feat: add embeddings provider configuration and factory

Introduce a flexible embeddings configuration system:
- Add EMBEDDINGS_PROVIDER to .env.example and config
- Create EmbeddingsFactory to dynamically select embedding providers
- Refactor existing code to use the new embedding factory
- Prepare for future multi-provider embedding support

Author: JohannLai

backend/.env.example

@@ -8,6 +8,10 @@ MYSQL_DATABASE=ragwebui
 SECRET_KEY=your-secret-key-here
 ACCESS_TOKEN_EXPIRE_MINUTES=30
 
+# Embeddings settings
+# if set to openai, use the openai api key and base url
+EMBEDDINGS_PROVIDER=openai
+
 # Vector DB settings
 CHROMA_DB_HOST=localhost
 CHROMA_DB_PORT=8001

backend/app/api/api_v1/knowledge_base.py

@@ -3,7 +3,6 @@
 from fastapi import APIRouter, Depends, HTTPException, UploadFile, File, BackgroundTasks, Query
 from sqlalchemy.orm import Session
 from langchain_chroma import Chroma
-from langchain_openai import OpenAIEmbeddings
 from sqlalchemy import text
 import logging
 from datetime import datetime, timedelta
@@ -25,6 +24,7 @@
 from app.core.minio import get_minio_client
 from minio.error import MinioException
 from app.services.vector_store import VectorStoreFactory
+from app.services.embedding.embedding_factory import EmbeddingsFactory
 
 router = APIRouter()
 
@@ -162,10 +162,7 @@ async def delete_knowledge_base(
 
         # Initialize services
         minio_client = get_minio_client()
-        embeddings = OpenAIEmbeddings(
-            openai_api_key=settings.OPENAI_API_KEY,
-            openai_api_base=settings.OPENAI_API_BASE
-        )
+        embeddings = EmbeddingsFactory.create()
 
         vector_store = VectorStoreFactory.create(
             store_type=settings.VECTOR_STORE_TYPE,
@@ -511,10 +508,7 @@ async def test_retrieval(
                 detail=f"Knowledge base {request.kb_id} not found",
             )
 
-        embeddings = OpenAIEmbeddings(
-            openai_api_key=settings.OPENAI_API_KEY,
-            openai_api_base=settings.OPENAI_API_BASE
-        )
+        embeddings = EmbeddingsFactory.create()
 
         vector_store = VectorStoreFactory.create(
             store_type=settings.VECTOR_STORE_TYPE,

backend/app/api/openapi/knowledge.py

@@ -2,13 +2,13 @@
 from fastapi import APIRouter, Depends, HTTPException
 from sqlalchemy.orm import Session
 from langchain_chroma import Chroma
-from langchain_openai import OpenAIEmbeddings
 from app.services.vector_store import VectorStoreFactory
 
 from app import models
 from app.db.session import get_db
 from app.core.security import get_api_key_user
 from app.core.config import settings
+from app.services.embedding.embedding_factory import EmbeddingsFactory
 
 router = APIRouter()
 
@@ -36,10 +36,7 @@ def query_knowledge_base(
                 detail=f"Knowledge base {knowledge_base_id} not found",
             )
 
-        embeddings = OpenAIEmbeddings(
-            openai_api_key=settings.OPENAI_API_KEY,
-            openai_api_base=settings.OPENAI_API_BASE
-        )
+        embeddings = EmbeddingsFactory.create()
 
         vector_store = VectorStoreFactory.create(
             store_type=settings.VECTOR_STORE_TYPE,

backend/app/core/config.py

@@ -36,6 +36,9 @@ def get_database_url(self) -> str:
     ALGORITHM: str = "HS256"
     ACCESS_TOKEN_EXPIRE_MINUTES: int = 10080
 
