Openbase

Using the API

SDKs

There is no Openbase SDK to install. Anything that speaks the OpenAI Chat Completions API works once you change the base URL.

Python

pip install openai
from openai import OpenAI

client = OpenAI(
    base_url="https://api.openbase.ai/v1",
    api_key=os.environ["OPENBASE_API_KEY"],
)

Node and TypeScript

npm install openai
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.openbase.ai/v1",
  apiKey: process.env.OPENBASE_API_KEY,
});

Vercel AI SDK

Use the OpenAI-compatible provider factory and point it at the gateway. Streaming, tool calls and structured output all behave as they would against OpenAI directly.

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { streamText } from "ai";

const openbase = createOpenAICompatible({
  name: "openbase",
  baseURL: "https://api.openbase.ai/v1",
  apiKey: process.env.OPENBASE_API_KEY,
});

const result = streamText({
  model: openbase("anthropic/claude-sonnet-5"),
  prompt: "Explain failover in one paragraph.",
});

LangChain

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="deepseek/deepseek-v4-pro",
    base_url="https://api.openbase.ai/v1",
    api_key=os.environ["OPENBASE_API_KEY"],
)

LlamaIndex

from llama_index.llms.openai_like import OpenAILike

llm = OpenAILike(
    model="google/gemini-3.1-pro",
    api_base="https://api.openbase.ai/v1",
    api_key=os.environ["OPENBASE_API_KEY"],
    is_chat_model=True,
)

Model names are not validated locally

Some frameworks keep an allowlist of known OpenAI model names and reject anything else before a request is sent. Where that happens, use the library's generic or "openai-like" class, as shown above, so the slug reaches the gateway untouched.

Anything else

If a tool lets you set a base URL and an API key, it will work. Set them to https://api.openbase.ai/v1 and your Openbase key.