Frameworks

AI Setu doesn't ship a dedicated plugin for any particular framework — because it's a drop-in OpenAI-compatible endpoint (https://gateway.aisetu.ai/v1), any framework that lets you point its OpenAI client at a custom base URL and key works without one. This page shows the standard override for a few common frameworks. These aren't officially maintained integrations, and the exact configuration option names are each framework's own — check their docs if something below is out of date.

The general pattern

Every framework below reduces to the same two settings: swap the base URL to https://gateway.aisetu.ai/v1, and pass your AI Setu key instead of a raw OpenAI key. Where the framework asks for a model name, use any of the three model forms AI Setu accepts.

LangChain / LangChain.js

ChatOpenAI (and the Python langchain-openai equivalent) accepts a base URL override:

import { ChatOpenAI } from '@langchain/openai';

const model = new ChatOpenAI({
  apiKey: process.env.AI_SETU_API_KEY,
  model: 'openai/gpt-4o-mini',
  configuration: { baseURL: 'https://gateway.aisetu.ai/v1' },
});

from langchain_openai import ChatOpenAI

model = ChatOpenAI(
    api_key=os.environ["AI_SETU_API_KEY"],
    model="openai/gpt-4o-mini",
    base_url="https://gateway.aisetu.ai/v1",
)

Vercel AI SDK

Use the OpenAI-compatible provider factory and point it at the gateway:

import { createOpenAICompatible } from '@ai-sdk/openai-compatible';
import { generateText } from 'ai';

const aiSetu = createOpenAICompatible({
  name: 'ai-setu',
  apiKey: process.env.AI_SETU_API_KEY,
  baseURL: 'https://gateway.aisetu.ai/v1',
});

const { text } = await generateText({
  model: aiSetu('openai/gpt-4o-mini'),
  prompt: 'Hello!',
});

LlamaIndex

The OpenAI LLM class takes the same override:

import { OpenAI } from 'llamaindex';

const llm = new OpenAI({
  apiKey: process.env.AI_SETU_API_KEY,
  model: 'openai/gpt-4o-mini',
  additionalSessionOptions: { baseURL: 'https://gateway.aisetu.ai/v1' },
});

Things to check when adapting a framework

• Streaming — AI Setu's stream: true responses are OpenAI-shaped SSE; frameworks that consume the OpenAI SDK's stream iterator work unchanged.
• Tool calling — forwarded verbatim to whichever provider serves the request; not every upstream model supports every tool-calling shape a framework might send. Bedrock text-only models, for example, reject tool calls with a clear error.
• Embeddings — only the OpenAI and Gemini providers implement /v1/embeddings today (see Models); a framework's embedding step needs a model routed to one of those.
• Headers — AI Setu's routing, caching, and BYOK headers (documented throughout these Concepts pages) are plain HTTP headers. If your framework exposes a way to attach custom headers to the underlying OpenAI client, you can still use them alongside a framework integration.