← Back to Documentation

Quickstart

Start capturing LLM traces in 2 minutes.

1Install

Python

pip install kalibr

TypeScript

npm install @kalibr/sdk

2Set API Key

Get your key from dashboard.kalibr.systems

export KALIBR_API_KEY=sk_live_...
export KALIBR_TENANT_ID=my-tenant

3Use

Python (Auto-Instrumentation)

Just import kalibr first. All OpenAI/Anthropic/Google calls are traced automatically:

import kalibr  # Must be first!
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)
# ✅ Trace captured automatically

TypeScript (SpanBuilder)

TypeScript requires explicit span creation:

import { Kalibr, SpanBuilder } from '@kalibr/sdk';
import OpenAI from 'openai';

// Initialize once
Kalibr.init({
  apiKey: process.env.KALIBR_API_KEY!,
  tenantId: process.env.KALIBR_TENANT_ID!,
});

const openai = new OpenAI();

// Create span
const span = new SpanBuilder()
  .setProvider('openai')
  .setModel('gpt-4o-mini')
  .setOperation('chat')
  .start();

const response = await openai.chat.completions.create({
  model: 'gpt-4o-mini',
  messages: [{ role: 'user', content: 'Hello!' }],
});

// Finish with tokens
await span.finish({
  inputTokens: response.usage?.prompt_tokens ?? 0,
  outputTokens: response.usage?.completion_tokens ?? 0,
});
// ✅ Trace captured

View Traces

See your traces at dashboard.kalibr.systems

What you'll see:

  • Provider & Model: openai/gpt-4o-mini
  • Tokens: Input/output counts
  • Cost: USD (auto-calculated)
  • Latency: Response time
  • Trace ID: For debugging

Python vs TypeScript

Python TypeScript
Auto-instrumentation ✓ Yes ✗ No
Pattern import kalibr SpanBuilder
Tokens Auto-captured Pass to finish()
Initialization Env vars only Kalibr.init()

Next Steps

← Back to Documentation