Auto-Instrumentation

Zero-code LLM tracing. Just import and go.

You will learn: - Which providers are auto-tracked - What data is captured - How to add context

How It Works

import kalibr  # This patches the SDKs
import openai

# All calls are now tracked automatically
client = openai.OpenAI()
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello"}]
)

Supported Providers

Works with LangChain, LlamaIndex, CrewAI (they use these SDKs internally).

What's Captured

For every LLM call: - Provider and model - Input/output tokens - Duration - Cost (calculated automatically) - Success/error status

Adding Context

import os

os.environ["KALIBR_WORKFLOW_ID"] = "research_pipeline"
os.environ["KALIBR_TENANT_ID"] = "acme_corp"

# All subsequent calls tagged with this context

Import Order Matters

# ✓ Correct
import kalibr
import openai

# ✗ Wrong - openai won't be instrumented
import openai
import kalibr

Disabling Auto-Instrumentation

export KALIBR_AUTO_INSTRUMENT=false

Then use manual @trace decorators instead.

Next Steps

• Manual Tracing — Track non-LLM operations
• Workflows — Group related calls