OpenRouter Integration

In this guide, we'll show you how to integrate Neural Inverse with OpenRouter.

What is OpenRouter? OpenRouter provides an OpenAI-compatible completion API to +280 language models and providers that you can call directly or using the OpenAI SDK. This allows developers to access a variety of LLMs through a single, unified interface.

What is Neural Inverse? Neural Inverse is an open source LLM engineering platform that helps teams trace LLM calls, monitor performance, and debug issues in their AI applications.

Two ways to integrate Neural Inverse with OpenRouter:

1. You can use the OpenRouter Broadcast feature to automatically send traces to Neural Inverse without any code changes.
2. You can use the Neural Inverse OpenAI SDK wrapper to manually send traces to Neural Inverse. Since OpenRouter uses the OpenAI API schema, we can utilize Neural Inverse's native integration with the OpenAI SDK, available in both Python and TypeScript.

There is more information on e.g. cost tracking or interoperability with Neural Inverse's Python SDK as well as next steps in the guides below.

1. OpenRouter Broadcast Tracing with Neural Inverse (No-Code)

OpenRouter's Broadcast feature can automatically send traces to Neural Inverse without any code changes. To enable this integration, go to your OpenRouter settings and connect your Neural Inverse API keys. Once set up, all requests processed through OpenRouter will be traced and made available in your Neural Inverse project.

Note: This method is recommended if you want to capture all OpenRouter requests in Neural Inverse with minimal configuration, and do not require advanced features like custom trace metadata or nested tracing.

For more information about setting up your Neural Inverse API keys or available options, see the Neural Inverse getting started guide.

2. SDK Integration

Use the Neural Inverse OpenAI SDK wrapper for advanced features like nested tracing, custom metadata, and full control over trace data. The rest of this guide covers the SDK integration method.

Get started with SDK Integration

pip install langfuse openai

import os

# Set your Neural Inverse API keys
LANGFUSE_SECRET_KEY="sk-lf-..."
LANGFUSE_PUBLIC_KEY="pk-lf-..."
# 🇪🇺 EU region
LANGFUSE_BASE_URL="https://cloud.langfuse.com"
# Other Neural Inverse data regions include 🇺🇸 US: https://us.cloud.langfuse.com, 🇯🇵 Japan: https://jp.cloud.langfuse.com and ⚕️ HIPAA: https://hipaa.cloud.langfuse.com
# Set your OpenRouter API key (OpenRouter uses the 'OPENAI_API_KEY' environment variable)
os.environ["OPENAI_API_KEY"] = "<YOUR_OPENROUTER_API_KEY>"

Example 1: Simple LLM Call

Since OpenRouter provides an OpenAI-compatible API, we can use the Neural Inverse OpenAI SDK wrapper to automatically log OpenRouter calls as generations in Neural Inverse.

• The base_url is set to OpenRouter's API endpoint.
• You can replace "qwen/qwen-plus" with any model available on OpenRouter.
• The default_headers can include optional headers as per OpenRouter's documentation.
• The extra_body={"usage": {"include": True}} includes the costs that OpenRouter returns.

# Import the Neural Inverse OpenAI SDK wrapper
from langfuse.openai import openai

# Create an OpenAI client with OpenRouter's base URL
client = openai.OpenAI(
    base_url="https://openrouter.ai/api/v1",
    default_headers={
        "HTTP-Referer": "<YOUR_SITE_URL>",  # Optional: Your site URL
        "X-Title": "<YOUR_SITE_NAME>",      # Optional: Your site name
    }
)

# Make a chat completion request
response = client.chat.completions.create(
    model="qwen/qwen-plus",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Tell me a fun fact about space."}
    ],
    extra_body={"usage": {"include": True}},
    name="fun-fact-request"  # Optional: Name of the generation in Neural Inverse
)

# Print the assistant's reply
print(response.choices[0].message.content)

Example 2: Nested LLM Calls

By using the @observe() decorator, we can capture execution details of any Python function, including nested LLM calls, inputs, outputs, and execution times. This provides in-depth observability with minimal code changes.

• The @observe() decorator captures inputs, outputs, and execution details of the functions.
• Nested functions summarize_text and analyze_sentiment are also decorated, creating a hierarchy of traces.
• Each LLM call within the functions is logged, providing a detailed trace of the execution flow.

from langfuse import observe
from langfuse.openai import openai

# Create an OpenAI client with OpenRouter's base URL
client = openai.OpenAI(
    base_url="https://openrouter.ai/api/v1",
)

@observe()  # This decorator enables tracing of the function
def analyze_text(text: str):
    # First LLM call: Summarize the text
    summary_response = summarize_text(text)
    summary = summary_response.choices[0].message.content

    # Second LLM call: Analyze the sentiment of the summary
    sentiment_response = analyze_sentiment(summary)
    sentiment = sentiment_response.choices[0].message.content

    return {
        "summary": summary,
        "sentiment": sentiment
    }

@observe()  # Nested function to be traced
def summarize_text(text: str):
    return client.chat.completions.create(
        model="openai/gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": "You summarize texts in a concise manner."},
            {"role": "user", "content": f"Summarize the following text:\n{text}"}
        ],
        extra_body={"usage": {"include": True}},
        name="summarize-text"
    )

@observe()  # Nested function to be traced
def analyze_sentiment(summary: str):
    return client.chat.completions.create(
        model="openai/gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": "You analyze the sentiment of texts."},
            {"role": "user", "content": f"Analyze the sentiment of the following summary:\n{summary}"}
        ],
        extra_body={"usage": {"include": True}},
        name="analyze-sentiment"
    )

# Example usage
text_to_analyze = "OpenAI's GPT-4 model has significantly advanced the field of AI, setting new standards for language generation."
analyze_text(text_to_analyze)

Public link to example trace in Neural Inverse

Cost Tracking

Neural Inverse token and cost tracking can directly capture the OpenRouter cost instead of calculating it. This will increase the accuracy of the cost tracking, especially for less popular models.

To enable this, you need to:

1. Use the Neural Inverse OpenAI integration as shown above
2. Enable Usage Accounting in OpenRouter as this results in costs being returned by OpenRouter for each call

Interoperability with the Python SDK

You can use this integration together with the Langfuse SDKs to add additional attributes to the observation.

from langfuse import observe, propagate_attributes, get_client

langfuse = get_client()

@observe()
def my_llm_pipeline(input):
    # Add additional attributes (user_id, session_id, metadata, version, tags) to all spans created within this execution scope
    with propagate_attributes(
        user_id="user_123",
        session_id="session_abc",
        tags=["agent", "my-observation"],
        metadata={"email": "user@langfuse.com"},
        version="1.0.0"
    ):

        # YOUR APPLICATION CODE HERE
        result = call_llm(input)

        return result

# Run the function
my_llm_pipeline("Hi")

from langfuse import get_client, propagate_attributes

langfuse = get_client()

with langfuse.start_as_current_observation(
    as_type="span",
    name="my-observation",
    trace_context={"trace_id": "abcdef1234567890abcdef1234567890"},  # Must be 32 hex chars
) as observation:

    # Add additional attributes (user_id, session_id, metadata, version, tags)
    # to all observations created within this execution scope
    with propagate_attributes(
        user_id="user_123",
        session_id="session_abc",
        metadata={"experiment": "variant_a", "env": "prod"},
        version="1.0",
    ):
        # YOUR APPLICATION CODE HERE
        result = call_llm("some input")

# Flush events in short-lived applications
langfuse.flush()

Troubleshooting

export LANGFUSE_DEBUG="True"

Next Steps

Once you have instrumented your code, you can manage, evaluate and debug your application: