https://github.com/langchain-ai/retrieval-agent-template?utm_source=chatgpt.com 如何部署

userhttps://github.com/langchain-ai/retrieval-agent-template?utm_sourcechatgpt.com这个如何部署ChatGPT这个项目是LangGraph RAG 检索 Agent 模板部署前要先准备一个向量库默认是Elasticsearch也支持 MongoDB Atlas、Pinecone。仓库里已经有langgraph.json包含两个图indexer和retrieval_graph。本地跑起来git clone https://github.com/langchain-ai/retrieval-agent-template.gitcd retrieval-agent-templatecp .env.example .env安装pip install -U langgraph-cli[inmem]pip install -e .如果用 OpenAI.env至少填OPENAI_API_KEY你的keyLANGSMITH_PROJECTretrieval-agent默认检索后端是 Elasticsearch。官方 README 说默认retriever_provider: elastic。本地 Elasticsearchdocker network create elastic-netdocker run -p 127.0.0.1:9200:9200 -d \--name elasticsearch \--network elastic-net \-e ELASTIC_PASSWORDchangeme \-e discovery.typesingle-node \-e xpack.security.http.ssl.enabledfalse \-e xpack.license.self_generated.typetrial \docker.elastic.co/elasticsearch/elasticsearch:8.15.1.env加ELASTICSEARCH_URLhttp://host.docker.internal:9200ELASTICSEARCH_USERelasticELASTICSEARCH_PASSWORDchangeme然后启动 LangGraphlanggraph dev打开 Studio 后先选indexer图传入文档[{ page_content: My cat knows python. }]再切到retrieval_graph图提问。README 说明这个模板就是先用indexer建索引再用retrieval_graph按user_id检索回答。部署到 LangGraph Cloud / LangSmith把项目 push 到你自己的 GitHub 仓库。LangSmith → Deployments → New Deployment。选择这个仓库。配置环境变量OPENAI_API_KEY...ELASTICSEARCH_URL...ELASTICSEARCH_API_KEY...# 或者 ELASTICSEARCH_USER / ELASTICSEARCH_PASSWORD部署后调用图名from langgraph_sdk import get_sync_clientclient get_sync_client(url你的部署URL,api_key你的LangSmith API Key)client.runs.stream(None,retrieval_graph,input{messages: [{role: human, content: 我的文档里有什么}]},stream_modeupdates,)关键点不要只部署聊天图。这个模板有两个图indexer: ./src/retrieval_graph/index_graph.py:graph,retrieval_graph: ./src/retrieval_graph/graph.py:graph你必须先调用indexer把资料写入向量库再调用retrieval_graph问答。found something on the website that might be helpful to you. Here is the link and the content: