Langchain中RetrievalQA.from_chain_type报错ValidationError求助

问题场景 #

运行以下LangChain代码构建RetrievalQA链时，触发了ValidationError：

from langchain_google_genai import GoogleGenerativeAIEmbeddings

google_generative_ai_Embeddings =  GoogleGenerativeAIEmbeddings(model="models/embedding-001", google_api_key=api_key)
from langchain_community.vectorstores import Chroma
vectordb = Chroma.from_documents(data,
                           embedding=google_generative_ai_Embeddings,
                           persist_directory='./chromadb')
retriever_google = vectordb.as_retriever(score_threshold = 0.7)

from langchain.prompts import PromptTemplate

prompt_template = """Given the following context and a question, generate an answer based on this context only.
In the answer try to provide as much text as possible from "response" section in the source document context without making much changes.
If the answer is not found in the context, kindly state "I don't know." Don't try to make up an answer.

CONTEXT: {context}

QUESTION: {question}"""


PROMPT = PromptTemplate(
    template=prompt_template, input_variables=["context", "question"]
)
chain_type_kwargs = {"prompt": PROMPT}


from langchain.chains import RetrievalQA

chain_type = "stuff"

chain = RetrievalQA.from_chain_type(llm=llm,
                            chain_type=chain_type,
                            retriever=retriever_google,
                            input_key="query",
                            return_source_documents=True,
                            chain_type_kwargs=chain_type_kwargs)

错误信息如下：

ValidationError                           Traceback (most recent call last)
~\AppData\Local\Temp\ipykernel_4328\826232400.py in <module>
     20 chain_type = "stuff"
     21 
---> 22 chain = RetrievalQA.from_chain_type(llm=llm,
     23                             chain_type=chain_type,
     24                             retriever=retriever_google,

~\AppData\Roaming\Python\Python39\site-packages\langchain\chains\retrieval_qa\base.py in from_chain_type(cls, llm, chain_type, chain_type_kwargs, **kwargs)
     98         """Load chain from chain type."""
     99         _chain_type_kwargs = chain_type_kwargs or {}
--> 100         combine_documents_chain = load_qa_chain(
    101             llm, chain_type=chain_type, **_chain_type_kwargs
    102         )

~\AppData\Roaming\Python\Python39\site-packages\langchain\chains\question_answering\__init__.py in load_qa_chain(llm, chain_type, verbose, callback_manager, **kwargs)
    247             f"Should be one of {loader_mapping.keys()}"
    248         )
--> 249     return loader_mapping[chain_type](
    250         llm, verbose=verbose, callback_manager=callback_manager, **kwargs
    251     )

~\AppData\Roaming\Python\Python39\site-packages\langchain\chains\question_answering\__init__.py in _load_stuff_chain(llm, prompt, document_variable_name, verbose, callback_manager, callbacks, **kwargs)
     71 ) -> StuffDocumentsChain:
     72     _prompt = prompt or stuff_prompt.PROMPT_SELECTOR.get_prompt(llm)
--> 73     llm_chain = LLMChain(
     74         llm=llm,
     75         prompt=_prompt,

~\AppData\Roaming\Python\Python39\site-packages\langchain\load\serializable.py in __init__(self, **kwargs)
     73 
     74     def __init__(self, **kwargs: Any) -> None:
--> 75         super().__init__(**kwargs)
     76         self._lc_kwargs = kwargs
     77 

~\AppData\Roaming\Python\Python39\site-packages\pydantic\v1\main.py in __init__(__pydantic_self__, **data)
    339         values, fields_set, validation_error = validate_model(__pydantic_self__.__class__, data)
    340         if validation_error:
--> 341             raise validation_error
    342         try:
    343             object_setattr(__pydantic_self__, '__dict__', values)

ValidationError: 1 validation error for LLMChain
llm
  Can't instantiate abstract class BaseLanguageModel with abstract methods agenerate_prompt, apredict, apredict_messages, generate_prompt, invoke, predict, predict_messages (type=type_error)

错误原因 #

代码中的llm变量未被正确实例化：

• 仅引用了llm但未定义，或错误使用LangChain的抽象基类BaseLanguageModel而非具体模型实现
• RetrievalQA需要接收已实例化的大语言模型对象，而非抽象类或未定义变量

修复步骤 #

1. 导入对应LLM的实现类（这里搭配你使用的Google Embedding模型，选择Google Gemini）
2. 用你的Google API Key实例化LLM对象
3. 将实例化后的llm传入RetrievalQA

修复后的完整代码 #

from langchain_google_genai import GoogleGenerativeAIEmbeddings, GoogleGenerativeAI

# 实例化Embedding模型
google_generative_ai_Embeddings = GoogleGenerativeAIEmbeddings(model="models/embedding-001", google_api_key=api_key)

# 实例化LLM（关键补充步骤）
llm = GoogleGenerativeAI(model="gemini-pro", google_api_key=api_key, temperature=0.0)

from langchain_community.vectorstores import Chroma
vectordb = Chroma.from_documents(data,
                           embedding=google_generative_ai_Embeddings,
                           persist_directory='./chromadb')
retriever_google = vectordb.as_retriever(score_threshold = 0.7)

from langchain.prompts import PromptTemplate

prompt_template = """Given the following context and a question, generate an answer based on this context only.
In the answer try to provide as much text as possible from "response" section in the source document context without making much changes.
If the answer is not found in the context, kindly state "I don't know." Don't try to make up an answer.

CONTEXT: {context}

QUESTION: {question}"""


PROMPT = PromptTemplate(
    template=prompt_template, input_variables=["context", "question"]
)
chain_type_kwargs = {"prompt": PROMPT}


from langchain.chains import RetrievalQA

chain_type = "stuff"

chain = RetrievalQA.from_chain_type(llm=llm,
                            chain_type=chain_type,
                            retriever=retriever_google,
                            input_key="query",
                            return_source_documents=True,
                            chain_type_kwargs=chain_type_kwargs)

注意事项 #

• 确保api_key变量已正确设置为你的Google API密钥
• 根据需求调整LLM参数，比如temperature控制生成内容的随机性
• 如果使用其他LLM（如OpenAI、Anthropic等），替换对应的导入和实例化代码即可

内容的提问来源于stack exchange，提问作者Monidipa Das