """
+
+    name: str = "api"
+    """the unique name of external data tool"""
+
+    @classmethod
+    def validate_config(cls, tenant_id: str, config: dict) -> None:
+        """
+        Validate the incoming form config data.
+
+        :param tenant_id: the id of workspace
+        :param config: the form config data
+        :return:
+        """
+        # own validation logic
+        api_based_extension_id = config.get("api_based_extension_id")
+        if not api_based_extension_id:
+            raise ValueError("api_based_extension_id is required")
+
+        # get api_based_extension
+        api_based_extension = db.session.query(APIBasedExtension).filter(
+            APIBasedExtension.tenant_id == tenant_id,
+            APIBasedExtension.id == api_based_extension_id
+        ).first()
+
+        if not api_based_extension:
+            raise ValueError("api_based_extension_id is invalid")
+
+    def query(self, inputs: dict, query: Optional[str] = None) -> str:
+        """
+        Query the external data tool.
+
+        :param inputs: user inputs
+        :param query: the query of chat app
+        :return: the tool query result
+        """
+        # get params from config
+        api_based_extension_id = self.config.get("api_based_extension_id")
+
+        # get api_based_extension
+        api_based_extension = db.session.query(APIBasedExtension).filter(
+            APIBasedExtension.tenant_id == self.tenant_id,
+            APIBasedExtension.id == api_based_extension_id
+        ).first()
+
+        if not api_based_extension:
+            raise ValueError("[External data tool] API query failed, variable: {}, "
+                             "error: api_based_extension_id is invalid"
+                             .format(self.config.get('variable')))
+
+        # decrypt api_key
+        api_key = encrypter.decrypt_token(
+            tenant_id=self.tenant_id,
+            token=api_based_extension.api_key
+        )
+
+        try:
+            # request api
+            requestor = APIBasedExtensionRequestor(
+                api_endpoint=api_based_extension.api_endpoint,
+                api_key=api_key
+            )
+        except Exception as e:
+            raise ValueError("[External data tool] API query failed, variable: {}, error: {}".format(
+                self.config.get('variable'),
+                e
+            ))
+
+        response_json = requestor.request(point=APIBasedExtensionPoint.APP_EXTERNAL_DATA_TOOL_QUERY, params={
+            'app_id': self.app_id,
+            'tool_variable': self.variable,
+            'inputs': inputs,
+            'query': query
+        })
+
+        if 'result' not in response_json:
+            raise ValueError("[External data tool] API query failed, variable: {}, error: result not found in response"
+                             .format(self.config.get('variable')))
+
+        return response_json['result']

api/core/external_data_tool/base.py

@@ -0,0 +1,45 @@
+from abc import abstractmethod, ABC
+from typing import Optional
+
+from core.extension.extensible import Extensible, ExtensionModule
+
+
+class ExternalDataTool(Extensible, ABC):
+    """
+    The base class of external data tool.
+    """
+
+    module: ExtensionModule = ExtensionModule.EXTERNAL_DATA_TOOL
+
+    app_id: str
+    """the id of app"""
+    variable: str
+    """the tool variable name of app tool"""
+
+    def __init__(self, tenant_id: str, app_id: str, variable: str, config: Optional[dict] = None) -> None:
+        super().__init__(tenant_id, config)
+        self.app_id = app_id
+        self.variable = variable
+
+    @classmethod
+    @abstractmethod
+    def validate_config(cls, tenant_id: str, config: dict) -> None:
+        """
+        Validate the incoming form config data.
+
+        :param tenant_id: the id of workspace
+        :param config: the form config data
+        :return:
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def query(self, inputs: dict, query: Optional[str] = None) -> str:
+        """
+        Query the external data tool.
+
+        :param inputs: user inputs
+        :param query: the query of chat app
+        :return: the tool query result
+        """
+        raise NotImplementedError

api/core/external_data_tool/factory.py

@@ -0,0 +1,40 @@
+from typing import Optional
+
+from core.extension.extensible import ExtensionModule
+from extensions.ext_code_based_extension import code_based_extension
+
+
+class ExternalDataToolFactory:
+
+    def __init__(self, name: str, tenant_id: str, app_id: str, variable: str, config: dict) -> None:
+        extension_class = code_based_extension.extension_class(ExtensionModule.EXTERNAL_DATA_TOOL, name)
+        self.__extension_instance = extension_class(
+            tenant_id=tenant_id,
+            app_id=app_id,
+            variable=variable,
+            config=config
+        )
+
+    @classmethod
+    def validate_config(cls, name: str, tenant_id: str, config: dict) -> None:
+        """
+        Validate the incoming form config data.
+
+        :param name: the name of external data tool
+        :param tenant_id: the id of workspace
+        :param config: the form config data
+        :return:
+        """
+        code_based_extension.validate_form_schema(ExtensionModule.EXTERNAL_DATA_TOOL, name, config)
+        extension_class = code_based_extension.extension_class(ExtensionModule.EXTERNAL_DATA_TOOL, name)
+        extension_class.validate_config(tenant_id, config)
+
+    def query(self, inputs: dict, query: Optional[str] = None) -> str:
+        """
+        Query the external data tool.
+
+        :param inputs: user inputs
+        :param query: the query of chat app
+        :return: the tool query result
+        """
+        return self.__extension_instance.query(inputs, query)

api/core/file/file_obj.py

@@ -0,0 +1,79 @@
+import enum
+from typing import Optional
+
+from pydantic import BaseModel
+
+from core.file.upload_file_parser import UploadFileParser
+from core.model_providers.models.entity.message import PromptMessageFile, ImagePromptMessageFile
+from extensions.ext_database import db
+from models.model import UploadFile
+
+
+class FileType(enum.Enum):
+    IMAGE = 'image'
+
+    @staticmethod
+    def value_of(value):
+        for member in FileType:
+            if member.value == value:
+                return member
+        raise ValueError(f"No matching enum found for value '{value}'")
+
+
+class FileTransferMethod(enum.Enum):
+    REMOTE_URL = 'remote_url'
+    LOCAL_FILE = 'local_file'
+
+    @staticmethod
+    def value_of(value):
+        for member in FileTransferMethod:
+            if member.value == value:
+                return member
+        raise ValueError(f"No matching enum found for value '{value}'")
+
+
+class FileObj(BaseModel):
+    id: Optional[str]
+    tenant_id: str
+    type: FileType
+    transfer_method: FileTransferMethod
+    url: Optional[str]
+    upload_file_id: Optional[str]
+    file_config: dict
+
+    @property
+    def data(self) -> Optional[str]:
+        return self._get_data()
+
+    @property
+    def preview_url(self) -> Optional[str]:
+        return self._get_data(force_url=True)
+
+    @property
+    def prompt_message_file(self) -> PromptMessageFile:
+        if self.type == FileType.IMAGE:
+            image_config = self.file_config.get('image')
+
+            return ImagePromptMessageFile(
+                data=self.data,
+                detail=ImagePromptMessageFile.DETAIL.HIGH
+                if image_config.get("detail") == "high" else ImagePromptMessageFile.DETAIL.LOW
+            )
+
+    def _get_data(self, force_url: bool = False) -> Optional[str]:
+        if self.type == FileType.IMAGE:
+            if self.transfer_method == FileTransferMethod.REMOTE_URL:
+                return self.url
+            elif self.transfer_method == FileTransferMethod.LOCAL_FILE:
+                upload_file = (db.session.query(UploadFile)
+                               .filter(
+                    UploadFile.id == self.upload_file_id,
+                    UploadFile.tenant_id == self.tenant_id
+                ).first())
+
+                return UploadFileParser.get_image_data(
+                    upload_file=upload_file,
+                    force_url=force_url
+                )
+
+        return None

