gyvm/PR Insights — Collect & Build Dashboard

In one step: collect GitHub pull requests into a committed DuckDB warehouse (incremental, with on-demand history backfill) AND build the static Explore + Reports dashboard into dist/. Run inside a consumer repo that holds the config and the accumulating data/dwh; commit the DWH and deploy dist/ afterwards.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Activelast commit Jun 26, 2026
License
MIT

Pinned Snippet

workflow.ymlSHA-pinned
uses: gyvm/dev-prism@f38dcf8b1e4c19846b4adf6066018a9c2e1ce0e3 # v0.3.0

tags can be moved; commit SHAs can't. why a SHA?

namedescriptionrequireddefault
configPath to config.toml, relative to the repo root.noconfig.toml
dwh-dirDirectory for the committed DWH parquet files.nodata/dwh
fromBackfill history down to this date (YYYY-MM-DD), fetching only the uncovered older slice per repo. Omit for a normal incremental run.no""
baseAstro base path for the built site. "/" for a custom domain or Cloudflare Pages; "/<repo>/" for a GitHub Pages project page (otherwise every asset 404s).no/
siteAbsolute site URL for canonical/OG links (optional).no""
output-dirWhere to write the built site in the consumer workspace.nodist
reports-configPath to reports.toml; when set, frozen reports populate the gallery.no""
github-tokenRead-only PAT with access to the target repos, used by the collector (reads as GITHUB_TOKEN). Pass it here via `with:` — env set on the calling `uses:` step is NOT propagated into a composite action's run steps. For GitHub App auth, set GITHUB_APP_ID / GITHUB_APP_PRIVATE_KEY / GITHUB_APP_INSTALLATION_ID as job- or workflow-level env instead (those ARE shared with composite steps).no""

no outputs