Appearance

LangChain & LlamaIndex ​

Both frameworks talk to LibertAI through their standard OpenAI classes, pointed at the LibertAI base URL. Your chains, agents, and indexes run on open-weights models, and moving to or from any other OpenAI-compatible provider is a constructor argument.

Prerequisites ​

  • A LibertAI API key — see Get an API key
  • pip install langchain-openai and/or pip install llama-index llama-index-llms-openai-like

LangChain ​

Use ChatOpenAI from langchain_openai with a custom base_url:

python
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://api.libertai.io/v1",
    api_key="YOUR_API_KEY",
    model="qwen3.5-122b-a10b",
)

print(llm.invoke("In one sentence: what is decentralized inference?").content)

Tool calling, streaming, and structured output work as with any OpenAI-compatible chat model, so bind_tools, with_structured_output, and LangGraph agents run unchanged.

LlamaIndex ​

Use OpenAILike, which is built for OpenAI-compatible endpoints serving non-OpenAI models:

python
from llama_index.llms.openai_like import OpenAILike

llm = OpenAILike(
    api_base="https://api.libertai.io/v1",
    api_key="YOUR_API_KEY",
    model="qwen3.5-122b-a10b",
    is_chat_model=True,
    is_function_calling_model=True,
    context_window=262144,
)

print(llm.complete("In one sentence: what is decentralized inference?"))

Pass the llm into Settings.llm or any query engine as usual. For RAG pipelines, LibertAI's Embeddings API is OpenAI-compatible too, so the standard OpenAI embedding classes work the same way with the same base URL.

Verify it works ​

Run either snippet above. A one-sentence answer means everything is wired up. On a 401, check the key directly:

sh
curl -i https://api.libertai.io/libertai/auth/check \
  -H "Authorization: Bearer YOUR_API_KEY"

Models ​

qwen3.5-122b-a10b is a strong general default (262k context, vision, tools). See the full model and pricing table, including the TEE option for confidential workloads.