Will there be a ChatOpenRouter (or some equivalent) implementation?

OSS Product Help LangGraph
1

Hello,

I was wondering if there was an implementation for chat models using the OpenRouter platform/API. I have viewed this link which works fine for building and testing purposes but I am not finding it helpful while deploying the application. If anyone would happen to know if there is some official implementation (either Langchain or community) in the works, I would really appreciate knowing about any related developments.

I am aware that I can likely build a custom class to interact with this but ideally I would like it to be similar to the other chat models and implement exception handling, message management for the LLM, etc., natively. I am also unsure of how to set up a feature where we can easily switch between models as we can with the API (and automatically alter the request/message formats, tool calling, and so on).

Thank you.

2

My understanding is that you should be able to use ChatOpenAI directly with OpenRouter as they implement OpenAI’s completions endpoint? All that’s required is updating the BaseURL.

3

This sample code already works as the practical equivalent of a ChatOpenRouter implementation using OpenRouter’s OpenAI-compatible API.

you need to add OPENROUTER_API_KEY to work this code

import os
from langchain.agents import create_agent
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI

# Load environment variables
load_dotenv()

# Get API key
OPENROUTER_KEY = os.getenv("OPENROUTER_API_KEY")
if not OPENROUTER_KEY:
    raise ValueError("Missing OPENROUTER_API_KEY in environment")

# Set OpenAI-compatible environment variables for OpenRouter
os.environ["OPENAI_API_KEY"] = OPENROUTER_KEY
os.environ["OPENAI_BASE_URL"] = "https://openrouter.ai/api/v1"

# Create LLM
llm = ChatOpenAI(
    model="mistralai/mistral-7b-instruct:free",
    temperature=0.7,
    api_key=OPENROUTER_KEY,
)

# Create agent
agent = create_agent(
    llm,
    tools=[],
    system_prompt="You are a helpful assistant. Answer in simple words."
)

# User input
user_question = "What is artificial intelligence in simple terms?"

# Invoke agent
response = agent.invoke({
    "messages": [
        {"role": "user", "content": user_question}
    ]
})

# Extract and print AI response safely
if "messages" in response and len(response["messages"]) > 0:
    ai_message = response["messages"][-1]
    print(ai_message.content)
else:
    print("No response received from the agent.")