Pytest Dagger Toolchain

A toolchain for testing Python application with automatic OpenTelemetry tracing

This toolchain automatically injects a pytest_otel library for test tracing visibility in Dagger TUI and Dagger Cloud. No configuration required.

Usage:

On a python project, using pytest as the test runner:

• initialize a dagger module if not already done: dagger init .
• install the pytest toolchain: dagger toolchain install github.com/dagger/pytest
• run tests: dagger check pytest:test. It will:
  • creates an Alpine based container with uv (a uv cache volume is set)
  • prepares Python for the requested version via uv, failing fast if it can't be resolved
  • injects the pytest_otel dependency for automatic test tracing (before your source is added, so it stays cached)
  • installs your project dependencies (automatically via uv run for pyproject.toml projects, or from requirements.txt)
  • export captured stdout, stderr, and Python logging records as OTel logs
  • run pytest

pytest:test runs against a self-contained Alpine + uv base, so it needs no setup. To trace pytest inside your own container instead (your Python, your dependencies, your environment), pass that container to the toolchain or use the lower-level functions directly.

Usage modes

The toolchain serves three levels of control:

1. Plain toolchain - install it and run dagger check. Your project is taken from the workspace automatically; pythonVersion (and sourcePath, to target a subdirectory of the workspace) are the only knobs commonly set. Everything else just works against the bundled Alpine + uv base.
2. Custom image (drop-in) - set container to your own image (your Python, your tooling, uv or not). It replaces the default base; the module still installs pytest_otel and runs pytest. If your image is not uv-based, set runner to PIP or leave AUTO to auto-detect.
3. Embedded / custom - call the building blocks directly:
   • test - full pipeline (base -> source -> deps -> otel -> pytest), with skipInstallDeps: true to skip mounting the source and installing dependencies when your image already has both.
   • testUv / testPip - run-only: given a container that already has source + dependencies, install pytest_otel and run pytest. Use these to skip detection entirely.
   • installPytestOtel(ctr, runner, envTarget) - install only pytest_otel into your container (AUTO detects the tool; UV/PIP force one), then run pytest yourself. Use envTarget to control where it lands (see below) - in particular SYSTEM for a container without a .venv.
   • pytestOtel - get the bundled pytest_otel library as a Directory for fully-manual integration (it is not yet published to PyPI).

runner is AUTO | UV | PIP (default AUTO: prefer uv, else pip).

envTarget is AUTO | VENV | SYSTEM (default AUTO): AUTO installs into an existing virtual environment when present and otherwise falls back to a system install; VENV forces a virtual environment (created when missing); SYSTEM forces a system install (--system --break-system-packages). Containers without a .venv work with AUTO and SYSTEM.

Example, using fastly/fastly-cli:

$ git clone https://github.com/fastly/cli

$ cd cli

$ # Initialize an empty dagger module
$ dagger init

Initialized module fastapi-cli in ~/dev/src/github.com/fastapi/fastapi-cli

$ # Add the pytest toolchain
$ dagger toolchain install github.com/dagger/pytest

toolchain installed

$ # Run tests, with tracing enabled
$ dagger check
✔ pytest:test 17.4s ⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣧⣿⣿⣿⣿⣿⣿⣿⡆⡄⣿⣿⡄ OK