mirage-langchain

The Mirage LangChain adapter. Use Mirage AI models with LangChain.

Copyright 2025 Crisp IM SAS. See LICENSE for copying information.

• 📝 Implements: API Reference (V1) at revision: 07/04/2026
• 😘 Maintainer: @baptistejamin

Usage

Install the library:

npm install mirage-langchain --save

Then, import it:

import { ChatMirage } from "mirage-langchain";

Construct a new authenticated ChatMirage client with your user_id and secret_key tokens:

const model = new ChatMirage({
  userId: "ui_xxxxxx",
  secretKey: "sk_xxxxxx",
  model: "medium"
});

Then, use it with LangChain:

import { HumanMessage } from "@langchain/core/messages";

const response = await model.invoke([
  new HumanMessage("What is the capital of France?")
]);

console.log(response.content);

Authentication

To authenticate against the API, get your tokens (user_id and secret_key).

Then, pass those tokens when you instantiate the ChatMirage client:

const model = new ChatMirage({
  userId: "user_id",
  secretKey: "secret_key"
});

Configuration Options

The ChatMirage constructor accepts the following options:

| Option | Type | Description | |--------|------|-------------| | userId | string | Required. Your Mirage API user ID | | secretKey | string | Required. Your Mirage API secret key | | model | "small" \| "medium" \| "large" | Model size to use (default: "medium") | | temperature | number | Sampling temperature | | maxTokens | number | Maximum tokens to generate | | logprobs | boolean | Whether to return log probabilities | | topLogprobs | number | Number of top log probabilities to return |

Features

Streaming

ChatMirage supports streaming responses:

const stream = await model.stream([
  new HumanMessage("Tell me a story")
]);

for await (const chunk of stream) {
  process.stdout.write(chunk.content);
}

Tool Calling

ChatMirage supports tool calling:

import { tool } from "@langchain/core/tools";
import { z } from "zod";

const weatherTool = tool(
  async ({ location }) => {
    return `The weather in ${location} is sunny.`;
  },
  {
    name: "get_weather",
    description: "Get the weather for a location",
    schema: z.object({
      location: z.string().describe("The city name")
    })
  }
);

const modelWithTools = model.bindTools([weatherTool]);

const response = await modelWithTools.invoke([
  new HumanMessage("What's the weather in Paris?")
]);

Structured Output

ChatMirage supports structured output with Zod schemas:

import { z } from "zod";

const schema = z.object({
  name: z.string(),
  age: z.number()
});

const structuredModel = model.withStructuredOutput(schema);

const result = await structuredModel.invoke([
  new HumanMessage("John is 25 years old")
]);

console.log(result); // { name: "John", age: 25 }

Peer Dependencies

This package requires @langchain/core version 1.0.0 or higher as a peer dependency:

npm install @langchain/core