api/core/file/message_file_parser.py

@@ -0,0 +1,180 @@
+from typing import List, Union, Optional, Dict
+
+import requests
+
+from core.file.file_obj import FileObj, FileType, FileTransferMethod
+from core.file.upload_file_parser import SUPPORT_EXTENSIONS
+from extensions.ext_database import db
+from models.account import Account
+from models.model import MessageFile, EndUser, AppModelConfig, UploadFile
+
+
+class MessageFileParser:
+
+    def __init__(self, tenant_id: str, app_id: str) -> None:
+        self.tenant_id = tenant_id
+        self.app_id = app_id
+
+    def validate_and_transform_files_arg(self, files: List[dict], app_model_config: AppModelConfig,
+                                         user: Union[Account, EndUser]) -> List[FileObj]:
+        """
+        validate and transform files arg
+
+        :param files:
+        :param app_model_config:
+        :param user:
+        :return:
+        """
+        file_upload_config = app_model_config.file_upload_dict
+
+        for file in files:
+            if not isinstance(file, dict):
+                raise ValueError('Invalid file format, must be dict')
+            if not file.get('type'):
+                raise ValueError('Missing file type')
+            FileType.value_of(file.get('type'))
+            if not file.get('transfer_method'):
+                raise ValueError('Missing file transfer method')
+            FileTransferMethod.value_of(file.get('transfer_method'))
+            if file.get('transfer_method') == FileTransferMethod.REMOTE_URL.value:
+                if not file.get('url'):
+                    raise ValueError('Missing file url')
+                if not file.get('url').startswith('http'):
+                    raise ValueError('Invalid file url')
+            if file.get('transfer_method') == FileTransferMethod.LOCAL_FILE.value and not file.get('upload_file_id'):
+                raise ValueError('Missing file upload_file_id')
+
+        # transform files to file objs
+        type_file_objs = self._to_file_objs(files, file_upload_config)
+
+        # validate files
+        new_files = []
+        for file_type, file_objs in type_file_objs.items():
+            if file_type == FileType.IMAGE:
+                # parse and validate files
+                image_config = file_upload_config.get('image')
+
+                # check if image file feature is enabled
+                if not image_config['enabled']:
+                    continue
+
+                # Validate number of files
+                if len(files) > image_config['number_limits']:
+                    raise ValueError(f"Number of image files exceeds the maximum limit {image_config['number_limits']}")
+
+                for file_obj in file_objs:
+                    # Validate transfer method
+                    if file_obj.transfer_method.value not in image_config['transfer_methods']:
+                        raise ValueError(f'Invalid transfer method: {file_obj.transfer_method.value}')
+
+                    # Validate file type
+                    if file_obj.type != FileType.IMAGE:
+                        raise ValueError(f'Invalid file type: {file_obj.type}')
+
+                    if file_obj.transfer_method == FileTransferMethod.REMOTE_URL:
+                        # check remote url valid and is image
+                        result, error = self._check_image_remote_url(file_obj.url)
+                        if result is False:
+                            raise ValueError(error)
+                    elif file_obj.transfer_method == FileTransferMethod.LOCAL_FILE:
+                        # get upload file from upload_file_id
+                        upload_file = (db.session.query(UploadFile)
+                                       .filter(
+                            UploadFile.id == file_obj.upload_file_id,
+                            UploadFile.tenant_id == self.tenant_id,
+                            UploadFile.created_by == user.id,
+                            UploadFile.created_by_role == ('account' if isinstance(user, Account) else 'end_user'),
+                            UploadFile.extension.in_(SUPPORT_EXTENSIONS)
+                        ).first())
+
+                        # check upload file is belong to tenant and user
+                        if not upload_file:
+                            raise ValueError('Invalid upload file')
+
+                    new_files.append(file_obj)
+
+        # return all file objs
+        return new_files
+
+    def transform_message_files(self, files: List[MessageFile], app_model_config: Optional[AppModelConfig]) -> List[FileObj]:
+        """
+        transform message files
+
+        :param files:
+        :param app_model_config:
+        :return:
+        """
+        # transform files to file objs
+        type_file_objs = self._to_file_objs(files, app_model_config.file_upload_dict)
+
+        # return all file objs
+        return [file_obj for file_objs in type_file_objs.values() for file_obj in file_objs]
+
+    def _to_file_objs(self, files: List[Union[Dict, MessageFile]],
+                      file_upload_config: dict) -> Dict[FileType, List[FileObj]]:
+        """
+        transform files to file objs
+
+        :param files:
+        :param file_upload_config:
+        :return:
+        """
+        type_file_objs: Dict[FileType, List[FileObj]] = {
+            # Currently only support image
+            FileType.IMAGE: []
+        }
+
+        if not files:
+            return type_file_objs
+
+        # group by file type and convert file args or message files to FileObj
+        for file in files:
+            file_obj = self._to_file_obj(file, file_upload_config)
+            if file_obj.type not in type_file_objs:
+                continue
+
+            type_file_objs[file_obj.type].append(file_obj)
+
+        return type_file_objs
+
+    def _to_file_obj(self, file: Union[dict, MessageFile], file_upload_config: dict) -> FileObj:
+        """
+        transform file to file obj
+
+        :param file:
+        :return:
+        """
+        if isinstance(file, dict):
+            transfer_method = FileTransferMethod.value_of(file.get('transfer_method'))
+            return FileObj(
+                tenant_id=self.tenant_id,
+                type=FileType.value_of(file.get('type')),
+                transfer_method=transfer_method,
+                url=file.get('url') if transfer_method == FileTransferMethod.REMOTE_URL else None,
+                upload_file_id=file.get('upload_file_id') if transfer_method == FileTransferMethod.LOCAL_FILE else None,
+                file_config=file_upload_config
+            )
+        else:
+            return FileObj(
+                id=file.id,
+                tenant_id=self.tenant_id,
+                type=FileType.value_of(file.type),
+                transfer_method=FileTransferMethod.value_of(file.transfer_method),
+                url=file.url,
+                upload_file_id=file.upload_file_id or None,
+                file_config=file_upload_config
+            )
+
+    def _check_image_remote_url(self, url):
+        try:
+            headers = {
+                "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36"
+            }
+
+            response = requests.head(url, headers=headers, allow_redirects=True)
+            if response.status_code == 200:
+                return True, ""
+            else:
+                return False, "URL does not exist."
+        except requests.RequestException as e:
+            return False, f"Error checking URL: {e}"

api/core/file/upload_file_parser.py

@@ -0,0 +1,79 @@
+import base64
+import hashlib
+import hmac
+import logging
+import os
+import time
+from typing import Optional
+
+from flask import current_app
+
+from extensions.ext_storage import storage
+
+SUPPORT_EXTENSIONS = ['jpg', 'jpeg', 'png', 'webp', 'gif']
+
+
+class UploadFileParser:
+    @classmethod
+    def get_image_data(cls, upload_file, force_url: bool = False) -> Optional[str]:
+        if not upload_file:
+            return None
+
+        if upload_file.extension not in SUPPORT_EXTENSIONS:
+            return None
+
+        if current_app.config['MULTIMODAL_SEND_IMAGE_FORMAT'] == 'url' or force_url:
+            return cls.get_signed_temp_image_url(upload_file)
+        else:
+            # get image file base64
+            try:
+                data = storage.load(upload_file.key)
+            except FileNotFoundError:
+                logging.error(f'File not found: {upload_file.key}')
+                return None
+
+            encoded_string = base64.b64encode(data).decode('utf-8')
+            return f'data:{upload_file.mime_type};base64,{encoded_string}'
+
+    @classmethod
+    def get_signed_temp_image_url(cls, upload_file) -> str:
+        """
+        get signed url from upload file
+
+        :param upload_file: UploadFile object
+        :return:
+        """
+        base_url = current_app.config.get('FILES_URL')
+        image_preview_url = f'{base_url}/files/{upload_file.id}/image-preview'
+
+        timestamp = str(int(time.time()))
+        nonce = os.urandom(16).hex()
+        data_to_sign = f"image-preview|{upload_file.id}|{timestamp}|{nonce}"
+        secret_key = current_app.config['SECRET_KEY'].encode()
+        sign = hmac.new(secret_key, data_to_sign.encode(), hashlib.sha256).digest()
+        encoded_sign = base64.urlsafe_b64encode(sign).decode()
+
+        return f"{image_preview_url}?timestamp={timestamp}&nonce={nonce}&sign={encoded_sign}"
+
+    @classmethod
+    def verify_image_file_signature(cls, upload_file_id: str, timestamp: str, nonce: str, sign: str) -> bool:
+        """
+        verify signature
+
+        :param upload_file_id: file id
+        :param timestamp: timestamp
+        :param nonce: nonce
+        :param sign: signature
+        :return:
+        """
+        data_to_sign = f"image-preview|{upload_file_id}|{timestamp}|{nonce}"
+        secret_key = current_app.config['SECRET_KEY'].encode()
+        recalculated_sign = hmac.new(secret_key, data_to_sign.encode(), hashlib.sha256).digest()
+        recalculated_encoded_sign = base64.urlsafe_b64encode(recalculated_sign).decode()
+
+        # verify signature
+        if sign != recalculated_encoded_sign:
+            return False
+
+        current_time = int(time.time())
+        return current_time - int(timestamp) <= 300  # expired after 5 minutes

api/core/generator/llm_generator.py

@@ -16,7 +16,7 @@ from core.prompt.prompts import CONVERSATION_TITLE_PROMPT, GENERATOR_QA_PROMPT
 
 class LLMGenerator:
     @classmethod
-    def generate_conversation_name(cls, tenant_id: str, query, answer):
+    def generate_conversation_name(cls, tenant_id: str, query):
         prompt = CONVERSATION_TITLE_PROMPT
 
         if len(query) > 2000:
@@ -40,8 +40,12 @@ class LLMGenerator:
 
         result_dict = json.loads(answer)
         answer = result_dict['Your Output']
+        name = answer.strip()
 
