Fixing the TypeError promptValue.toChatMessages is not a function in LangChain.js

DanielJuly 26, 2024July 26, 2024

Yesterday, I made a trivial LangChain mistake that led to more than an hour wasted in debugging.

While googling this error, I did not find any useful articles, so hope this post will save someone else's time.

I was working on a code like the one below (mixed with some other stuff):

import { ChatOpenAI } from "@langchain/openai";
import { PromptTemplate } from "@langchain/core/prompts";
import { StringOutputParser } from "@langchain/core/output_parsers";
import * as dotenv from "dotenv";

dotenv.config();

const prompt = PromptTemplate.fromTemplate(
`Answer the question below:
Question: {question}`
);

const model = new ChatOpenAI({ temperature: 0 });

const chain = model
                .pipe(model)
                .pipe(new StringOutputParser());

const answer = await chain.invoke({
  question: 'What is the speed of light?'
});

console.log(answer);

When running the example, I was getting this error:

TypeError: promptValue.toChatMessages is not a function
const promptMessages = promptValues.map((promptValue) => promptValue.toChatMessages());

The reason for this error was how I was building the LCEL chain.

// ⛔️ TypeError: promptValue.toChatMessages is not a function
const chain = model
                .pipe(prompt)
                .pipe(new StringOutputParser());

const answer = await chain.invoke({
  question: 'What is the speed of light?'
});

It turns out that I was passing the output of the model to the prompt 🤦‍♂️.

It needs to be the other way around. The output of the prompt needs to be passed in as input for the model.

// 👍 this will work
const chain = prompt
                .pipe(model)
                .pipe(new StringOutputParser());

const answer = await chain.invoke({
  question: 'What is the speed of light?'
});

There you have it! This was the solution in my case. Hope it helps!