langchain-topxai

LangChain chat model for TopxAI, an API gateway that serves Claude, GPT, Grok, GLM, Kimi and DeepSeek models through one OpenAI-compatible endpoint at fixed USD prices per route.

Installation

pip install -U langchain-topxai

Create a key at https://ai.topxea.com/keys and set it:

export TOPXAI_API_KEY="your-api-key"

Chat models

from langchain_topxai import ChatTopxAI

model = ChatTopxAI(model="kimi-k3", temperature=0)
model.invoke("Sing a ballad of LangChain.")

model takes any text model id from https://ai.topxea.com/pricing, for example claude-sonnet-5-5, gpt-6.1-sol, grok-4.7, kimi-k3, deepseek-v4-pro or deepseek-flash. ChatTopxAI builds on langchain-openai's BaseChatOpenAI and always calls the Chat Completions API: streaming, async calls, tool calling, structured output and token usage work as they do there. For TopxAI's Responses API, use ChatOpenAI from langchain-openai with base_url="https://ai.topxea.com/v1" and use_responses_api=True.

Models that return their reasoning (for example deepseek-flash and kimi-k3) put it in additional_kwargs["reasoning_content"], also when streaming:

msg = ChatTopxAI(model="deepseek-flash").invoke("What is 17 * 3?")
msg.additional_kwargs.get("reasoning_content")

with_structured_output uses JSON Schema by default (method="json_schema", as in ChatOpenAI). Keep the default for the Claude models (claude-sonnet-5-5, claude-opus-5-5, claude-fable-5-1): they do not accept a forced tool call, so method="function_calling" fails for them.

Configuration

Links

• Guide: https://ai.topxea.com/docs/use-topxai-with-langchain
• Models and prices: https://ai.topxea.com/pricing
• API reference: https://ai.topxea.com/docs/api-reference

Development

make tests              # unit tests, no network
make integration_tests  # LangChain standard integration tests, needs TOPXAI_API_KEY
make lint type

MIT licensed. Maintained by TopXEA, the operator of TopxAI.