langchain-revana

LangChain integration for Revana: Claude, Gemini, DeepSeek and other text models from several labs behind one API key, through Revana's OpenAI-compatible Chat Completions endpoint. Each request is routed to the cheapest working provider.

Install

pip install -U langchain-revana
export REVANA_API_KEY="sr_live_..."   # create one at https://revana.si/dashboard

Chat model

from langchain_revana import ChatRevana

llm = ChatRevana(model="claude-haiku-5-5", max_tokens=256)

print(llm.invoke("Explain what an API gateway does in one sentence.").content)

for chunk in llm.stream("Count from 1 to 5."):
    print(chunk.content, end="")

ChatRevana supports everything ChatOpenAI does that the chosen model supports: streaming, async, tool calling (bind_tools), and structured output.

from langchain_core.tools import tool

@tool
def get_weather(city: str) -> str:
    """Get the weather for a city."""
    return "sunny"

msg = llm.bind_tools([get_weather]).invoke("What's the weather in Paris?")
print(msg.tool_calls)

Models

Model IDs and live per-token prices: https://api.revana.si/v1/models (text models have "kind": "text"; no key needed to list them).

Revana also serves image, video and music models (Seedance 2.5, Veo 3.1, Kling, Nano Banana Pro and more) through its generations API; see https://revana.si/docs and the revana Python SDK.

Configuration

License

MIT