Throws Error: Missing value for input variable question on subsequent chatting

[ShisuiMadara]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

import { ChatOpenAI, OpenAI } from "@langchain/openai";
import { PineconeStore } from "@langchain/pinecone";
import { ConversationChain, loadQAChain } from 'langchain/chains';
import { PromptTemplate } from "@langchain/core/prompts";

import {
  ChatPromptTemplate,
  MessagesPlaceholder,
} from "@langchain/core/prompts";
import { createStuffDocumentsChain } from "langchain/chains/combine_documents";
import { createHistoryAwareRetriever } from "langchain/chains/history_aware_retriever";
import { createRetrievalChain } from "langchain/chains/retrieval";

const contextualizeQSystemPrompt = `Given a chat history and the latest user question
which might reference context in the chat history, formulate a standalone question
which can be understood without the chat history. Do NOT answer the question,
just reformulate it if needed and otherwise return it as is.`;


  
  const condenseQuestionTemplate = `Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.

  Chat History:
  {chat_history}
  Follow Up Input: {question}
  Standalone question:`;
  const CONDENSE_QUESTION_PROMPT = PromptTemplate.fromTemplate(
    condenseQuestionTemplate
  );
  
  const answerTemplate = `Answer the question based only on the following context:
  {context}
  
  Question: {question}
  `;
  const ANSWER_PROMPT = PromptTemplate.fromTemplate(answerTemplate);

const makeChain = async (vectorstore: PineconeStore) => {
  const llm = new ChatOpenAI()

  console.log("wttttffffffffffffffffffffffffffffffffffffffffffffff")
  console.log("{input}")

  const contextualizeQPrompt = ChatPromptTemplate.fromMessages([
    ["system", condenseQuestionTemplate],
    new MessagesPlaceholder("chat_history"),
    ["human", "{input}"],
    
  ]);

  console.log(contextualizeQPrompt)

  const historyAwareRetriever = await createHistoryAwareRetriever({
    llm,
    retriever: vectorstore.asRetriever(),
    rephrasePrompt: contextualizeQPrompt,
  });

  // Answer question
  const qaSystemPrompt = `
  You are an assistant for question-answering tasks. Use
  the following pieces of retrieved context to answer the
  question. If you don't know the answer, just say that you
  don't know.
  \n\n
  {context}`;
  
  const qaPrompt = ChatPromptTemplate.fromMessages([
    ["system", qaSystemPrompt],
    new MessagesPlaceholder("chat_history"),
    ["human", "{input}"],
  ]);



  const docChain = await createStuffDocumentsChain({
    llm,
    prompt: qaPrompt,
  });

  const ragChain = await createRetrievalChain({
    retriever: historyAwareRetriever,
    combineDocsChain: docChain,
  });

  return ragChain
 
};

export default makeChain;

Description

I am trying to make a conversational bot. Which answers questions based on the given document. For this I created a stuffDocumentChain and then a retrieval change. I can run the code for one input from the UI. but for the future inputs, it throws the error. Although it receives the question from the frontend.

System Info

"@langchain/openai": "^0.1.3",
"@langchain/pinecone": "^0.0.7",
"langchain": "0.2.5",

[dosubot[bot]]

Hey there, @ShisuiMadara! I'm here to help you out. If you need assistance with bugs, questions, or want to contribute, I've got your back. Let's tackle this issue together!

To resolve the error "Missing value for input variable question" for subsequent inputs in your conversational bot code, you need to ensure that all required input variables are provided when formatting the prompt. Here’s how you can adjust your code to include the question variable correctly:

1. Ensure question is included in the input object:
   When invoking the chain, make sure to provide the question and chat_history variables correctly.

2. Update the ChatPromptTemplate to use question instead of input:
   Modify the ChatPromptTemplate to use question instead of input to match the expected variable names.

Here’s an updated version of your makeChain function:

import { ChatOpenAI, OpenAI } from "@langchain/openai";
import { PineconeStore } from "@langchain/pinecone";
import { ConversationChain, loadQAChain } from 'langchain/chains';
import { PromptTemplate } from "@langchain/core/prompts";
import {
  ChatPromptTemplate,
  MessagesPlaceholder,
} from "@langchain/core/prompts";
import { createStuffDocumentsChain } from "langchain/chains/combine_documents";
import { createHistoryAwareRetriever } from "langchain/chains/history_aware_retriever";
import { createRetrievalChain } from "langchain/chains/retrieval";

const condenseQuestionTemplate = `Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.

