publish-community-curve.yml

# Publish the community curve commons
#
# Pools the human contributions reviewed on main (community-curves/*.json) with
# the accumulated self-samples (seed-self.json on the community-data branch) into
# one community-curve.json. That file's raw URL is what users set COMMUNITY_CURVE
# to, so any zone others have sampled gains an hour-of-day and day-of-week profile
# even with no free history.
#
# The pool is published to the orphan `community-data` branch (not main), so the
# bot's refreshes never touch main's history.

name: Publish community curve

on:
  push:
    branches: [main]
    paths:
      - 'community-curves/*.json'
  schedule:
    - cron: '23 5 * * *'   # daily, after most contributions settle
  workflow_dispatch:

permissions:
  contents: write          # to push to the community-data branch

concurrency:
  group: community-data     # serialize with the sampler: both write community-data
  cancel-in-progress: false

jobs:
  publish:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v5    # main: code + reviewed community-curves/
      - uses: astral-sh/setup-uv@v7

      - name: Check out the community-data branch into a worktree
        run: |
          git config user.name  "github-actions[bot]"
          git config user.email "github-actions[bot]@users.noreply.github.com"
          git fetch origin community-data || true
          if git worktree add -B community-data ../data origin/community-data 2>/dev/null; then
            echo "Tracking existing community-data branch."
          else
            echo "Creating orphan community-data branch."
            git worktree add --detach ../data
            git -C ../data checkout --orphan community-data
            git -C ../data rm -rf . >/dev/null 2>&1 || true
          fi

      - name: Merge contributions and self-samples into the pool
        run: |
          shopt -s nullglob
          inputs=(community-curves/*.json)          # reviewed human contributions on main
          [ -f ../data/seed-self.json ] && inputs+=(../data/seed-self.json)
          if [ ${#inputs[@]} -eq 0 ]; then
            echo "Nothing to pool yet."
            exit 0
          fi
          # Defense in depth: refuse to pool malformed or implausible data.
          # --min-hours 0 because accumulating seed files are legitimately sparse
          # early on; volume-weighting and --cap-n already neutralize any skew.
          uv run cli.py validate-curves --min-hours 0 "${inputs[@]}"
          # Cap each contributor's per-cell weight so no single repo dominates a zone
          uv run cli.py merge-curves "${inputs[@]}" --cap-n 1000 --output ../data/community-curve.json

      - name: Commit and push if changed
        working-directory: ../data
        run: |
          if [ ! -f community-curve.json ]; then
            echo "No pool produced."
            exit 0
          fi
          git add community-curve.json
          if git diff --cached --quiet; then
            echo "Pool unchanged."
          else
            git commit -m "chore: refresh community curve pool"
            git push origin community-data
          fi