inference-tracing

Short company-qualified install name for the Catalyst by Inference.net Python tracing SDK.

This package depends on catalyst-tracing, the primary Python SDK package, and re-exports the same public API from the inference_tracing import path. Use this package when you want the shortest install command that includes the Inference company name.

This package is currently in beta. APIs may change before 1.0, but the package name and import path are intended to remain stable.

Full documentation lives at docs.inference.net/integrations/traces/overview.

Install

pip install inference-tracing

Install integration extras through the alias package:

pip install 'inference-tracing[openai]'
pip install 'inference-tracing[anthropic]'
pip install 'inference-tracing[langchain]'
pip install 'inference-tracing[langsmith]'
pip install 'inference-tracing[livekit-agents]'
pip install 'inference-tracing[openai,langchain,langsmith]'
pip install 'inference-tracing[all]'

Quick Start

from inference_tracing import setup
from openai import OpenAI

tracing = setup()

client = OpenAI()
client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Summarize this order."}],
)

tracing.shutdown()

You can also import from the primary namespace after installing this package:

from catalyst_tracing import setup

Relationship To catalyst-tracing

The alias package is intentionally small. The tracing runtime, integration installers, OpenTelemetry setup, and OpenInference attribute handling live in catalyst-tracing.

Supported Integrations

The underlying SDK supports OpenAI, Anthropic, LangChain, LangGraph, LangSmith, OpenAI Agents, Claude Agent SDK, Pydantic AI, ElevenLabs Agents, and LiveKit Agents integrations. It also exposes manual manual_span() and agent_span() helpers for custom agent, chain, and tool workflows.

For LangSmith, install inference-tracing[langsmith] or inference-tracing[langchain,langsmith], keep using LangSmith's @traceable decorators, and call setup() once at process startup. The primary SDK configures LangSmith OTel mode and enriches those decorator spans with OpenInference attributes before exporting them to Catalyst. When a LangChain or LangGraph agent runs inside a decorated function, Catalyst uses the LangSmith OTel context as the parent for nested chain, tool, retriever, and provider spans; customer code does not need a separate Catalyst wrapper for trace grouping.

See the catalyst-tracing package page for the full API reference and configuration guide.