How to use ClearRun with LangChain

Run your chain or model as usual, take the final answer text, and pass it with the prompt to clearrun.evaluate(). ClearRun works on the text, so any LangChain model works.

Last updated

Steps

  1. Install

    npm install @langchain/openai @clearrun/sdk

  2. Score the final answer

    import { ChatOpenAI } from '@langchain/openai';
import { createKernelClient } from '@clearrun/sdk';
const clearrun = createKernelClient(); // reads CLEARRUN_API_KEY

const model = new ChatOpenAI({ model: process.env.OPENAI_MODEL });
const prompt = 'When was the Eiffel Tower finished?';
const message = await model.invoke(prompt);
const answer = typeof message.content === 'string' ? message.content : JSON.stringify(message.content);

const { verdict } = await clearrun.evaluate({ prompt, responseText: answer });

Where to score

Score the answer a person or system will act on. Skip middle steps. Each call uses credits. See Where should ClearRun go in an AI pipeline?