+    # Embeddings settings
+    EMBEDDINGS_PROVIDER: str = os.getenv("EMBEDDINGS_PROVIDER", "openai")
+
     # Vector DB settings
     CHROMA_DB_HOST: str = os.getenv("CHROMA_DB_HOST", "localhost")
     CHROMA_DB_PORT: int = int(os.getenv("CHROMA_DB_PORT", "8001"))

backend/app/services/chat_service.py

@@ -2,8 +2,6 @@
 import base64
 from typing import List, AsyncGenerator
 from sqlalchemy.orm import Session
-from langchain_chroma import Chroma
-from langchain_openai import OpenAIEmbeddings
 from langchain_openai import ChatOpenAI
 from langchain.chains import create_history_aware_retriever, create_retrieval_chain
 from langchain.chains.combine_documents import create_stuff_documents_chain
@@ -14,6 +12,7 @@
 from app.models.knowledge import KnowledgeBase, Document
 from langchain.globals import set_verbose, set_debug
 from app.services.vector_store import VectorStoreFactory
+from app.services.embedding.embedding_factory import EmbeddingsFactory
 
 set_verbose(True)
 set_debug(True)
@@ -52,10 +51,7 @@ async def generate_response(
         )
 
         # Initialize embeddings
-        embeddings = OpenAIEmbeddings(
-            openai_api_key=settings.OPENAI_API_KEY,
-            openai_api_base=settings.OPENAI_API_BASE
-        )
+        embeddings = EmbeddingsFactory.create()
 
         # Create a vector store for each knowledge base
         vector_stores = []

backend/app/services/document_processor.py

@@ -16,8 +16,6 @@
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_core.documents import Document as LangchainDocument
 from pydantic import BaseModel
-from langchain_openai import OpenAIEmbeddings
-from langchain_chroma import Chroma
 from sqlalchemy import create_engine, text
 from sqlalchemy.orm import Session
 from app.core.config import settings
@@ -31,6 +29,7 @@
 from minio import Minio
 from minio.commonconfig import CopySource
 from app.services.vector_store import VectorStoreFactory
+from app.services.embedding.embedding_factory import EmbeddingsFactory
 
 class UploadResult(BaseModel):
     file_path: str
@@ -56,10 +55,7 @@ async def process_document(file_path: str, file_name: str, kb_id: int, document_
 
         # Initialize embeddings
         logger.info("Initializing OpenAI embeddings...")
-        embeddings = OpenAIEmbeddings(
-            openai_api_key=settings.OPENAI_API_KEY,
-            openai_api_base=settings.OPENAI_API_BASE
-        )
+        embeddings = EmbeddingsFactory.create()
 
         logger.info(f"Initializing vector store with collection: kb_{kb_id}")
         vector_store = VectorStoreFactory.create(
@@ -303,10 +299,7 @@ async def process_document_background(
 
             # 3. 创建向量存储
             logger.info(f"Task {task_id}: Initializing vector store")
-            embeddings = OpenAIEmbeddings(
-                openai_api_key=settings.OPENAI_API_KEY,
-                openai_api_base=settings.OPENAI_API_BASE
-            )
+            embeddings = EmbeddingsFactory.create()
 
             vector_store = VectorStoreFactory.create(
                 store_type=settings.VECTOR_STORE_TYPE,

backend/app/services/embedding/embedding_factory.py

@@ -0,0 +1,24 @@
+from app.core.config import settings
+from langchain_openai import OpenAIEmbeddings
+# If you plan on adding other embeddings, import them here
+# from some_other_module import AnotherEmbeddingClass
+
+class EmbeddingsFactory:
+    @staticmethod
+    def create():
+        """
+        Factory method to create an embeddings instance based on .env config.
+        """
+        # Suppose your .env has a value like EMBEDDINGS_PROVIDER=openai
+        embeddings_provider = settings.EMBEDDINGS_PROVIDER.lower()
+
+        if embeddings_provider == "openai":
+            return OpenAIEmbeddings(
+                openai_api_key=settings.OPENAI_API_KEY,
+                openai_api_base=settings.OPENAI_API_BASE
+            )
+        # Extend with other providers:
+        # elif embeddings_provider == "another_provider":
+        #     return AnotherEmbeddingClass(...)
+        else:
+            raise ValueError(f"Unsupported embeddings provider: {embeddings_provider}")