langchain-openzoo

LangChain integration for OpenZoo — an OpenAI-compatible inference endpoint with no signup: any API key value works and billing happens per call via x402 or card.

Installation

pip install -U langchain-openzoo

No API key is required. Optionally:

export OPENZOO_API_KEY=sk-openzoo   # any value works; useful for tagging calls
export OPENZOO_BASE_URL=http://localhost:8402/v1   # route through a local `npx openzoo` gateway

Chat model

ChatOpenZoo is a thin subclass of ChatOpenAI pointed at https://api.openzoo.fun/v1. Everything ChatOpenAI supports (streaming, tool calling, structured output, async) works the same way.

from langchain_openzoo import ChatOpenZoo

llm = ChatOpenZoo(model="z-ai/glm-5.3-flash")
llm.invoke("Sing about zebras")   # note: billed per call

With parameters:

llm = ChatOpenZoo(
    model="z-ai/glm-5.3-flash",
    temperature=0.2,
    max_tokens=512,
    # api_key="sk-anything",                 # optional
    # base_url="http://localhost:8402/v1",   # optional local gateway
)

Browse the free model catalog at https://api.openzoo.fun/v1/models — model ids look like z-ai/glm-5.3-flash.

Development

python -m venv venv && source venv/bin/activate
pip install -e . pytest
pytest

The tests only cover instantiation and configuration — no network calls are made.