-        return answer.strip()
+        if len(name) > 75:
+            name = name[:75] + '...'
+
+        return name
 
     @classmethod
     def generate_suggested_questions_after_answer(cls, tenant_id: str, histories: str):

api/core/indexing_runner.py

@@ -89,22 +89,6 @@ class IndexingRunner:
                 dataset_document.stopped_at = datetime.datetime.utcnow()
                 db.session.commit()
 
-    def format_split_text(self, text):
-        regex = r"Q\d+:\s*(.*?)\s*A\d+:\s*([\s\S]*?)(?=Q|$)"
-        matches = re.findall(regex, text, re.MULTILINE)
-
-        result = []
-        for match in matches:
-            q = match[0]
-            a = match[1]
-            if q and a:
-                result.append({
-                    "question": q,
-                    "answer": re.sub(r"\n\s*", "\n", a.strip())
-                })
-
-        return result
-
     def run_in_splitting_status(self, dataset_document: DatasetDocument):
         """Run the indexing process when the index_status is splitting."""
         try:
@@ -647,21 +631,16 @@ class IndexingRunner:
         return text
 
     def format_split_text(self, text):
-        regex = r"Q\d+:\s*(.*?)\s*A\d+:\s*([\s\S]*?)(?=Q|$)"  # 匹配Q和A的正则表达式
-        matches = re.findall(regex, text, re.MULTILINE)  # 获取所有匹配到的结果
+        regex = r"Q\d+:\s*(.*?)\s*A\d+:\s*([\s\S]*?)(?=Q|$)"
+        matches = re.findall(regex, text, re.MULTILINE)
 
-        result = []  # 存储最终的结果
-        for match in matches:
-            q = match[0]
-            a = match[1]
-            if q and a:
-                # 如果Q和A都存在，就将其添加到结果中
-                result.append({
-                    "question": q,
-                    "answer": re.sub(r"\n\s*", "\n", a.strip())
-                })
-
-        return result
+        return [
+            {
+                "question": q,
+                "answer": re.sub(r"\n\s*", "\n", a.strip())
+            }
+            for q, a in matches if q and a
+        ]
 
     def _build_index(self, dataset: Dataset, dataset_document: DatasetDocument, documents: List[Document]) -> None:
         """

api/core/memory/read_only_conversation_token_db_buffer_shared_memory.py

@@ -3,6 +3,7 @@ from typing import Any, List, Dict
 from langchain.memory.chat_memory import BaseChatMemory
 from langchain.schema import get_buffer_string, BaseMessage
 
+from core.file.message_file_parser import MessageFileParser
 from core.model_providers.models.entity.message import PromptMessage, MessageType, to_lc_messages
 from core.model_providers.models.llm.base import BaseLLM
 from extensions.ext_database import db
@@ -21,6 +22,8 @@ class ReadOnlyConversationTokenDBBufferSharedMemory(BaseChatMemory):
     @property
     def buffer(self) -> List[BaseMessage]:
         """String buffer of memory."""
+        app_model = self.conversation.app
+
         # fetch limited messages desc, and return reversed
         messages = db.session.query(Message).filter(
             Message.conversation_id == self.conversation.id,
@@ -28,10 +31,25 @@ class ReadOnlyConversationTokenDBBufferSharedMemory(BaseChatMemory):
         ).order_by(Message.created_at.desc()).limit(self.message_limit).all()
 
         messages = list(reversed(messages))
+        message_file_parser = MessageFileParser(tenant_id=app_model.tenant_id, app_id=self.conversation.app_id)
 
         chat_messages: List[PromptMessage] = []
         for message in messages:
-            chat_messages.append(PromptMessage(content=message.query, type=MessageType.USER))
+            files = message.message_files
+            if files:
+                file_objs = message_file_parser.transform_message_files(
+                    files, message.app_model_config
+                )
+
+                prompt_message_files = [file_obj.prompt_message_file for file_obj in file_objs]
+                chat_messages.append(PromptMessage(
+                    content=message.query,
+                    type=MessageType.USER,
+                    files=prompt_message_files
+                ))
+            else:
+                chat_messages.append(PromptMessage(content=message.query, type=MessageType.USER))
+
             chat_messages.append(PromptMessage(content=message.answer, type=MessageType.ASSISTANT))
 
         if not chat_messages:

api/core/model_providers/model_provider_factory.py

@@ -211,6 +211,9 @@ class ModelProviderFactory:
                                 Provider.quota_type == ProviderQuotaType.TRIAL.value
                             ).first()
 
+                        if provider.quota_limit == 0:
+                            return None
+
                         return provider
 
             no_system_provider = True

api/core/model_providers/models/embedding/xinference_embedding.py

@@ -1,8 +1,7 @@
-from core.third_party.langchain.embeddings.xinference_embedding import XinferenceEmbedding as XinferenceEmbeddings
-
 from core.model_providers.error import LLMBadRequestError
 from core.model_providers.providers.base import BaseModelProvider
 from core.model_providers.models.embedding.base import BaseEmbedding
+from core.third_party.langchain.embeddings.xinference_embedding import XinferenceEmbeddings
 
 
 class XinferenceEmbedding(BaseEmbedding):

api/core/model_providers/models/entity/message.py

@@ -1,4 +1,5 @@
 import enum
+from typing import Any, cast, Union, List, Dict
 
 from langchain.schema import HumanMessage, AIMessage, SystemMessage, BaseMessage, FunctionMessage
 from pydantic import BaseModel
@@ -18,17 +19,53 @@ class MessageType(enum.Enum):
     SYSTEM = 'system'
 
 
+class PromptMessageFileType(enum.Enum):
+    IMAGE = 'image'
+
+    @staticmethod
+    def value_of(value):
+        for member in PromptMessageFileType:
+            if member.value == value:
+                return member
+        raise ValueError(f"No matching enum found for value '{value}'")
+
+
+
+class PromptMessageFile(BaseModel):
+    type: PromptMessageFileType
+    data: Any
+
+
+class ImagePromptMessageFile(PromptMessageFile):
+    class DETAIL(enum.Enum):
+        LOW = 'low'
+        HIGH = 'high'
+
+    type: PromptMessageFileType = PromptMessageFileType.IMAGE
+    detail: DETAIL = DETAIL.LOW
+
+
 class PromptMessage(BaseModel):
     type: MessageType = MessageType.USER
     content: str = ''
+    files: list[PromptMessageFile] = []
     function_call: dict = None
 
 
+class LCHumanMessageWithFiles(HumanMessage):
+    # content: Union[str, List[Union[str, Dict]]]
+    content: str
+    files: list[PromptMessageFile]
+
+
 def to_lc_messages(messages: list[PromptMessage]):
     lc_messages = []
     for message in messages:
         if message.type == MessageType.USER:
-            lc_messages.append(HumanMessage(content=message.content))
+            if not message.files:
+                lc_messages.append(HumanMessage(content=message.content))
+            else:
+                lc_messages.append(LCHumanMessageWithFiles(content=message.content, files=message.files))
         elif message.type == MessageType.ASSISTANT:
             additional_kwargs = {}
             if message.function_call:
@@ -44,7 +81,14 @@ def to_prompt_messages(messages: list[BaseMessage]):
     prompt_messages = []
     for message in messages:
         if isinstance(message, HumanMessage):
-            prompt_messages.append(PromptMessage(content=message.content, type=MessageType.USER))
+            if isinstance(message, LCHumanMessageWithFiles):
+                prompt_messages.append(PromptMessage(
+                    content=message.content,
+                    type=MessageType.USER,
+                    files=message.files
+                ))
+            else:
+                prompt_messages.append(PromptMessage(content=message.content, type=MessageType.USER))
         elif isinstance(message, AIMessage):
             message_kwargs = {
                 'content': message.content,

api/core/model_providers/models/entity/provider.py

@@ -8,3 +8,4 @@ class ProviderQuotaUnit(Enum):
 
 class ModelFeature(Enum):
     AGENT_THOUGHT = 'agent_thought'
+    VISION = 'vision'

api/core/model_providers/models/llm/azure_openai_model.py

@@ -19,6 +19,13 @@ from core.model_providers.models.entity.model_params import ModelMode, ModelKwar
 AZURE_OPENAI_API_VERSION = '2023-07-01-preview'
 
