How to add more inputs to my Retrieval QA Prompt Template

[sergiogcb]

I'm trying to add more variables to my Retrieval QA Prompt Template, but there is no way to make it work

template =  """You are a human assist that is expert with our app.
    Always say "Best regards" at the end of the answer, and "Thanks for your request, at the beginning.
    {context}
    App: {app}
    At the end of your sentence return the used app"""

    template.format(app = myapp, context = "")

    QA_CHAIN_PROMPT = PromptTemplate.from_template(template)

    qa_chain = RetrievalQA.from_chain_type(llm,retriever=vectorStore.as_retriever(search_kwargs={'k':8}), chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}, return_source_documents=True)
    result = qa_chain({"query": myquery, "app": myapp})

    print(qa_chain)

What is exactly failing in my code? I can't make it work. Also, it forces me to give a value to the context, I dont know if theres any impact for having the context empty.

This is the error that appears:
ValueError: Missing some input keys: {'app'}

I also tried giving the value manually instead that from a variable but nothing worked

[dosubot[bot]]

🤖

Thanks for your request,

The error message you're seeing, ValueError: Missing some input keys: {'app'}, is indicating that the app variable is not being found when the PromptTemplate is being used. This is because the app variable is not included in the input_variables list when the PromptTemplate is instantiated.

To fix this, you need to modify both the template string and the input_variables list in the PromptTemplate instantiation. Here's how you can do it:

template =  """You are a human assist that is expert with our app.
    Always say "Best regards" at the end of the answer, and "Thanks for your request, at the beginning.
    {context}
    App: {app}
    At the end of your sentence return the used app"""

QA_CHAIN_PROMPT = PromptTemplate(
    input_variables=["app", "context"],
    template=template,
)

qa_chain = RetrievalQA.from_chain_type(llm,retriever=vectorStore.as_retriever(search_kwargs={'k':8}), chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}, return_source_documents=True)
result = qa_chain({"query": myquery, "app": myapp})

print(qa_chain)

In this code, {app} is a placeholder for the new variable in the template string, and "app" is added to the input_variables list. This tells the PromptTemplate that it should expect an additional input key named app when the template is used. If you don't provide a value for app when using the template, you will encounter a ValueError indicating missing input keys.

As for the context variable, it's okay to have it empty. However, you need to ensure that it's included in the input_variables list when the PromptTemplate is instantiated, otherwise, you will encounter a ValueError indicating missing input keys.

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

Best regards,

Sources

• 'ValueError: Missing some input keys: {'query'}' but i added it?
• libs/langchain/langchain/chains/retrieval_qa/prompt.py
• libs/langchain/langchain/evaluation/qa/generate_prompt.py
• libs/langchain/langchain/schema/prompt_template.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.

[sergiogcb]

@dosu-bot

The same error still appears

ValueError: Missing some input keys: {'app'}

My code now is

   template =  """You are a human assist that is expert with our app.
    {context}
    App: {app}
    At the end of your sentence return the used app"""


    QA_CHAIN_PROMPT = PromptTemplate(
    input_variables=["app", "context"],
    template=template,
)

    qa_chain = RetrievalQA.from_chain_type(llm,retriever=vectorStore.as_retriever(search_kwargs={'k':8}), chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}, return_source_documents=True)
    result = qa_chain({"query": myquery, "app": myapp})

[dosubot]

🤖

Hello,

The error you're encountering, ValueError: Missing some input keys: {'app'}, is raised when the format method of the PromptTemplate class is called with keyword arguments that do not contain all the keys specified in the input_variables list of the PromptTemplate instance.

In your case, you have defined input_variables=["app", "context"] in your PromptTemplate instantiation. This means that when you're calling the format method, you need to ensure that the keyword arguments you're passing contain both "app" and "context".

However, in your qa_chain call, you're only passing "query" and "app" as keys in the dictionary. The "context" key is missing, hence the error.

Here's how you can modify your code to resolve this issue:

# Ensure 'context' is included in the dictionary passed to qa_chain
result = qa_chain({"query": myquery, "app": myapp, "context": "your_context_here"})

In the above code, replace "your_context_here" with the actual context to be used. This will ensure that the "context" key is present in the dictionary, and the format method will be able to find it when formatting the document based on the prompt template.

