kalibr 1.2.0

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

Kalibr Python SDK

Production-grade observability and execution intelligence for LLM applications. Automatically instrument OpenAI, Anthropic, and Google AI SDKs with zero code changes.

Features

• Zero-code instrumentation - Automatic tracing for OpenAI, Anthropic, and Google AI SDKs
• Outcome-conditioned routing - Query for optimal models based on historical success rates
• TraceCapsule - Cross-agent context propagation for multi-agent systems
• Cost tracking - Real-time cost calculation for all LLM calls
• Token monitoring - Track input/output tokens across providers
• Framework integrations - LangChain, CrewAI, OpenAI Agents SDK

Installation

pip install kalibr

Quick Start

Auto-instrumentation (Recommended)

Simply import kalibr at the start of your application - all LLM calls are automatically traced:

import kalibr  # Must be FIRST import
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)
# That's it. The call is automatically traced.

Manual Tracing with @trace Decorator

For more control, use the @trace decorator:

from kalibr import trace
from openai import OpenAI

@trace(operation="summarize", provider="openai", model="gpt-4o")
def summarize_text(text: str) -> str:
    client = OpenAI()
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": "Summarize the following text."},
            {"role": "user", "content": text}
        ]
    )
    return response.choices[0].message.content

Multi-Provider Example

import kalibr
from openai import OpenAI
from anthropic import Anthropic

# Both are automatically traced
openai_client = OpenAI()
anthropic_client = Anthropic()

gpt_response = openai_client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Explain quantum computing"}]
)

claude_response = anthropic_client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain machine learning"}]
)

Outcome-Conditioned Routing

Query Kalibr for optimal model recommendations based on real execution outcomes:

from kalibr import get_policy, report_outcome

# Before executing - get the best model for your goal
policy = get_policy(goal="book_meeting")
print(f"Use {policy['recommended_model']} - {policy['outcome_success_rate']:.0%} success rate")

# Execute with the recommended model
# ...

# After executing - report what happened
report_outcome(
    trace_id="abc123",
    goal="book_meeting",
    success=True
)

With Constraints

from kalibr import get_policy

policy = get_policy(
    goal="resolve_ticket",
    constraints={
        "max_cost_usd": 0.05,
        "max_latency_ms": 3000,
        "min_quality": 0.8
    }
)

TraceCapsule - Cross-Agent Tracing

Propagate trace context across agent boundaries:

from kalibr import TraceCapsule, get_or_create_capsule

# Agent 1: Create capsule and add hop
capsule = get_or_create_capsule()
capsule.append_hop({
    "provider": "openai",
    "operation": "chat_completion",
    "model": "gpt-4o",
    "duration_ms": 150,
    "cost_usd": 0.002,
    "status": "success"
})

# Pass to Agent 2 via HTTP header
headers = {"X-Kalibr-Capsule": capsule.to_json()}

# Agent 2: Receive and continue
capsule = TraceCapsule.from_json(headers["X-Kalibr-Capsule"])
capsule.append_hop({
    "provider": "anthropic",
    "operation": "chat_completion",
    "model": "claude-3-5-sonnet-20241022",
    "duration_ms": 200,
    "cost_usd": 0.003,
    "status": "success"
})

Framework Integrations

LangChain

pip install kalibr[langchain]

from kalibr_langchain import KalibrCallbackHandler
from langchain_openai import ChatOpenAI

handler = KalibrCallbackHandler()
llm = ChatOpenAI(model="gpt-4o", callbacks=[handler])
response = llm.invoke("What is the capital of France?")

See LangChain Integration Guide for full documentation.

CrewAI

pip install kalibr[crewai]

from kalibr_crewai import KalibrCrewAIInstrumentor
from crewai import Agent, Task, Crew

instrumentor = KalibrCrewAIInstrumentor()
instrumentor.instrument()

# Use CrewAI normally - all operations are traced

See CrewAI Integration Guide for full documentation.

OpenAI Agents SDK

pip install kalibr[openai-agents]

from kalibr_openai_agents import setup_kalibr_tracing
from agents import Agent, Runner

setup_kalibr_tracing()

agent = Agent(name="Assistant", instructions="You are helpful.")
result = Runner.run_sync(agent, "Hello!")

See OpenAI Agents Integration Guide for full documentation.

Configuration

Configure via environment variables:

Variable	Description	Default
KALIBR_API_KEY	API key for authentication	Required
KALIBR_TENANT_ID	Tenant identifier	default
KALIBR_COLLECTOR_URL	Collector endpoint URL	https://api.kalibr.systems/api/ingest
KALIBR_INTELLIGENCE_URL	Intelligence API URL	https://kalibr-intelligence.fly.dev
KALIBR_SERVICE_NAME	Service name for spans	kalibr-app
KALIBR_ENVIRONMENT	Environment (prod/staging/dev)	prod
KALIBR_WORKFLOW_ID	Workflow identifier	default
KALIBR_AUTO_INSTRUMENT	Enable auto-instrumentation	true

CLI Commands

# Serve your app with tracing
kalibr serve myapp.py

# Run with managed runtime
kalibr run myapp.py --port 8000

# Deploy to cloud platforms
kalibr deploy myapp.py --runtime fly.io

# Fetch trace capsule by ID
kalibr capsule <trace-id>

# Show version
kalibr version

Supported Providers

Provider	Models	Auto-Instrumentation
OpenAI	GPT-4, GPT-4o, GPT-3.5	Yes
Anthropic	Claude 3.5 Sonnet, Claude 3 Opus/Sonnet/Haiku	Yes
Google	Gemini Pro, Gemini Flash	Yes

Development

git clone https://github.com/kalibr-ai/kalibr-sdk-python.git
cd kalibr-sdk-python

pip install -e ".[dev]"

# Run tests
pytest

# Format code
black kalibr/
ruff check kalibr/

Contributing

We welcome contributions! See CONTRIBUTING.md.

License

Apache 2.0 - see LICENSE.

Links

• Documentation
• Dashboard
• GitHub
• PyPI