Skip to main content

Deploying the History Repository

No clone and no separate extension repository required:

curl -fsSL https://raw.githubusercontent.com/mzbrau/actions-insights/main/scripts/init-history-repo.sh | bash -s -- init

With options:

curl -fsSL https://raw.githubusercontent.com/mzbrau/actions-insights/main/scripts/init-history-repo.sh | \
  bash -s -- init my-org-test-history --org my-org --default-repository my-org/my-project

Requires gh and git on your PATH. The script shallow-clones mzbrau/actions-insights into a temporary directory to copy templates, then creates your history repository.

Before making changes, the script shows a summary of the repository name, visibility, and planned steps, then asks you to confirm, cancel, or edit settings. When installed via curl | bash, prompts still work when a terminal is available.

From a local checkout you can run the same script directly:

bash scripts/init-history-repo.sh init --dry-run

Use --dry-run to preview the same summary without executing or prompting. Use --yes (or -y) to skip confirmation in non-interactive or CI environments.

Optional: GitHub CLI extension​

gh extension install requires a repository whose name starts with gh-, so the main actions-insights repository cannot be used directly. If you prefer a gh subcommand, install the optional extension:

gh extension install mzbrau/gh-actions-insights
gh actions-insights init
gh actions-insights update <owner>/<history-repo>
gh actions-insights import <owner>/<source-repo> <owner>/<history-repo>

See Importing historical test data for backfilling runs from workflow artifacts.

Maintainers publish that extension from this monorepo with npm run publish:extension.

Manual setup​

  1. Create a new repository
  2. Copy contents from templates/history-repo/ and web/ from this monorepo
  3. Run bash scripts/prepare-standalone-web.sh web packages/history-models to vendor shared types and generate web/package-lock.json
  4. Enable GitHub Pages with source: GitHub Actions
  5. Push to main — the pages.yml workflow builds and deploys

Dashboard URL​

For project Pages: https://{owner}.github.io/{repo-name}/

The dashboard uses hash-based routing so deep links work on static GitHub Pages hosting. Run detail URLs look like:

https://{owner}.github.io/{repo-name}/#/r/{repo-key}/b/{branch}/run/{run-id}

For example: https://my-org.github.io/test-history/#/r/owner.repo/b/main/run/42

Updating the dashboard​

To pull in dashboard and workflow improvements from actions-insights, run the update script. It validates the target repository, syncs web/ and .github/workflows/pages.yml, and opens a pull request. Your data/ and config.json are left unchanged.

Like the init script, update shows a summary and asks you to confirm, cancel, or edit settings (target repository, source ref, verify build) before cloning or opening a pull request.

curl -fsSL https://raw.githubusercontent.com/mzbrau/actions-insights/main/scripts/update-history-repo.sh | \
  bash -s -- update <owner>/<history-repo>

From a local checkout:

bash scripts/update-history-repo.sh update <owner>/<history-repo>

Or with the GitHub CLI extension:

gh actions-insights update <owner>/<history-repo>

Use --dry-run to preview the planned actions without executing or prompting. Use --verify to run npm ci and npm run build while preparing web/. Use --yes (or -y) to skip confirmation in non-interactive or CI environments.

You can still modify web/ directly in the history repository and push. The Pages workflow rebuilds on changes to web/, data/, or config.json.

JSON-only pushes from the action do not require React source changes — the workflow rebuilds the static site and copies current data/ into the deploy artifact.

What not to commit​

The history repository template includes a .gitignore that excludes web/node_modules/ and web/dist/. CI installs dependencies from web/package-lock.json during the Pages build. Commit web/package-lock.json and web/vendor/history-models/, but not install or build output.

Fixing an already-created history repository​

If your history repository was created before standalone web preparation was added, the Pages workflow may fail on the first run with:

Some specified paths were not resolved, unable to cache dependencies.

The easiest fix is the update script, which also adds .gitignore and removes tracked node_modules if present:

bash scripts/update-history-repo.sh update <owner>/<history-repo>

Manual fix from a checkout of actions-insights (note the /web suffix on the history repo path):

git clone https://github.com/<owner>/<history-repo>.git /tmp/history-repo-fix
bash scripts/prepare-standalone-web.sh /tmp/history-repo-fix/web packages/history-models

You can also pass the history repo root instead of web/; the script will detect that and use web/ automatically.

Then update .github/workflows/pages.yml in the history repository to match the current template (Node 24, npm ci, and cache-dependency-path: web/package-lock.json), add .gitignore from templates/history-repo/.gitignore, commit, and push:

cd /tmp/history-repo-fix
cp /path/to/actions-insights/templates/history-repo/.github/workflows/pages.yml .github/workflows/pages.yml
cp /path/to/actions-insights/templates/history-repo/.gitignore .gitignore
git rm -r --cached web/node_modules 2>/dev/null || true
git add web .github/workflows/pages.yml .gitignore
git commit -m "Fix Pages workflow for standalone web build"
git push