Tracelit Python SDK

Drop-in OpenTelemetry SDK for Python that exports traces, metrics, and logs to Tracelit via OTLP/HTTP.

Install

pip install tracelit

Install OpenTelemetry instrumentation packages for your stack (see each framework section below). Tracelit only instruments libraries whose instrumentation package is installed — missing packages are skipped silently.

pip install opentelemetry-instrumentation-flask      # Flask apps
pip install opentelemetry-instrumentation-django     # Django apps
pip install opentelemetry-instrumentation-fastapi    # FastAPI apps
pip install opentelemetry-instrumentation-celery     # Celery workers
pip install opentelemetry-instrumentation-redis        # Redis
pip install opentelemetry-instrumentation-requests   # requests HTTP client

Which setup do I use?

There is no tracelit.init() API.

Django

1. Add to settings.py:

INSTALLED_APPS = [
    "tracelit.integrations.django.TracelitConfig",  # full path required
    "django.contrib.admin",
    # ...
]

2. Set environment variables (.env, deployment config, etc.):

TRACELIT_API_KEY=your-api-key
TRACELIT_SERVICE_NAME=payments-api
TRACELIT_ENVIRONMENT=production
TRACELIT_ENDPOINT=https://ingest.tracelit.app

3. Install Django instrumentation:

pip install opentelemetry-instrumentation-django

4. Restart Django (runserver, gunicorn, uvicorn, etc.).

Tracelit starts automatically in AppConfig.ready() after Django is configured.

python-decouple / django-environ

Bridge secrets into os.environ before INSTALLED_APPS:

import os
from decouple import config

os.environ.setdefault("TRACELIT_API_KEY", config("TRACELIT_API_KEY"))
os.environ.setdefault("TRACELIT_SERVICE_NAME", config("TRACELIT_SERVICE_NAME", default="backend"))
os.environ.setdefault("TRACELIT_ENVIRONMENT", config("TRACELIT_ENVIRONMENT", default="production"))

Do not call tracelit.auto_start() at the top of settings.py.

Celery (Django projects with background workers)

In celery.py:

from tracelit.integrations.celery import install_celery_hook

install_celery_hook()

pip install opentelemetry-instrumentation-celery

Flask

1. app.py — complete working example:

import os

import tracelit

tracelit.auto_start(
    api_key=os.environ["TRACELIT_API_KEY"],
    service_name="my-flask-app",
    environment=os.getenv("FLASK_ENV", "production"),
)

from flask import Flask
from tracelit.integrations.wsgi import TracelitWSGIMiddleware

app = Flask(__name__)
app.wsgi_app = TracelitWSGIMiddleware(app.wsgi_app)


@app.get("/")
def home():
    return "ok"

2. Install Flask instrumentation:

pip install opentelemetry-instrumentation-flask opentelemetry-instrumentation-wsgi

3. Run:

export TRACELIT_API_KEY=your-api-key
FLASK_APP=app.py flask run

FLASK_APP=app.py requires a module-level variable named app. If your factory is named differently, use FLASK_APP=app:create_app instead.

TracelitWSGIMiddleware adds HTTP request metrics. Flask route tracing comes from opentelemetry-instrumentation-flask.

FastAPI

1. main.py — complete working example:

import os

import tracelit

tracelit.auto_start(
    api_key=os.environ["TRACELIT_API_KEY"],
    service_name="my-api",
    environment=os.getenv("ENV", "production"),
)

from fastapi import FastAPI
from tracelit.integrations.asgi import TracelitASGIMiddleware

app = FastAPI()
app = TracelitASGIMiddleware(app)


@app.get("/")
def home():
    return {"ok": True}

2. Install FastAPI instrumentation:

pip install opentelemetry-instrumentation-fastapi opentelemetry-instrumentation-asgi

3. Run with uvicorn:

export TRACELIT_API_KEY=your-api-key
uvicorn main:app

Scripts and batch jobs (no web framework)

Use tracelit.auto_start() at the top of your entry point. This is for CLIs, cron jobs, and one-off scripts — not for Django (use TracelitConfig instead).

import os
import tracelit

tracelit.auto_start(
    api_key=os.environ["TRACELIT_API_KEY"],
    service_name="nightly-report",
    environment=os.getenv("ENV", "production"),
)

with tracelit.tracer().start_as_current_span("generate-report"):
    # your logic here
    pass

Or rely entirely on environment variables:

import tracelit

tracelit.auto_start()

Manual spans and metrics

import tracelit

with tracelit.tracer().start_as_current_span("process-order"):
    tracelit.metrics.counter("orders.processed").add(1)

Configuration

Set TRACELIT_ENABLED=false in tests to disable telemetry with no code changes.

Behavior guarantees

• Non-blocking span export with async error-span queue
• Never raises setup or exporter failures into customer code
• Graceful shutdown flush at process exit
• Error spans are always exported, even when sampling is below 1.0
• Missing instrumentation packages are skipped silently (no startup warning spam)