Chat History:
{chat_history}
Follow Up Input: {question}
Standalone question:`;
const CONDENSE_QUESTION_PROMPT = PromptTemplate.fromTemplate(condenseQuestionTemplate);

const answerTemplate = `Answer the question based only on the following context:
{context}

Question: {question}`;
const ANSWER_PROMPT = PromptTemplate.fromTemplate(answerTemplate);

const makeChain = async (vectorstore: PineconeStore) => {
  const llm = new ChatOpenAI();

  const contextualizeQPrompt = ChatPromptTemplate.fromMessages([
    ["system", condenseQuestionTemplate],
    new MessagesPlaceholder("chat_history"),
    ["human", "{question}"],
  ]);

  const historyAwareRetriever = await createHistoryAwareRetriever({
    llm,
    retriever: vectorstore.asRetriever(),
    rephrasePrompt: contextualizeQPrompt,
  });

  const qaSystemPrompt = `
  You are an assistant for question-answering tasks. Use
  the following pieces of retrieved context to answer the
  question. If you don't know the answer, just say that you
  don't know.
  \n\n
  {context}`;

  const qaPrompt = ChatPromptTemplate.fromMessages([
    ["system", qaSystemPrompt],
    new MessagesPlaceholder("chat_history"),
    ["human", "{question}"],
  ]);

  const docChain = await createStuffDocumentsChain({
    llm,
    prompt: qaPrompt,
  });

  const ragChain = await createRetrievalChain({
    retriever: historyAwareRetriever,
    combineDocsChain: docChain,
  });

  return ragChain;
};

export default makeChain;

When invoking the chain, ensure you provide the question and chat_history:

const chain = await makeChain(vectorstore);
const res = await chain.invoke({
  question: "What is the meaning of life?",
  chat_history: "Previous conversation history here",
});
console.log(res);

This should resolve the error by ensuring that the question variable is correctly included in the input object for subsequent inputs ^[1][2].

To continue talking to Dosu, mention @dosu.

[ShisuiMadara]

This gives Error: Expected "input" to be a prompt variable, but got ["chat_history","question"].

[dosubot[bot]]

To resolve the error "Expected input to be a prompt variable, but got ["chat_history","question"]", you need to ensure that the ChatPromptTemplate is correctly configured to match the input variables you are passing. Specifically, the ChatPromptTemplate should use question and chat_history as input variables.

Here's how you can adjust your code:

1. Ensure the ChatPromptTemplate uses question and chat_history:
   Update the ChatPromptTemplate to correctly reference question and chat_history.

2. Ensure the input object matches the expected variables:
   When invoking the chain, make sure to provide both question and chat_history.

Here’s the updated version of your makeChain function:

import { ChatOpenAI } from "@langchain/openai";
import { PineconeStore } from "@langchain/pinecone";
import { PromptTemplate } from "@langchain/core/prompts";
import {
  ChatPromptTemplate,
  MessagesPlaceholder,
} from "@langchain/core/prompts";
import { createStuffDocumentsChain } from "langchain/chains/combine_documents";
import { createHistoryAwareRetriever } from "langchain/chains/history_aware_retriever";
import { createRetrievalChain } from "langchain/chains/retrieval";

const condenseQuestionTemplate = `Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.

Chat History:
{chat_history}
Follow Up Input: {question}
Standalone question:`;
const CONDENSE_QUESTION_PROMPT = PromptTemplate.fromTemplate(condenseQuestionTemplate);

const answerTemplate = `Answer the question based only on the following context:
{context}

Question: {question}`;
const ANSWER_PROMPT = PromptTemplate.fromTemplate(answerTemplate);

const makeChain = async (vectorstore: PineconeStore) => {
  const llm = new ChatOpenAI();

  const contextualizeQPrompt = ChatPromptTemplate.fromMessages([
    ["system", condenseQuestionTemplate],
    new MessagesPlaceholder("chat_history"),
    ["human", "{question}"],
  ]);

  const historyAwareRetriever = await createHistoryAwareRetriever({
    llm,
    retriever: vectorstore.asRetriever(),
    rephrasePrompt: contextualizeQPrompt,
  });

  const qaSystemPrompt = `
  You are an assistant for question-answering tasks. Use
  the following pieces of retrieved context to answer the
  question. If you don't know the answer, just say that you
  don't know.
  \n\n
  {context}`;

  const qaPrompt = ChatPromptTemplate.fromMessages([
    ["system", qaSystemPrompt],
    new MessagesPlaceholder("chat_history"),
    ["human", "{question}"],
  ]);

  const docChain = await createStuffDocumentsChain({
    llm,
    prompt: qaPrompt,
  });

  const ragChain = await createRetrievalChain({
    retriever: historyAwareRetriever,
    combineDocsChain: docChain,
  });

  return ragChain;
};

export default makeChain;

