Skip to main content

Python SDK API

The TrainLoop Evals Python SDK provides zero-touch instrumentation for Python applications.

Installation​

pip install trainloop-sdk

Quick Start​

import trainloop

# Initialize the SDK
trainloop.init(endpoint="http://localhost:8000")

# Your LLM calls are automatically instrumented
import openai
client = openai.OpenAI()
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello, world!"}]
)

Configuration​

The SDK can be configured using environment variables or initialization parameters:

Environment Variables​

  • TRAINLOOP_ENDPOINT - The endpoint to send traces to (default: http://localhost:8000)
  • TRAINLOOP_API_KEY - API key for authentication (optional)
  • TRAINLOOP_DEBUG - Enable debug logging (default: false)

Initialization Parameters​

import trainloop

trainloop.init(
    endpoint="http://localhost:8000",
    api_key="your-api-key",
    debug=True
)

Supported Libraries​

The Python SDK automatically instruments the following libraries:

  • OpenAI
  • Anthropic
  • LiteLLM
  • LangChain
  • LlamaIndex

Manual Instrumentation​

For custom instrumentation, you can use the manual tracing API:

import trainloop

with trainloop.trace("my-llm-call") as span:
    span.set_input({"prompt": "Hello, world!"})
    # Your LLM call here
    span.set_output({"response": "Hello! How can I help you?"})