 
+FUNCTION_CALL_MODELS = [
+    'gpt-4',
+    'gpt-4-32k',
+    'gpt-35-turbo',
+    'gpt-35-turbo-16k'
+]
+
 class AzureOpenAIModel(BaseLLM):
     def __init__(self, model_provider: BaseModelProvider,
                  name: str,
@@ -157,3 +164,7 @@ class AzureOpenAIModel(BaseLLM):
     @property
     def support_streaming(self):
         return True
+
+    @property
+    def support_function_call(self):
+        return self.base_model_name in FUNCTION_CALL_MODELS

api/core/model_providers/models/llm/baichuan_model.py

@@ -37,12 +37,6 @@ class BaichuanModel(BaseLLM):
         prompts = self._get_prompt_from_messages(messages)
         return self._client.generate([prompts], stop, callbacks)
 
-    def prompt_file_name(self, mode: str) -> str:
-        if mode == 'completion':
-            return 'baichuan_completion'
-        else:
-            return 'baichuan_chat'
-
     def get_num_tokens(self, messages: List[PromptMessage]) -> int:
         """
         get num tokens of prompt messages.

api/core/model_providers/models/llm/base.py

@@ -1,28 +1,18 @@
-import json
-import os
-import re
-import time
 from abc import abstractmethod
-from typing import List, Optional, Any, Union, Tuple
+from typing import List, Optional, Any, Union
 import decimal
+import logging
 
 from langchain.callbacks.manager import Callbacks
-from langchain.memory.chat_memory import BaseChatMemory
-from langchain.schema import LLMResult, SystemMessage, AIMessage, HumanMessage, BaseMessage, ChatGeneration
+from langchain.schema import LLMResult, BaseMessage, ChatGeneration
 
 from core.callback_handler.std_out_callback_handler import DifyStreamingStdOutCallbackHandler, DifyStdOutCallbackHandler
 from core.helper import moderation
 from core.model_providers.models.base import BaseProviderModel
-from core.model_providers.models.entity.message import PromptMessage, MessageType, LLMRunResult, to_prompt_messages, \
-    to_lc_messages
+from core.model_providers.models.entity.message import PromptMessage, MessageType, LLMRunResult, to_lc_messages
 from core.model_providers.models.entity.model_params import ModelType, ModelKwargs, ModelMode, ModelKwargsRules
 from core.model_providers.providers.base import BaseModelProvider
-from core.prompt.prompt_builder import PromptBuilder
-from core.prompt.prompt_template import PromptTemplateParser
 from core.third_party.langchain.llms.fake import FakeLLM
-import logging
-
-from extensions.ext_database import db
 
 logger = logging.getLogger(__name__)
 
@@ -320,206 +310,12 @@ class BaseLLM(BaseProviderModel):
     def support_streaming(self):
         return False
 
-    def get_prompt(self, mode: str,
-                   pre_prompt: str, inputs: dict,
-                   query: str,
-                   context: Optional[str],
-                   memory: Optional[BaseChatMemory]) -> \
-            Tuple[List[PromptMessage], Optional[List[str]]]:
-        prompt_rules = self._read_prompt_rules_from_file(self.prompt_file_name(mode))
-        prompt, stops = self._get_prompt_and_stop(prompt_rules, pre_prompt, inputs, query, context, memory)
-        return [PromptMessage(content=prompt)], stops
-
-    def get_advanced_prompt(self, app_mode: str,
-                   app_model_config: str, inputs: dict,
-                   query: str,
-                   context: Optional[str],
-                   memory: Optional[BaseChatMemory]) -> List[PromptMessage]:
-
-        model_mode = app_model_config.model_dict['mode']
-        conversation_histories_role = {}
-
-        raw_prompt_list = []
-        prompt_messages = []
-
-        if app_mode == 'chat' and model_mode == ModelMode.COMPLETION.value:
-            prompt_text = app_model_config.completion_prompt_config_dict['prompt']['text']
-            raw_prompt_list = [{
-                'role': MessageType.USER.value,
-                'text': prompt_text
-            }]
-            conversation_histories_role = app_model_config.completion_prompt_config_dict['conversation_histories_role']
-        elif app_mode == 'chat' and model_mode == ModelMode.CHAT.value:
-            raw_prompt_list = app_model_config.chat_prompt_config_dict['prompt']
-        elif app_mode == 'completion' and model_mode == ModelMode.CHAT.value:
-            raw_prompt_list = app_model_config.chat_prompt_config_dict['prompt']
-        elif app_mode == 'completion' and model_mode == ModelMode.COMPLETION.value:
-            prompt_text = app_model_config.completion_prompt_config_dict['prompt']['text']
-            raw_prompt_list = [{
-                'role': MessageType.USER.value,
-                'text': prompt_text
-            }]
-        else:
-            raise Exception("app_mode or model_mode not support")
-
-        for prompt_item in raw_prompt_list:
-            prompt = prompt_item['text']
-
-            # set prompt template variables
-            prompt_template = PromptTemplateParser(template=prompt)
-            prompt_inputs = {k: inputs[k] for k in prompt_template.variable_keys if k in inputs}
-
-            if '#context#' in prompt:
-                if context:
-                    prompt_inputs['#context#'] = context
-                else:
-                    prompt_inputs['#context#'] = ''
-
-            if '#query#' in prompt:
-                if query:
-                    prompt_inputs['#query#'] = query
-                else:
-                    prompt_inputs['#query#'] = ''
-
-            if '#histories#' in prompt:
-                if memory and app_mode == 'chat' and model_mode == ModelMode.COMPLETION.value:
-                    memory.human_prefix = conversation_histories_role['user_prefix']
-                    memory.ai_prefix = conversation_histories_role['assistant_prefix']
-                    histories = self._get_history_messages_from_memory(memory, 2000)
-                    prompt_inputs['#histories#'] = histories
-                else:
-                    prompt_inputs['#histories#'] = ''
-
-            prompt = prompt_template.format(
-                prompt_inputs
-            )
-
-            prompt = re.sub(r'<\|.*?\|>', '', prompt)
-
-            prompt_messages.append(PromptMessage(type = MessageType(prompt_item['role']) ,content=prompt))
-
-        if memory and app_mode == 'chat' and model_mode == ModelMode.CHAT.value:
-            memory.human_prefix = MessageType.USER.value
-            memory.ai_prefix = MessageType.ASSISTANT.value
-            histories = self._get_history_messages_list_from_memory(memory, 2000)
-            prompt_messages.extend(histories)
-
-        if app_mode == 'chat' and model_mode == ModelMode.CHAT.value:
-            prompt_messages.append(PromptMessage(type = MessageType.USER ,content=query))
-
-        return prompt_messages
-
-    def prompt_file_name(self, mode: str) -> str:
-        if mode == 'completion':
-            return 'common_completion'
-        else:
-            return 'common_chat'
-
-    def _get_prompt_and_stop(self, prompt_rules: dict, pre_prompt: str, inputs: dict,
-                             query: str,
-                             context: Optional[str],
-                             memory: Optional[BaseChatMemory]) -> Tuple[str, Optional[list]]:
-        context_prompt_content = ''
-        if context and 'context_prompt' in prompt_rules:
-            prompt_template = PromptTemplateParser(template=prompt_rules['context_prompt'])
-            context_prompt_content = prompt_template.format(
-                {'context': context}
-            )
-
-        pre_prompt_content = ''
-        if pre_prompt:
-            prompt_template = PromptTemplateParser(template=pre_prompt)
-            prompt_inputs = {k: inputs[k] for k in prompt_template.variable_keys if k in inputs}
-            pre_prompt_content = prompt_template.format(
-                prompt_inputs
-            )
-
-        prompt = ''
-        for order in prompt_rules['system_prompt_orders']:
-            if order == 'context_prompt':
-                prompt += context_prompt_content
-            elif order == 'pre_prompt':
-                prompt += pre_prompt_content
-
-        query_prompt = prompt_rules['query_prompt'] if 'query_prompt' in prompt_rules else '{{query}}'
-
-        if memory and 'histories_prompt' in prompt_rules:
-            # append chat histories
-            tmp_human_message = PromptBuilder.to_human_message(
-                prompt_content=prompt + query_prompt,
-                inputs={
-                    'query': query
-                }
-            )
-
-            if self.model_rules.max_tokens.max:
-                curr_message_tokens = self.get_num_tokens(to_prompt_messages([tmp_human_message]))
-                max_tokens = self.model_kwargs.max_tokens
-                rest_tokens = self.model_rules.max_tokens.max - max_tokens - curr_message_tokens
-                rest_tokens = max(rest_tokens, 0)
-            else:
-                rest_tokens = 2000
-
-            memory.human_prefix = prompt_rules['human_prefix'] if 'human_prefix' in prompt_rules else 'Human'
-            memory.ai_prefix = prompt_rules['assistant_prefix'] if 'assistant_prefix' in prompt_rules else 'Assistant'
-
-            histories = self._get_history_messages_from_memory(memory, rest_tokens)
-            prompt_template = PromptTemplateParser(template=prompt_rules['histories_prompt'])
-            histories_prompt_content = prompt_template.format({'histories': histories})
-
-            prompt = ''
-            for order in prompt_rules['system_prompt_orders']:
-                if order == 'context_prompt':
-                    prompt += context_prompt_content
-                elif order == 'pre_prompt':
-                    prompt += (pre_prompt_content + '\n') if pre_prompt_content else ''
-                elif order == 'histories_prompt':
-                    prompt += histories_prompt_content
-
-        prompt_template = PromptTemplateParser(template=query_prompt)
-        query_prompt_content = prompt_template.format({'query': query})
-
-        prompt += query_prompt_content
-
-        prompt = re.sub(r'<\|.*?\|>', '', prompt)
-
-        stops = prompt_rules.get('stops')
-        if stops is not None and len(stops) == 0:
-            stops = None
-
-        return prompt, stops
-
-    def _read_prompt_rules_from_file(self, prompt_name: str) -> dict:
-        # Get the absolute path of the subdirectory
-        prompt_path = os.path.join(
-            os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.realpath(__file__))))),
-            'prompt/generate_prompts')
-
-        json_file_path = os.path.join(prompt_path, f'{prompt_name}.json')
-        # Open the JSON file and read its content
-        with open(json_file_path, 'r') as json_file:
-            return json.load(json_file)
-
-    def _get_history_messages_from_memory(self, memory: BaseChatMemory,
-                                          max_token_limit: int) -> str:
-        """Get memory messages."""
-        memory.max_token_limit = max_token_limit
-        memory_key = memory.memory_variables[0]
-        external_context = memory.load_memory_variables({})
-        return external_context[memory_key]
-
-    def _get_history_messages_list_from_memory(self, memory: BaseChatMemory,
-                                          max_token_limit: int) -> List[PromptMessage]:
-        """Get memory messages."""
-        memory.max_token_limit = max_token_limit
-        memory.return_messages = True
-        memory_key = memory.memory_variables[0]
-        external_context = memory.load_memory_variables({})
-        memory.return_messages = False
-        return to_prompt_messages(external_context[memory_key])
+    @property
+    def support_function_call(self):
+        return False
 