When invoking the chain, ensure you provide the question and chat_history:

const chain = await makeChain(vectorstore);
const res = await chain.invoke({
  question: "What is the meaning of life?",
  chat_history: "Previous conversation history here",
});
console.log(res);

This should resolve the error by ensuring that the question and chat_history variables are correctly included in the input object for subsequent inputs ^[1][2].

To continue talking to Dosu, mention @dosu.

[ShisuiMadara]

no, the error still persists. Is the usage of docChain and ragChain fine?

[dosubot[bot]]

The usage of docChain and ragChain seems fine, but the error "Expected input to be a prompt variable, but got ["chat_history","question"]" indicates that the ChatPromptTemplate is not correctly configured to match the input variables you are passing.

To resolve this, ensure that the ChatPromptTemplate and the input variables are correctly aligned. Here’s a revised version of your makeChain function with the necessary adjustments:

1. Ensure the ChatPromptTemplate uses question and chat_history:
   Update the ChatPromptTemplate to correctly reference question and chat_history.

2. Ensure the input object matches the expected variables:
   When invoking the chain, make sure to provide both question and chat_history.

Here’s the updated version of your makeChain function:

import { ChatOpenAI } from "@langchain/openai";
import { PineconeStore } from "@langchain/pinecone";
import { PromptTemplate } from "@langchain/core/prompts";
import {
  ChatPromptTemplate,
  MessagesPlaceholder,
} from "@langchain/core/prompts";
import { createStuffDocumentsChain } from "langchain/chains/combine_documents";
import { createHistoryAwareRetriever } from "langchain/chains/history_aware_retriever";
import { createRetrievalChain } from "langchain/chains/retrieval";

const condenseQuestionTemplate = `Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.

Chat History:
{chat_history}
Follow Up Input: {question}
Standalone question:`;
const CONDENSE_QUESTION_PROMPT = PromptTemplate.fromTemplate(condenseQuestionTemplate);

const answerTemplate = `Answer the question based only on the following context:
{context}

Question: {question}`;
const ANSWER_PROMPT = PromptTemplate.fromTemplate(answerTemplate);

const makeChain = async (vectorstore: PineconeStore) => {
  const llm = new ChatOpenAI();

  const contextualizeQPrompt = ChatPromptTemplate.fromMessages([
    ["system", condenseQuestionTemplate],
    new MessagesPlaceholder("chat_history"),
    ["human", "{question}"],
  ]);

  const historyAwareRetriever = await createHistoryAwareRetriever({
    llm,
    retriever: vectorstore.asRetriever(),
    rephrasePrompt: contextualizeQPrompt,
  });

  const qaSystemPrompt = `
  You are an assistant for question-answering tasks. Use
  the following pieces of retrieved context to answer the
  question. If you don't know the answer, just say that you
  don't know.
  \n\n
  {context}`;

  const qaPrompt = ChatPromptTemplate.fromMessages([
    ["system", qaSystemPrompt],
    new MessagesPlaceholder("chat_history"),
    ["human", "{question}"],
  ]);

  const docChain = await createStuffDocumentsChain({
    llm,
    prompt: qaPrompt,
  });

  const ragChain = await createRetrievalChain({
    retriever: historyAwareRetriever,
    combineDocsChain: docChain,
  });

  return ragChain;
};

export default makeChain;

When invoking the chain, ensure you provide the question and chat_history:

const chain = await makeChain(vectorstore);
const res = await chain.invoke({
  question: "What is the meaning of life?",
  chat_history: "Previous conversation history here",
});
console.log(res);

This should resolve the error by ensuring that the question and chat_history variables are correctly included in the input object for subsequent inputs ^[1].

To continue talking to Dosu, mention @dosu.

[DaX-523]

Hi @ShisuiMadara, may I know if the issue is resolved??

[ShisuiMadara]

no this was not resolved

[DaX-523]

i copied your code and made some changes, though there is no error but the chain is not running. If i ran the chain after wrapping it in the runnable sequence it forgets the context..