使用LangChain与OpenAI集成现有ChromaDB集合的兼容性指南

from langchain_community.vectorstores import Chroma
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain.chains import RetrievalQA
from chromadb.utils import embedding_functions as ef
import chromadb

# 假设persistent_client和collection已创建并包含数据
# persistent_client = chromadb.PersistentClient(path="/path/to/your/chroma/db")
# collection = persistent_client.get_or_create_collection(name="my_collection")

# 初始化OpenAI嵌入函数
# 注意：旧版本的LangChain可能直接使用OpenAIEmbeddingFunction，
# 而新版本通常通过langchain_openai.OpenAIEmbeddings进行封装
openai_ef = OpenAIEmbeddings(
    model="text-embedding-ada-002", # 或其他合适的嵌入模型
    openai_api_key="YOUR_OPENAI_API_KEY",
    # 其他OpenAI配置，如azure_endpoint, api_version等
)

# 使用现有ChromaDB集合初始化LangChain的Chroma向量存储
langchain_chroma = Chroma(
    client=persistent_client,
    collection_name=collection.name,
    embedding_function=openai_ef, # 错误可能源于此处的embedding_function对象
)

# 初始化LLM模型
llm_model = "gpt-3.5-turbo" # 或其他Azure OpenAI模型名称
llm = ChatOpenAI(
    api_key="YOUR_OPENAI_API_KEY",
    api_version="YOUR_API_VERSION", # 如 "2023-05-15"
    azure_endpoint="YOUR_AZURE_ENDPOINT",
    model=llm_model
)

# 创建检索问答链
qa_chain = RetrievalQA.from_chain_type(
    llm,
    retriever=langchain_chroma.as_retriever(),
    chain_type="stuff" # 或 "refine", "map_reduce" 等
)

# 运行查询
try:
    response = qa_chain.run("How many datascientist do I need for a Object detection problem")
    print(response)
except AttributeError as e:
    print(f"An AttributeError occurred: {e}")
    print("This likely indicates a version incompatibility issue with your libraries.")

python --version
pip show langchain openai chromadb

# 创建并激活虚拟环境 (如果尚未完成)
python -m venv venv_rag
source venv_rag/bin/activate # macOS/Linux
# venv_rag\Scripts\activate # Windows

# 卸载现有版本 (可选，如果遇到冲突)
pip uninstall -y langchain openai chromadb

# 安装指定版本
pip install langchain==0.0.339 openai==1.3.4 chromadb==0.4.17
# 注意：如果您的LangChain版本高于0.1.0，OpenAI相关的组件已移至langchain-openai包
# 例如：pip install langchain==0.1.0 langchain-openai==0.0.1 chromadb==0.4.17

# 假设您已安装：
# pip install langchain langchain-openai langchain-community chromadb

from langchain_community.vectorstores import Chroma
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain.chains import RetrievalQA
import chromadb

# 初始化OpenAI嵌入函数 (在新版本中，OpenAIEmbeddings自带embed_query方法)
openai_embeddings = OpenAIEmbeddings(
    model="text-embedding-ada-002",
    openai_api_key="YOUR_OPENAI_API_KEY",
    # ... 其他OpenAI配置
)

# 使用现有ChromaDB集合初始化LangChain的Chroma向量存储
# Chroma的embedding_function参数现在通常直接接收OpenAIEmbeddings实例
langchain_chroma = Chroma(
    client=persistent_client,
    collection_name=collection.name,
    embedding_function=openai_embeddings,
)

# 初始化LLM模型 (同样来自langchain_openai)
llm = ChatOpenAI(
    api_key="YOUR_OPENAI_API_KEY",
    api_version="YOUR_API_VERSION",
    azure_endpoint="YOUR_AZURE_ENDPOINT",
    model="gpt-3.5-turbo"
)

# 创建检索问答链
qa_chain = RetrievalQA.from_chain_type(
    llm,
    retriever=langchain_chroma.as_retriever(),
    chain_type="stuff"
)

# 运行查询
response = qa_chain.run("How many datascientist do I need for a Object detection problem")
print(response)