     def _get_prompt_from_messages(self, messages: List[PromptMessage],
-                                  model_mode: Optional[ModelMode] = None) -> Union[str | List[BaseMessage]]:
+                                  model_mode: Optional[ModelMode] = None) -> Union[str , List[BaseMessage]]:
         if not model_mode:
             model_mode = self.model_mode
 

api/core/model_providers/models/llm/huggingface_hub_model.py

@@ -66,15 +66,6 @@ class HuggingfaceHubModel(BaseLLM):
         prompts = self._get_prompt_from_messages(messages)
         return self._client.get_num_tokens(prompts)
 
-    def prompt_file_name(self, mode: str) -> str:
-        if 'baichuan' in self.name.lower():
-            if mode == 'completion':
-                return 'baichuan_completion'
-            else:
-                return 'baichuan_chat'
-        else:
-            return super().prompt_file_name(mode)
-
     def _set_model_kwargs(self, model_kwargs: ModelKwargs):
         provider_model_kwargs = self._to_model_kwargs_input(self.model_rules, model_kwargs)
         self.client.model_kwargs = provider_model_kwargs

api/core/model_providers/models/llm/openai_model.py

@@ -1,11 +1,9 @@
-import decimal
 import logging
 from typing import List, Optional, Any
 
 import openai
 from langchain.callbacks.manager import Callbacks
 from langchain.schema import LLMResult
-from openai import api_requestor
 
 from core.model_providers.providers.base import BaseModelProvider
 from core.third_party.langchain.llms.chat_open_ai import EnhanceChatOpenAI
@@ -23,21 +21,36 @@ COMPLETION_MODELS = [
 ]
 
 CHAT_MODELS = [
+    'gpt-4-1106-preview',  # 128,000 tokens
+    'gpt-4-vision-preview',  # 128,000 tokens
     'gpt-4',  # 8,192 tokens
     'gpt-4-32k',  # 32,768 tokens
+    'gpt-3.5-turbo-1106',  # 16,384 tokens
     'gpt-3.5-turbo',  # 4,096 tokens
     'gpt-3.5-turbo-16k',  # 16,384 tokens
 ]
 
 MODEL_MAX_TOKENS = {
+    'gpt-4-1106-preview': 128000,
+    'gpt-4-vision-preview': 128000,
     'gpt-4': 8192,
     'gpt-4-32k': 32768,
+    'gpt-3.5-turbo-1106': 16384,
     'gpt-3.5-turbo': 4096,
-    'gpt-3.5-turbo-instruct': 8192,
+    'gpt-3.5-turbo-instruct': 4097,
     'gpt-3.5-turbo-16k': 16384,
     'text-davinci-003': 4097,
 }
 
+FUNCTION_CALL_MODELS = [
+    'gpt-4-1106-preview',
+    'gpt-4',
+    'gpt-4-32k',
+    'gpt-3.5-turbo-1106',
+    'gpt-3.5-turbo',
+    'gpt-3.5-turbo-16k'
+]
+
 
 class OpenAIModel(BaseLLM):
     def __init__(self, model_provider: BaseModelProvider,
@@ -50,7 +63,6 @@ class OpenAIModel(BaseLLM):
         else:
             self.model_mode = ModelMode.CHAT
         
-        # TODO load price config from configs(db)
         super().__init__(model_provider, name, model_kwargs, streaming, callbacks)
 
     def _init_client(self) -> Any:
@@ -100,7 +112,7 @@ class OpenAIModel(BaseLLM):
         :param callbacks:
         :return:
         """
-        if self.name == 'gpt-4' \
+        if self.name.startswith('gpt-4') \
                 and self.model_provider.provider.provider_type == ProviderType.SYSTEM.value \
                 and self.model_provider.provider.quota_type == ProviderQuotaType.TRIAL.value:
             raise ModelCurrentlyNotSupportError("Dify Hosted OpenAI GPT-4 currently not support.")
@@ -175,6 +187,10 @@ class OpenAIModel(BaseLLM):
     def support_streaming(self):
         return True
 
+    @property
+    def support_function_call(self):
+        return self.name in FUNCTION_CALL_MODELS
+
     # def is_model_valid_or_raise(self):
     #     """
     #     check is a valid model.

api/core/model_providers/models/llm/openllm_model.py

@@ -49,15 +49,6 @@ class OpenLLMModel(BaseLLM):
         prompts = self._get_prompt_from_messages(messages)
         return max(self._client.get_num_tokens(prompts), 0)
 
-    def prompt_file_name(self, mode: str) -> str:
-        if 'baichuan' in self.name.lower():
-            if mode == 'completion':
-                return 'baichuan_completion'
-            else:
-                return 'baichuan_chat'
-        else:
-            return super().prompt_file_name(mode)
-
     def _set_model_kwargs(self, model_kwargs: ModelKwargs):
         pass
 

api/core/model_providers/models/llm/wenxin_model.py

@@ -6,17 +6,16 @@ from langchain.schema import LLMResult
 
 from core.model_providers.error import LLMBadRequestError
 from core.model_providers.models.llm.base import BaseLLM
-from core.model_providers.models.entity.message import PromptMessage, MessageType
+from core.model_providers.models.entity.message import PromptMessage
 from core.model_providers.models.entity.model_params import ModelMode, ModelKwargs
 from core.third_party.langchain.llms.wenxin import Wenxin
 
 
 class WenxinModel(BaseLLM):
-    model_mode: ModelMode = ModelMode.COMPLETION
+    model_mode: ModelMode = ModelMode.CHAT
 
     def _init_client(self) -> Any:
         provider_model_kwargs = self._to_model_kwargs_input(self.model_rules, self.model_kwargs)
-        # TODO load price_config from configs(db)
         return Wenxin(
             model=self.name,
             streaming=self.streaming,
@@ -38,7 +37,13 @@ class WenxinModel(BaseLLM):
         :return:
         """
         prompts = self._get_prompt_from_messages(messages)
-        return self._client.generate([prompts], stop, callbacks)
+
+        generate_kwargs = {'stop': stop, 'callbacks': callbacks, 'messages': [prompts]}
+
+        if 'functions' in kwargs:
+            generate_kwargs['functions'] = kwargs['functions']
+
+        return self._client.generate(**generate_kwargs)
 
     def get_num_tokens(self, messages: List[PromptMessage]) -> int:
         """
@@ -48,7 +53,7 @@ class WenxinModel(BaseLLM):
         :return:
         """
         prompts = self._get_prompt_from_messages(messages)
-        return max(self._client.get_num_tokens(prompts), 0)
+        return max(self._client.get_num_tokens_from_messages(prompts), 0)
 
     def _set_model_kwargs(self, model_kwargs: ModelKwargs):
         provider_model_kwargs = self._to_model_kwargs_input(self.model_rules, model_kwargs)
@@ -58,3 +63,7 @@ class WenxinModel(BaseLLM):
 
     def handle_exceptions(self, ex: Exception) -> Exception:
         return LLMBadRequestError(f"Wenxin: {str(ex)}")
+
+    @property
+    def support_streaming(self):
+        return True

api/core/model_providers/models/llm/xinference_model.py

@@ -59,15 +59,6 @@ class XinferenceModel(BaseLLM):
         prompts = self._get_prompt_from_messages(messages)
         return max(self._client.get_num_tokens(prompts), 0)
 
-    def prompt_file_name(self, mode: str) -> str:
-        if 'baichuan' in self.name.lower():
-            if mode == 'completion':
-                return 'baichuan_completion'
-            else:
-                return 'baichuan_chat'
-        else:
-            return super().prompt_file_name(mode)
-
     def _set_model_kwargs(self, model_kwargs: ModelKwargs):
         pass
 

api/core/model_providers/models/llm/zhipuai_model.py

@@ -16,6 +16,7 @@ class ZhipuAIModel(BaseLLM):
     def _init_client(self) -> Any:
         provider_model_kwargs = self._to_model_kwargs_input(self.model_rules, self.model_kwargs)
         return ZhipuAIChatLLM(
+            model=self.name,
             streaming=self.streaming,
             callbacks=self.callbacks,
             **self.credentials,

api/core/model_providers/providers/anthropic_provider.py

@@ -172,7 +172,7 @@ class AnthropicProvider(BaseModelProvider):
 
     def should_deduct_quota(self):
         if hosted_model_providers.anthropic and \
-                hosted_model_providers.anthropic.quota_limit and hosted_model_providers.anthropic.quota_limit > 0:
+                hosted_model_providers.anthropic.quota_limit and hosted_model_providers.anthropic.quota_limit > -1:
             return True
 
         return False

api/core/model_providers/providers/azure_openai_provider.py

@@ -329,7 +329,7 @@ class AzureOpenAIProvider(BaseModelProvider):
 
     def should_deduct_quota(self):
         if hosted_model_providers.azure_openai \
-                and hosted_model_providers.azure_openai.quota_limit and hosted_model_providers.azure_openai.quota_limit > 0:
+                and hosted_model_providers.azure_openai.quota_limit and hosted_model_providers.azure_openai.quota_limit > -1:
             return True
 
         return False

api/core/model_providers/providers/hosted.py

@@ -11,7 +11,7 @@ class HostedOpenAI(BaseModel):
     api_organization: str = None
     api_key: str
     quota_limit: int = 0
-    """Quota limit for the openai hosted model. 0 means unlimited."""
+    """Quota limit for the openai hosted model. -1 means unlimited."""
     paid_enabled: bool = False
     paid_stripe_price_id: str = None
     paid_increase_quota: int = 1
@@ -21,14 +21,14 @@ class HostedAzureOpenAI(BaseModel):
     api_base: str
     api_key: str
     quota_limit: int = 0
-    """Quota limit for the azure openai hosted model. 0 means unlimited."""
+    """Quota limit for the azure openai hosted model. -1 means unlimited."""
 
 
 class HostedAnthropic(BaseModel):
     api_base: str = None
     api_key: str
     quota_limit: int = 0
-    """Quota limit for the anthropic hosted model. 0 means unlimited."""
+    """Quota limit for the anthropic hosted model. -1 means unlimited."""
     paid_enabled: bool = False
     paid_stripe_price_id: str = None
     paid_increase_quota: int = 1000000

api/core/model_providers/providers/openai_provider.py

@@ -41,9 +41,17 @@ class OpenAIProvider(BaseModelProvider):
                         ModelFeature.AGENT_THOUGHT.value
                     ]
                 },
+                {
+                    'id': 'gpt-3.5-turbo-1106',
+                    'name': 'gpt-3.5-turbo-1106',
+                    'mode': ModelMode.CHAT.value,
+                    'features': [
+                        ModelFeature.AGENT_THOUGHT.value
+                    ]
+                },
                 {
                     'id': 'gpt-3.5-turbo-instruct',
-                    'name': 'GPT-3.5-Turbo-Instruct',
+                    'name': 'gpt-3.5-turbo-instruct',
                     'mode': ModelMode.COMPLETION.value,
                 },
                 {
@@ -62,6 +70,22 @@ class OpenAIProvider(BaseModelProvider):
                         ModelFeature.AGENT_THOUGHT.value
                     ]
                 },
+                {
+                    'id': 'gpt-4-1106-preview',
+                    'name': 'gpt-4-1106-preview',
+                    'mode': ModelMode.CHAT.value,
+                    'features': [
+                        ModelFeature.AGENT_THOUGHT.value
+                    ]
+                },
+                {
+                    'id': 'gpt-4-vision-preview',
+                    'name': 'gpt-4-vision-preview',
+                    'mode': ModelMode.CHAT.value,
+                    'features': [
+                        ModelFeature.VISION.value
+                    ]
+                },
                 {
                     'id': 'gpt-4-32k',
                     'name': 'gpt-4-32k',
@@ -79,7 +103,7 @@ class OpenAIProvider(BaseModelProvider):
 
             if self.provider.provider_type == ProviderType.SYSTEM.value \
                     and self.provider.quota_type == ProviderQuotaType.TRIAL.value:
-                models = [item for item in models if item['id'] not in ['gpt-4', 'gpt-4-32k']]
+                models = [item for item in models if not item['id'].startswith('gpt-4')]
 
             return models
         elif model_type == ModelType.EMBEDDINGS:
@@ -141,10 +165,13 @@ class OpenAIProvider(BaseModelProvider):
         :return:
         """
         model_max_tokens = {
+            'gpt-4-1106-preview': 128000,
+            'gpt-4-vision-preview': 128000,
             'gpt-4': 8192,
             'gpt-4-32k': 32768,
+            'gpt-3.5-turbo-1106': 16384,
             'gpt-3.5-turbo': 4096,
-            'gpt-3.5-turbo-instruct': 8192,
+            'gpt-3.5-turbo-instruct': 4097,
             'gpt-3.5-turbo-16k': 16384,
             'text-davinci-003': 4097,
         }
@@ -250,7 +277,7 @@ class OpenAIProvider(BaseModelProvider):
 
     def should_deduct_quota(self):
         if hosted_model_providers.openai \
-                and hosted_model_providers.openai.quota_limit and hosted_model_providers.openai.quota_limit > 0:
+                and hosted_model_providers.openai.quota_limit and hosted_model_providers.openai.quota_limit > -1:
             return True
 
         return False

api/core/model_providers/providers/spark_provider.py

@@ -28,14 +28,19 @@ class SparkProvider(BaseModelProvider):
         if model_type == ModelType.TEXT_GENERATION:
             return [
                 {
-                    'id': 'spark',
-                    'name': 'Spark V1.5',
+                    'id': 'spark-v3',
+                    'name': 'Spark V3.0',
                     'mode': ModelMode.CHAT.value,
                 },
                 {
                     'id': 'spark-v2',
                     'name': 'Spark V2.0',
                     'mode': ModelMode.CHAT.value,
+                },
+                {
+                    'id': 'spark',
+                    'name': 'Spark V1.5',
+                    'mode': ModelMode.CHAT.value,
                 }
             ]
         else:
@@ -96,7 +101,7 @@ class SparkProvider(BaseModelProvider):
 
         try:
             chat_llm = ChatSpark(
-                model_name='spark-v2',
+                model_name='spark-v3',
                 max_tokens=10,
                 temperature=0.01,
                 **credential_kwargs
@@ -110,10 +115,10 @@ class SparkProvider(BaseModelProvider):
 
             chat_llm(messages)
         except SparkError as ex:
-            # try spark v1.5 if v2.1 failed
+            # try spark v2.1 if v3.1 failed
             try:
                 chat_llm = ChatSpark(
-                    model_name='spark',
+                    model_name='spark-v2',
                     max_tokens=10,
                     temperature=0.01,
                     **credential_kwargs
@@ -127,10 +132,27 @@ class SparkProvider(BaseModelProvider):
 
                 chat_llm(messages)
             except SparkError as ex:
-                raise CredentialsValidateFailedError(str(ex))
-            except Exception as ex:
-                logging.exception('Spark config validation failed')
-                raise ex
+                # try spark v1.5 if v2.1 failed
+                try:
+                    chat_llm = ChatSpark(
+                        model_name='spark',
+                        max_tokens=10,
+                        temperature=0.01,
+                        **credential_kwargs
+                    )
+
+                    messages = [
+                        HumanMessage(
+                            content="ping"
+                        )
+                    ]
+
+                    chat_llm(messages)
+                except SparkError as ex:
+                    raise CredentialsValidateFailedError(str(ex))
+                except Exception as ex:
+                    logging.exception('Spark config validation failed')
+                    raise ex
         except Exception as ex:
             logging.exception('Spark config validation failed')
             raise ex

api/core/model_providers/providers/wenxin_provider.py

@@ -2,6 +2,8 @@ import json
 from json import JSONDecodeError
 from typing import Type
 
