Observability for OpenRouter with Opik

This guide explains how to integrate Opik with OpenRouter using the OpenAI SDK. OpenRouter provides a unified API for accessing hundreds of AI models through a single OpenAI-compatible interface.

First, ensure you have both opik and openai packages installed:

pip install opik openai

You'll also need an OpenRouter API key which you can get from OpenRouter.

from opik.integrations.openai import track_openai
from openai import OpenAI

# Initialize the OpenAI client with OpenRouter base URL
client = OpenAI(
    base_url="https://openrouter.ai/api/v1",
    api_key="YOUR_OPENROUTER_API_KEY"
)
client = track_openai(client)

# Optional headers for OpenRouter leaderboard
headers = {
    "HTTP-Referer": "YOUR_SITE_URL",  # Optional. Site URL for rankings
    "X-Title": "YOUR_SITE_NAME"  # Optional. Site title for rankings
}

response = client.chat.completions.create(
    model="openai/gpt-4",  # You can use any model available on OpenRouter
    extra_headers=headers,
    messages=[
        {"role": "user", "content": "Hello, world!"}
    ],
    temperature=0.7,
    max_tokens=100
)

print(response.choices[0].message.content)

OpenRouter provides access to a wide variety of models, including many open source models from different providers.

• OpenAI models (GPT-4o, o1, o3-mini)
• Anthropic models (Opus, Sonnet, Haiku)
• Google models (Gemini Pro, Flash, Flash Thinking)
• And many open source models

You can find the complete list of available models in the OpenRouter documentation.

OpenRouter supports the following methods:

• client.chat.completions.create(): Works with all models
• Provides standard chat completion functionality
• Compatible with the OpenAI SDK interface

• client.beta.chat.completions.parse(): Only compatible with OpenAI models
• For non-OpenAI models, see OpenRouter's Structured Outputs documentation

For detailed information about available methods, parameters, and best practices, refer to the OpenRouter API documentation.