LangChain.js integration for EmpirioLabs AI: ChatEmpirioLabs chat models and EmpirioLabsEmbeddings.
- Chat with any chat model in the EmpirioLabs AI catalog, with streaming, tool calling, structured output, and image input on the models that support them.
- A reasoning model's thinking comes back in
additional_kwargs.reasoning_contentand as areasoningblock incontentBlocks. - Works with
createAgentfromlangchainand with LangGraph. - Embeddings for retrieval and vector stores.
npm install @empiriolabs/langchain @langchain/coreCreate an API key at platform.empiriolabs.ai and set it:
export EMPIRIOLABS_API_KEY="sk-empiriolabs-your_key_here"import { ChatEmpirioLabs } from "@empiriolabs/langchain";
const llm = new ChatEmpirioLabs({ model: "qwen3-8-flash" });
const message = await llm.invoke("Explain backpropagation in one paragraph.");
console.log(message.content);Stream the answer:
for await (const chunk of await llm.stream("Write a haiku about the sea.")) {
process.stdout.write(String(chunk.content));
}Read a reasoning model's thinking, or turn it off for a faster answer:
const answer = await llm.invoke("What is 17 times 23?");
console.log(answer.contentBlocks); // [{ type: "reasoning", ... }, { type: "text", text: "391" }]
const quick = new ChatEmpirioLabs({ model: "qwen3-8-flash", modelKwargs: { reasoning_effort: "none" } });import { tool } from "@langchain/core/tools";
import { createAgent } from "langchain";
import * as z from "zod";
const getWeather = tool(async ({ city }) => `It is 21 C and sunny in ${city}.`, {
name: "get_weather",
description: "Get the current weather for a city.",
schema: z.object({ city: z.string() }),
});
const agent = createAgent({ model: llm, tools: [getWeather] });
const result = await agent.invoke({ messages: [{ role: "user", content: "What is the weather in Tokyo?" }] });
console.log(result.messages.at(-1)?.content);llm.bindTools([getWeather]) binds tools without an agent loop.
const Movie = z.object({ title: z.string(), year: z.number() });
const movie = await llm.withStructuredOutput(Movie).invoke("Name one movie directed by Denis Villeneuve.");
// { title: "Dune", year: 2021 }Pass { method: "functionCalling" } or { method: "jsonSchema" } to choose how the schema is sent.
import { HumanMessage } from "@langchain/core/messages";
const reply = await llm.invoke([
new HumanMessage({
content: [
{ type: "text", text: "Describe this image." },
{ type: "image_url", image_url: { url: "https://example.com/photo.jpg" } },
],
}),
]);import { EmpirioLabsEmbeddings } from "@empiriolabs/langchain";
const embeddings = new EmpirioLabsEmbeddings({ model: "text-embedding-v4" });
const vector = await embeddings.embedQuery("What is EmpirioLabs AI?");
const vectors = await embeddings.embedDocuments(["First document", "Second document"]);A request takes up to 2,048 texts; embedDocuments sends 512 per request by default (batchSize). Set dimensions to choose the vector size on models that support it.
| Option | Default | |
|---|---|---|
model |
required | A model ID from Models & Pricing |
apiKey |
EMPIRIOLABS_API_KEY |
A key string, or a function that returns one |
configuration.baseURL |
EMPIRIOLABS_API_BASE, then https://api.empiriolabs.ai/v1 |
|
temperature, maxTokens, topP, stop, timeout, maxRetries |
the model's defaults | The usual LangChain chat model options |
modelKwargs |
Any other request parameter the model supports, for example reasoning_effort |
The OpenAI environment variables (OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_ORG_ID, OPENAI_PROJECT_ID) are ignored.
MIT