+from langchain.schema import HumanMessage
+
 from core.helper import encrypter
 from core.model_providers.models.base import BaseProviderModel
 from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
@@ -23,20 +25,25 @@ class WenxinProvider(BaseModelProvider):
     def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
         if model_type == ModelType.TEXT_GENERATION:
             return [
+                {
+                    'id': 'ernie-bot-4',
+                    'name': 'ERNIE-Bot-4',
+                    'mode': ModelMode.CHAT.value,
+                },
                 {
                     'id': 'ernie-bot',
                     'name': 'ERNIE-Bot',
-                    'mode': ModelMode.COMPLETION.value,
+                    'mode': ModelMode.CHAT.value,
                 },
                 {
                     'id': 'ernie-bot-turbo',
                     'name': 'ERNIE-Bot-turbo',
-                    'mode': ModelMode.COMPLETION.value,
+                    'mode': ModelMode.CHAT.value,
                 },
                 {
                     'id': 'bloomz-7b',
                     'name': 'BLOOMZ-7B',
-                    'mode': ModelMode.COMPLETION.value,
+                    'mode': ModelMode.CHAT.value,
                 }
             ]
         else:
@@ -68,11 +75,12 @@ class WenxinProvider(BaseModelProvider):
         :return:
         """
         model_max_tokens = {
+            'ernie-bot-4': 4800,
             'ernie-bot': 4800,
             'ernie-bot-turbo': 11200,
         }
 
-        if model_name in ['ernie-bot', 'ernie-bot-turbo']:
+        if model_name in ['ernie-bot-4', 'ernie-bot', 'ernie-bot-turbo']:
             return ModelKwargsRules(
                 temperature=KwargRule[float](min=0.01, max=1, default=0.95, precision=2),
                 top_p=KwargRule[float](min=0.01, max=1, default=0.8, precision=2),
@@ -111,7 +119,7 @@ class WenxinProvider(BaseModelProvider):
                 **credential_kwargs
             )
 
-            llm("ping")
+            llm([HumanMessage(content='ping')])
         except Exception as ex:
             raise CredentialsValidateFailedError(str(ex))
 

api/core/model_providers/providers/xinference_provider.py

@@ -2,7 +2,6 @@ import json
 from typing import Type
 
 import requests
-from langchain.embeddings import XinferenceEmbeddings
 
 from core.helper import encrypter
 from core.model_providers.models.embedding.xinference_embedding import XinferenceEmbedding
@@ -11,6 +10,7 @@ from core.model_providers.models.llm.xinference_model import XinferenceModel
 from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
 
 from core.model_providers.models.base import BaseProviderModel
+from core.third_party.langchain.embeddings.xinference_embedding import XinferenceEmbeddings
 from core.third_party.langchain.llms.xinference_llm import XinferenceLLM
 from models.provider import ProviderType
 

api/core/model_providers/providers/zhipuai_provider.py

@@ -26,6 +26,11 @@ class ZhipuAIProvider(BaseModelProvider):
     def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
         if model_type == ModelType.TEXT_GENERATION:
             return [
+                {
+                    'id': 'chatglm_turbo',
+                    'name': 'chatglm_turbo',
+                    'mode': ModelMode.CHAT.value,
+                },
                 {
                     'id': 'chatglm_pro',
                     'name': 'chatglm_pro',

api/core/model_providers/rules/anthropic.json

@@ -9,7 +9,7 @@
             "trial"
         ],
         "quota_unit": "tokens",
-        "quota_limit": 600000
+        "quota_limit": 0
     },
     "model_flexibility": "fixed",
     "price_config": {

api/core/model_providers/rules/openai.json

@@ -24,12 +24,30 @@
             "unit": "0.001",
             "currency": "USD"
         },
+        "gpt-4-1106-preview": {
+            "prompt": "0.01",
+            "completion": "0.03",
+            "unit": "0.001",
+            "currency": "USD"
+        },
+        "gpt-4-vision-preview": {
+            "prompt": "0.01",
+            "completion": "0.03",
+            "unit": "0.001",
+            "currency": "USD"
+        },
         "gpt-3.5-turbo": {
             "prompt": "0.0015",
             "completion": "0.002",
             "unit": "0.001",
             "currency": "USD"
         },
+        "gpt-3.5-turbo-1106": {
+            "prompt": "0.0010",
+            "completion": "0.002",
+            "unit": "0.001",
+            "currency": "USD"
+        },
         "gpt-3.5-turbo-instruct": {
             "prompt": "0.0015",
             "completion": "0.002",

api/core/model_providers/rules/spark.json

@@ -22,6 +22,12 @@
             "completion": "0.36",
             "unit": "0.0001",
             "currency": "RMB"
+        },
+        "spark-v3": {
+            "prompt": "0.36",
+            "completion": "0.36",
+            "unit": "0.0001",
+            "currency": "RMB"
         }
     }
 }

api/core/model_providers/rules/wenxin.json

@@ -5,6 +5,12 @@
     "system_config": null,
     "model_flexibility": "fixed",
     "price_config": {
+        "ernie-bot-4": {
+            "prompt": "0",
+            "completion": "0",
+            "unit": "0.001",
+            "currency": "RMB"
+        },
         "ernie-bot": {
             "prompt": "0.012",
             "completion": "0.012",

api/core/model_providers/rules/zhipuai.json

@@ -11,6 +11,12 @@
     },
     "model_flexibility": "fixed",
     "price_config": {
+        "chatglm_turbo": {
+            "prompt": "0.005",
+            "completion": "0.005",
+            "unit": "0.001",
+            "currency": "RMB"
+        },
         "chatglm_pro": {
             "prompt": "0.01",
             "completion": "0.01",

api/core/moderation/api/__builtin__

@@ -0,0 +1 @@
+3

api/core/moderation/api/api.py

@@ -0,0 +1,88 @@
+from pydantic import BaseModel
+
+from core.moderation.base import Moderation, ModerationInputsResult, ModerationOutputsResult, ModerationAction
+from core.extension.api_based_extension_requestor import APIBasedExtensionRequestor, APIBasedExtensionPoint
+from core.helper.encrypter import decrypt_token
+from extensions.ext_database import db
+from models.api_based_extension import APIBasedExtension
+
+
+class ModerationInputParams(BaseModel):
+    app_id: str = ""
+    inputs: dict = {}
+    query: str = ""
+
+
+class ModerationOutputParams(BaseModel):
+    app_id: str = ""
+    text: str
+
+
+class ApiModeration(Moderation):
+    name: str = "api"
+
+    @classmethod
+    def validate_config(cls, tenant_id: str, config: dict) -> None:
+        """
+        Validate the incoming form config data.
+
+        :param tenant_id: the id of workspace
+        :param config: the form config data
+        :return:
+        """
+        cls._validate_inputs_and_outputs_config(config, False)
+
+        api_based_extension_id = config.get("api_based_extension_id")
+        if not api_based_extension_id:
+            raise ValueError("api_based_extension_id is required")
+
+        extension = cls._get_api_based_extension(tenant_id, api_based_extension_id)
+        if not extension:
+            raise ValueError("API-based Extension not found. Please check it again.")
+
+    def moderation_for_inputs(self, inputs: dict, query: str = "") -> ModerationInputsResult:
+        flagged = False
+        preset_response = ""
+
+        if self.config['inputs_config']['enabled']:
+            params = ModerationInputParams(
+                app_id=self.app_id,
+                inputs=inputs,
+                query=query
+            )
+
+            result = self._get_config_by_requestor(APIBasedExtensionPoint.APP_MODERATION_INPUT, params.dict())
+            return ModerationInputsResult(**result)
+
+        return ModerationInputsResult(flagged=flagged, action=ModerationAction.DIRECT_OUTPUT, preset_response=preset_response)
+
+    def moderation_for_outputs(self, text: str) -> ModerationOutputsResult:
+        flagged = False
+        preset_response = ""
+
+        if self.config['outputs_config']['enabled']:
+            params = ModerationOutputParams(
+                app_id=self.app_id,
+                text=text
+            )
+
+            result = self._get_config_by_requestor(APIBasedExtensionPoint.APP_MODERATION_OUTPUT, params.dict())
+            return ModerationOutputsResult(**result)
+
+        return ModerationOutputsResult(flagged=flagged, action=ModerationAction.DIRECT_OUTPUT, preset_response=preset_response)
+
+    def _get_config_by_requestor(self, extension_point: APIBasedExtensionPoint, params: dict) -> dict:
+        extension = self._get_api_based_extension(self.tenant_id, self.config.get("api_based_extension_id"))
+        requestor = APIBasedExtensionRequestor(extension.api_endpoint, decrypt_token(self.tenant_id, extension.api_key))
+
+        result = requestor.request(extension_point, params)
+        return result
+
+    @staticmethod
+    def _get_api_based_extension(tenant_id: str, api_based_extension_id: str) -> APIBasedExtension:
+        extension = db.session.query(APIBasedExtension).filter(
+            APIBasedExtension.tenant_id == tenant_id,
+            APIBasedExtension.id == api_based_extension_id
+        ).first()
+
+        return extension

