Agents Langchain 集成

# LangChain 集成指南 本指南涵盖了向量数据库集成、LangSmith 可观测性以及部署相关内容。 ## 向量数据库集成 ### Chroma (本地、开源) python from langchain_chroma import Chroma from langchain_openai import OpenAIEmbeddings # 创建向量数据库 vectorstore = Chroma.from_documents( documents=docs, embedding=OpenAIEmbeddings(), persist_directory="./chroma_db" ) # 加载现有数据库 vectorstore = Chroma( persist_directory="./chroma_db", embedding_function=OpenAIEmbeddings() ) # 增量添加文档 vectorstore.add_documents([new_doc1, new_doc2]) # 删除文档 vectorstore.delete(ids=["doc1", "doc2"]) ### Pinecone (云端、可扩展) python from langchain_pinecone import PineconeVectorStore import pinecone # 初始化 Pinecone pinecone.init(api_key="your-api-key", environment="us-west1-gcp") # 创建索引（仅需一次） pinecone.create_index("my-index", dimension=1536, metric="cosine") # 创建向量数据库 vectorstore = PineconeVectorStore.from_documents( documents=docs, embedding=OpenAIEmbeddings(), index_name="my-index" ) # 使用元数据过滤进行查询 results = vectorstore.similarity_search( "Python tutorials", k=4, filter={"category": "beginner"} ) ### FAISS (快速相似度搜索) python from langchain_community.vectorstores import FAISS # 创建 FAISS 索引 vectorstore = FAISS.from_documents(docs, OpenAIEmbeddings()) # 保存到磁盘 vectorstore.save_local("./faiss_index") # 从磁盘加载 vectorstore = FAISS.load_local( "./faiss_index", OpenAIEmbeddings(), allow_dangerous_deserialization=True ) # 合并多个索引 vectorstore1 = FAISS.load_local("./index1", embeddings) vectorstore2 = FAISS.load_local("./index2", embeddings) vectorstore1.merge_from(vectorstore2) ### Weaviate (生产级、原生机器学习) python from langchain_weaviate import WeaviateVectorStore import weaviate # 连接 Weaviate client = weaviate.Client("http://baguaai.com") # 创建向量数据库 vectorstore = WeaviateVectorStore.from_documents( documents=docs, embedding=OpenAIEmbeddings(), client=client, index_name="LangChain" ) # 混合搜索 (向量 + 关键词) results = vectorstore.similarity_search( "Python async", k=4, alpha=0.5 # 0=关键词, 1=向量, 0.5=混合 ) ### Qdrant (快速、开源) python from langchain_qdrant import QdrantVectorStore from qdrant_client import QdrantClient # 连接 Qdrant client = QdrantClient(host="localhost", port=6333) # 创建向量数据库 vectorstore = QdrantVectorStore.from_documents( documents=docs, embedding=OpenAIEmbeddings(), collection_name="my_documents", client=client ) ## LangSmith 可观测性 ### 启用追踪 (Tracing) python import os # 设置环境变量 os.environ["LANGCHAIN_TRACING_V2"] = "true" os.environ["LANGCHAIN_API_KEY"] = "your-langsmith-api-key" os.environ["LANGCHAIN_PROJECT"] = "my-project" # 所有链调用将自动被追踪