For more information, you can refer to the following sources in the LangChain codebase:

• libs/langchain/langchain/schema/prompt_template.py
• libs/langchain/langchain/chains/qa_with_sources/stuff_prompt.py
• libs/langchain/langchain/prompts/prompt.py
• libs/langchain/langchain/chains/conversational_retrieval/prompts.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?
• libs/langchain/langchain/schema/prompt_template.py
• libs/langchain/langchain/chains/qa_with_sources/stuff_prompt.py
• libs/langchain/langchain/prompts/prompt.py
• libs/langchain/langchain/chains/conversational_retrieval/prompts.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.

[weissenbacherpwc]

Did you find a solution?

[eyurtsev]

The original code has a bug in it -- After formatting the value needs to be assigned to a variable

myapp = 'hello'
context = 'meow'

template =  """You are a human assist that is expert with our app.
Always say "Best regards" at the end of the answer, and "Thanks for your request, at the beginning.
{context}
App: {app}
At the end of your sentence return the used app"""

template.format(app=myapp, context=context)
print(template)

print("Correct")
template = template.format(app=myapp, context=context)
print(template)


QA_CHAIN_PROMPT = PromptTemplate.from_template(template)
QA_CHAIN_PROMPT.invoke({})

[weissenbacherpwc]

I am testing it and will let you know!

[weissenbacherpwc]

So got it working but figured out a weird behaviour. My prompt & code looks as follows and it works:

rag_prompt= """
<s> [INST] Im folgenden bekommst du eine Aufgabe. Erledige diese anhand des User Inputs.
Falls im Kontext relevante Informationen enthalten sind, die deine Antwort verbessern, füge diese in deine Antwort ein. 
Wichitg ist, dass du keine zusätzlichen Informationen erfindest, sondern gegebenenfalls nur die im Kontext enthaltenen.
Die Antwort soll sich nicht wiederholen.

### Hier die Aufgabe: ###
{typescript_string}

### Hier Kontext zum User Input: ###
{{context}}

### Hier der User Input: ###
{{question}}

Antwort: [/INST]
"""

rag_prompt = rag_prompt.format(typescript_string="Schreibe die Stichpunkte in einen Volltext um.")
def set_rag_prompt():
    return PromptTemplate(template=rag_prompt, input_variables=['typescript_string', 'context', 'question'])

from langchain_community.embeddings import HuggingFaceEmbeddings

embeddings = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large",
                                       model_kwargs={'device': 'mps'}, encode_kwargs={'device': 'mps', 'batch_size': 32})

vectorstore = FAISS.load_local("vectorstore/db_faiss_bkb", embeddings)
retriever = vectorstore.as_retriever(search_kwargs={'k': 1, 'score_treshold': 0.75}, search_type="similarity")
llm = build_llm("modelle/mixtral-8x7b-instruct-v0.1.Q5_K_M.gguf")
qa_prompt = set_rag_prompt()
def build_retrieval_qa(llm, prompt, vectordb):
    chain_type_kwargs={ "prompt": prompt,
                        "verbose": False
                        }
    dbqa = RetrievalQA.from_chain_type(llm=llm,
                                        chain_type='stuff',
                                        retriever=vectordb,
                                        return_source_documents=True,
                                        chain_type_kwargs=chain_type_kwargs,
                                        verbose=True
                                        )
    return dbqa


dbqa = build_retrieval_qa(llm=llm,prompt=qa_prompt,vectordb=retriever)
dbqa({"query": "Test Query"})

The weird behaviour is happening in naming the Prompt input variables. It is only working if the input_variable pair is "context" and "question". Before (when it did not work), "{question}" was named "{query}" in the Prompt and when calling the RetrievalQA chain with dbqa({"query": "Test Query"}) it gave me a "Key Error {query}". After renaming it to "question" the code runs through.

[eyurtsev]

OK Great! That makes sense, sorry didn't catch it earlier. If you think there's any documentation that should be improved in langchain to help a future you or something else, feel free to open a PR to expand documentation

[NickWyq]

How to change different typescript_string when inference? (The input key only contains 'query'/'question')

[weissenbacherpwc]

In the code you can use any name instead of typesript string. With rag_prompt = rag_prompt.format(typescript_string="Schreibe die Stichpunkte in einen Volltext um.") like @eyurtsev suggested, you need to first fill it with a value. If then only "context" and "question" are left, you can call the chain with dbqa({"query": "Test Query"}). This is the bug here I think but it works. So in this case "context" gets filled with the results from the retriever and "question" with the "query" value, when calling the chain

[anshumankmr]

Is there a fix available for this from Langchain's codebase?