import { Callout } from 'nextra/components'

Observability

DocsGPT bundles the OpenTelemetry SDK and auto-instrumentation packages in application/requirements.txt — they install with the rest of the backend deps. Telemetry is off by default; opt in by prefixing the launch command with opentelemetry-instrument and setting OTLP env vars.

Auto-instrumentation covers Flask, Starlette, Celery, SQLAlchemy, psycopg, Redis, requests, and Python logging. LLM/retriever calls are not captured at this layer — see Going further below.

Enabling

Set these env vars in your .env (or compose environment: block):

bash

OTEL_SDK_DISABLED=false
OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
OTEL_EXPORTER_OTLP_ENDPOINT=https://your-collector.example.com
OTEL_EXPORTER_OTLP_HEADERS=Authorization=Bearer%20<token>
OTEL_TRACES_EXPORTER=otlp
OTEL_METRICS_EXPORTER=otlp
OTEL_LOGS_EXPORTER=otlp
OTEL_PYTHON_LOG_CORRELATION=true
OTEL_RESOURCE_ATTRIBUTES=service.name=docsgpt-backend,deployment.environment=prod

Then prefix the process command with opentelemetry-instrument. The simplest way is a compose override (no image rebuild):

yaml

# deployment/docker-compose.override.yaml
services:
  backend:
    command: >
      opentelemetry-instrument gunicorn -w 1 -k uvicorn_worker.UvicornWorker
      --bind 0.0.0.0:7091 --config application/gunicorn_conf.py
      application.asgi:asgi_app
    environment:
      - OTEL_SERVICE_NAME=docsgpt-backend
  worker:
    command: opentelemetry-instrument celery -A application.app.celery worker -l INFO -B
    environment:
      - OTEL_SERVICE_NAME=docsgpt-celery-worker

For local dev, prepend dotenv run -- so the OTEL_* vars from .env reach opentelemetry-instrument before it boots the SDK:

bash

dotenv run -- opentelemetry-instrument flask --app application/app.py run --port=7091
dotenv run -- opentelemetry-instrument celery -A application.app.celery worker -l INFO --pool=solo

<Callout type="info" emoji="ℹ️"> Logs are exported in-process when `OTEL_LOGS_EXPORTER=otlp` is set — `application/core/logging_config.py` detects the flag and preserves the OTEL log handler. Without it, `logging` writes only to stdout. </Callout>

Backend examples

Axiom

bash

OTEL_EXPORTER_OTLP_ENDPOINT=https://api.axiom.co
OTEL_EXPORTER_OTLP_HEADERS=Authorization=Bearer%20xaat-XXXX,X-Axiom-Dataset=docsgpt
OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf

%20 is the URL-encoded space between Bearer and the token. Create the dataset in the Axiom UI before sending.

Self-hosted OTLP collector / Jaeger / Tempo

bash

OTEL_EXPORTER_OTLP_ENDPOINT=http://otel-collector:4317
OTEL_EXPORTER_OTLP_PROTOCOL=grpc

Honeycomb / Grafana Cloud / Datadog

Each vendor publishes a single-line OTEL_EXPORTER_OTLP_ENDPOINT plus OTEL_EXPORTER_OTLP_HEADERS recipe — drop them in alongside the service-name override.

Caveats

The Dockerfile uses gunicorn -w 1. If you raise worker count, move SDK init into a post_worker_init hook to avoid one-thread-per-process exporter contention.
asgi.py wraps Flask in Starlette's WSGIMiddleware. Both instrumentors are installed, so each request produces a Starlette span enclosing a Flask span. Drop opentelemetry-instrumentation-flask from requirements.txt if the duplication is noisy.
OTEL packages add ~50 MB to the image. They install on every build — the runtime cost is zero unless you set opentelemetry-instrument on the command and set the OTLP env vars.
The OTEL exporter ecosystem currently caps protobuf at <7, so the backend runs on protobuf 6.x. This will catch up in a future OTEL release.