A small Streamlit app to monitor the GitHub Actions workflows of a repository. Built on the GitHub REST API for workflow runs.
The UI is in German, but the code and this README are in English — the app was built for a German-speaking audience, but should be straightforward to reuse elsewhere.
- Overview of all workflows with their current status badges
- Time series of runs by started-at, colored by conclusion
- Distribution of runs by conclusion, trigger event, and head branch
- Mean run duration per workflow
- Table of currently running (non-completed) runs
- Table of failed runs
- Full data table of all fetched runs
Prerequisites: Python 3.11+ and uv.
uv sync
cp .streamlit/secrets.toml.example .streamlit/secrets.toml
# edit .streamlit/secrets.toml and fill in ACCESS_TOKEN and REPO
uv run streamlit run actions_monitor.pySecrets are read from .streamlit/secrets.toml (see Streamlit docs):
| Key | Required | Description |
|---|---|---|
ACCESS_TOKEN |
yes | GitHub personal access token (see below) |
REPO |
yes | OWNER/REPO of the repository to monitor |
Example: REPO = "opendatazurich/opendatazurich.github.io".
Create a personal access token with the following scopes:
- Fine-grained PAT (recommended):
Actions: Read-only+Metadata: Read-only, scoped to the target repository. - Classic PAT:
reposcope for private repositories; no scope needed for public repositories (but the unauthenticated rate limit is much lower, so a token is still recommended).
The app runs on Streamlit Community Cloud out of the box — connect the repository and set ACCESS_TOKEN and REPO via the Secrets UI. Running in Docker is possible but out of scope for this README.
- The authenticated GitHub REST API rate limit is 5000 requests/hour. The "Anzahl Ergebnisse" slider in the UI controls how many runs are fetched — higher values mean more API calls and slower loads.
- API responses are cached in-process for 5 minutes (
@st.cache_data(ttl=300)); restart the app or wait out the TTL to force a refresh.
Public repositories work with reduced token scopes (classic PAT without repo, or no token at all with sharply reduced rate limits). Private repositories require a token with read access.
uv sync # install dependencies (incl. dev group)
uv run pytest # run the test suite
uv run ruff check . # lintCI runs the same lint + test commands on push/PR across Python 3.11–3.14.
Contributions are welcome! Feel free to open an issue to discuss ideas or send a pull request directly — bug fixes, new features, UI improvements, or documentation tweaks are all appreciated. Please make sure tests and lint pass locally (uv run pytest, uv run ruff check .) before submitting.
MIT — © 2026 Open Data Zürich.