kalibr 1.1.0a0

Unified LLM Observability & Multi-Model AI Integration Framework - Deploy to GPT, Claude, Gemini, Copilot with full telemetry

Kalibr Python SDK

Production-grade observability for LLM applications.

Installation

pip install kalibr

Quickstart

from kalibr import trace
import openai

@trace(api_key="your-kalibr-api-key")
def my_agent():
    response = openai.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "user", "content": "Hello!"}]
    )
    return response

Features

• ✅ Zero-code instrumentation for OpenAI, Anthropic, Google AI
• ✅ Automatic parent-child trace relationships
• ✅ Real-time cost tracking
• ✅ Token usage monitoring
• ✅ Performance metrics

CLI Tools

# Run your app locally
kalibr serve myapp.py

# Deploy to Fly.io
kalibr deploy myapp.py

# Fetch trace data
kalibr capsule <trace-id>

Examples

See examples/ directory for complete examples.

Documentation

Full docs at https://docs.kalibr.systems

License

MIT