api/core/moderation/base.py

@@ -0,0 +1,113 @@
+from abc import ABC, abstractmethod
+from typing import Optional
+from pydantic import BaseModel
+from enum import Enum
+
+from core.extension.extensible import Extensible, ExtensionModule
+
+
+class ModerationAction(Enum):
+    DIRECT_OUTPUT = 'direct_output'
+    OVERRIDED = 'overrided'
+
+
+class ModerationInputsResult(BaseModel):
+    flagged: bool = False
+    action: ModerationAction
+    preset_response: str = ""
+    inputs: dict = {}
+    query: str = ""
+
+
+class ModerationOutputsResult(BaseModel):
+    flagged: bool = False
+    action: ModerationAction
+    preset_response: str = ""
+    text: str = ""
+
+
+class Moderation(Extensible, ABC):
+    """
+    The base class of moderation.
+    """
+    module: ExtensionModule = ExtensionModule.MODERATION
+
+    def __init__(self, app_id: str, tenant_id: str, config: Optional[dict] = None) -> None:
+        super().__init__(tenant_id, config)
+        self.app_id = app_id
+
+    @classmethod
+    @abstractmethod
+    def validate_config(cls, tenant_id: str, config: dict) -> None:
+        """
+        Validate the incoming form config data.
+
+        :param tenant_id: the id of workspace
+        :param config: the form config data
+        :return:
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def moderation_for_inputs(self, inputs: dict, query: str = "") -> ModerationInputsResult:
+        """
+        Moderation for inputs.
+        After the user inputs, this method will be called to perform sensitive content review
+        on the user inputs and return the processed results.
+
+        :param inputs: user inputs
+        :param query: query string (required in chat app)
+        :return:
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def moderation_for_outputs(self, text: str) -> ModerationOutputsResult:
+        """
+        Moderation for outputs.
+        When LLM outputs content, the front end will pass the output content (may be segmented)
+        to this method for sensitive content review, and the output content will be shielded if the review fails.
+
+        :param text: LLM output content
+        :return:
+        """
+        raise NotImplementedError
+
+    @classmethod
+    def _validate_inputs_and_outputs_config(self, config: dict, is_preset_response_required: bool) -> None:
+        # inputs_config
+        inputs_config = config.get("inputs_config")
+        if not isinstance(inputs_config, dict):
+            raise ValueError("inputs_config must be a dict")
+
+        # outputs_config
+        outputs_config = config.get("outputs_config")
+        if not isinstance(outputs_config, dict):
+            raise ValueError("outputs_config must be a dict")
+
+        inputs_config_enabled = inputs_config.get("enabled")
+        outputs_config_enabled = outputs_config.get("enabled")
+        if not inputs_config_enabled and not outputs_config_enabled:
+            raise ValueError("At least one of inputs_config or outputs_config must be enabled")
+
+        # preset_response
+        if not is_preset_response_required:
+            return
+
+        if inputs_config_enabled:
+            if not inputs_config.get("preset_response"):
+                raise ValueError("inputs_config.preset_response is required")
+
+            if len(inputs_config.get("preset_response")) > 100:
+                raise ValueError("inputs_config.preset_response must be less than 100 characters")
+
+        if outputs_config_enabled:
+            if not outputs_config.get("preset_response"):
+                raise ValueError("outputs_config.preset_response is required")
+
+            if len(outputs_config.get("preset_response")) > 100:
+                raise ValueError("outputs_config.preset_response must be less than 100 characters")
+
+
+class ModerationException(Exception):
+    pass

api/core/moderation/factory.py

@@ -0,0 +1,48 @@
+from core.extension.extensible import ExtensionModule
+from core.moderation.base import Moderation, ModerationInputsResult, ModerationOutputsResult
+from extensions.ext_code_based_extension import code_based_extension
+
+
+class ModerationFactory:
+    __extension_instance: Moderation
+
+    def __init__(self, name: str, app_id: str, tenant_id: str, config: dict) -> None:
+        extension_class = code_based_extension.extension_class(ExtensionModule.MODERATION, name)
+        self.__extension_instance = extension_class(app_id, tenant_id, config)
+
+    @classmethod
+    def validate_config(cls, name: str, tenant_id: str, config: dict) -> None:
+        """
+        Validate the incoming form config data.
+
+        :param name: the name of extension
+        :param tenant_id: the id of workspace
+        :param config: the form config data
+        :return:
+        """
+        code_based_extension.validate_form_schema(ExtensionModule.MODERATION, name, config)
+        extension_class = code_based_extension.extension_class(ExtensionModule.MODERATION, name)
+        extension_class.validate_config(tenant_id, config)
+
+    def moderation_for_inputs(self, inputs: dict, query: str = "") -> ModerationInputsResult:
+        """
+        Moderation for inputs.
+        After the user inputs, this method will be called to perform sensitive content review
+        on the user inputs and return the processed results.
+
+        :param inputs: user inputs
+        :param query: query string (required in chat app)
+        :return:
+        """
+        return self.__extension_instance.moderation_for_inputs(inputs, query)
+
+    def moderation_for_outputs(self, text: str) -> ModerationOutputsResult:
+        """
+        Moderation for outputs.
+        When LLM outputs content, the front end will pass the output content (may be segmented)
+        to this method for sensitive content review, and the output content will be shielded if the review fails.
+
+        :param text: LLM output content
+        :return:
+        """
+        return self.__extension_instance.moderation_for_outputs(text)

api/core/moderation/keywords/__builtin__

@@ -0,0 +1 @@
+2

api/core/moderation/keywords/keywords.py

@@ -0,0 +1,60 @@
+from core.moderation.base import Moderation, ModerationInputsResult, ModerationOutputsResult, ModerationAction
+
+
+class KeywordsModeration(Moderation):
+    name: str = "keywords"
+
+    @classmethod
+    def validate_config(cls, tenant_id: str, config: dict) -> None:
+        """
+        Validate the incoming form config data.
+
+        :param tenant_id: the id of workspace
+        :param config: the form config data
+        :return:
+        """
+        cls._validate_inputs_and_outputs_config(config, True)
+
+        if not config.get("keywords"):
+            raise ValueError("keywords is required")
+
+        if len(config.get("keywords")) > 1000:
+            raise ValueError("keywords length must be less than 1000")
+
+    def moderation_for_inputs(self, inputs: dict, query: str = "") -> ModerationInputsResult:
+        flagged = False
+        preset_response = ""
+
+        if self.config['inputs_config']['enabled']:
+            preset_response = self.config['inputs_config']['preset_response']
+
+            if query:
+                inputs['query__'] = query
+            keywords_list = self.config['keywords'].split('\n')
+            flagged = self._is_violated(inputs, keywords_list)
+
+        return ModerationInputsResult(flagged=flagged, action=ModerationAction.DIRECT_OUTPUT, preset_response=preset_response)
+
+    def moderation_for_outputs(self, text: str) -> ModerationOutputsResult:
+        flagged = False
+        preset_response = ""
+
+        if self.config['outputs_config']['enabled']:
+            keywords_list = self.config['keywords'].split('\n')
+            flagged = self._is_violated({'text': text}, keywords_list)
+            preset_response = self.config['outputs_config']['preset_response']
+
+        return ModerationOutputsResult(flagged=flagged, action=ModerationAction.DIRECT_OUTPUT, preset_response=preset_response)
+
+    def _is_violated(self, inputs: dict, keywords_list: list) -> bool:
+        for value in inputs.values():
+            if self._check_keywords_in_value(keywords_list, value):
+                return True
+
+        return False
+
+    def _check_keywords_in_value(self, keywords_list, value):
+        for keyword in keywords_list:
+            if keyword.lower() in value.lower():
+                return True
+        return False

api/core/moderation/openai_moderation/__builtin__

@@ -0,0 +1 @@
+1