Document variable name context was not found in llm_chain input variables

[sergiogcb]

I'm trying to add a specific prompt template to my QA Chain (RetrievalQA) so I can specify how the model will behave the answer. This is my current code:

vectorStore = Pinecone.from_existing_index(index_name, embedding_function)

query = "How can I avoid Jira issues from being created if the issue type is different than Story?"
docs = vectorStore.similarity_search(query)

llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0, openai_api_key="apiKey", max_tokens=300)

template = "ou have to try to provide solutions to customer questions with the information you've been trained. Keep the answer as concise as possible. Always say 'thanks for asking' at the end of the answer. Question: {question}"

QA_CHAIN_PROMPT = PromptTemplate.from_template(template);

qa_chain = RetrievalQA.from_chain_type(llm,retriever=vectorStore.as_retriever, chain_type_kwargs={"prompt": QA_CHAIN_PROMPT})
result = qa_chain({"question": query})

print(result)

This is the error I'm currently obtaining:

Traceback (most recent call last):
File "c:\Users\Sergio García\Desktop\AutomatedServiceDesk\code\PineConeVectorStore_Q.py", line 41, in
qa_chain = RetrievalQA.from_chain_type(llm,retriever=vectorStore.as_retriever, chain_type_kwargs={"prompt": QA_CHAIN_PROMPT})
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Sergio García\AppData\Local\Programs\Python\Python311\Lib\site-packages\langchain\chains\retrieval_qa\base.py", line 100, in from_chain_type
combine_documents_chain = load_qa_chain(
^^^^^^^^^^^^^^
File "C:\Users\Sergio García\AppData\Local\Programs\Python\Python311\Lib\site-packages\langchain\chains\question_answering_init_.py", line 249, in load_qa_chain
return loader_mapping[chain_type](
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Sergio García\AppData\Local\Programs\Python\Python311\Lib\site-packages\langchain\chains\question_answering_init_.py", line 81, in _load_stuff_chain return StuffDocumentsChain(
^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Sergio García\AppData\Local\Programs\Python\Python311\Lib\site-packages\langchain\load\serializable.py", line 90, in init
super().init(**kwargs)
File "C:\Users\Sergio García\AppData\Local\Programs\Python\Python311\Lib\site-packages\pydantic\v1\main.py", line 341, in init
raise validation_error
pydantic.v1.error_wrappers.ValidationError: 1 validation error for StuffDocumentsChain
root
document_variable_name context was not found in llm_chain input_variables: ['question'] (type=value_error)

Can't I define the Prompt Template with the variables I specify? Do I need to specify a context or? What is currently failing?

[dosubot[bot]]

🤖

Hello,

The error message "document_variable_name context was not found in llm_chain input_variables" suggests that the variable name you're using in your prompt template doesn't match the variable name you're passing to the PromptTemplate constructor in the input_variables parameter.

In your case, you're using the variable {question} in your QA_CHAIN_PROMPT template, and you're correctly passing {"question": query} as input to your qa_chain. However, the error might be occurring because you're trying to use a variable in your QA Chain that isn't included in the input_variables of your prompt templates.

To resolve this issue, you should ensure that all the variables you're using in your QA Chain are included in the input_variables parameter when you're creating your prompt templates. If you're using a variable named document_variable_name in your QA Chain, you should include it in the input_variables like this:

QA_CHAIN_PROMPT = PromptTemplate(
    input_variables=["question", "document_variable_name"],
    template=template,
)

This way, the PromptTemplate will know to expect a variable named document_variable_name when it's generating prompts.

Also, make sure that the template_format is set to "f-string" as the LangChain framework currently only supports f-string format for prompt templates. If you're using a different format, you might encounter errors.

Here are some relevant sources from the LangChain repository that might help you understand this better:

• libs/langchain/langchain/evaluation/qa/generate_prompt.py
• libs/langchain/langchain/chains/qa_with_sources/map_reduce_prompt.py
• libs/langchain/langchain/chains/retrieval_qa/prompt.py
• libs/langchain/langchain/prompts/prompt.py

I hope this helps! Let me know if you have any other questions.

Sources

• 'ValueError: Missing some input keys: {'query'}' but i added it?
• ValidationError: 1 validation error for ConversationalRetrievalChain chain_type_kwargs extra fields not permitted (type=value_error.extra)
• load_qa_with_sources_chain with custom prompt
• libs/langchain/langchain/evaluation/qa/generate_prompt.py
• libs/langchain/langchain/chains/qa_with_sources/map_reduce_prompt.py
• libs/langchain/langchain/chains/retrieval_qa/prompt.py
• libs/langchain/langchain/prompts/prompt.py
• libs/langchain/langchain/chains/retrieval_qa/base.py

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.