hermes-agent/.github/workflows/install-e2e.yml

name: Install & Update E2E

# Can a user on a released version get to this commit?
#
# The support matrix -- every {os, install-method, update-method} combination
# a user could be on -- lives in scripts/sandbox/generate-e2e-matrix.mjs.
# generate-matrix expands it against the picked release tags into one leg
# per {combination, tag}, split into one matrix job per OS:
#
#   Matrix: linux     the real curl|bash install one-liner, isolated by a
#                     git URL redirect to a local bare clone
#                     (install-e2e-run.yml)
#   Matrix: windows   the real desktop user flow: website Hermes-Setup.exe
#                     clicked by AutoHotkey, update via the app, Playwright
#                     clicking "Update now" (install-e2e-windows-run.yml)
#   Matrix: macos     script installs on the shared OS-agnostic driver,
#                     plus the real desktop user flow: website
#                     Hermes-Setup.dmg mounted and run, updates via the
#                     app under Playwright (install-e2e-macos-run.yml)
#
# Every combination is dispatched to its OS's run workflow; the run
# workflow natively skips (grey) what its driver cannot run yet -- an
# unimplemented method pair, or a starting tag that predates the surface
# under test (pick-releases annotates each tag with what its tree ships,
# e.g. whether the desktop app exists yet). Capability knowledge lives
# next to each driver, never here and never in the generator: declaring a
# method is a spec edit, implementing one is flipping the run workflow's
# gate.
#
# The starting versions are chosen at runtime from the repo's release tags
# (scripts/sandbox/pick-release-tags.sh): newest, oldest, and a spread
# between. A hardcoded list would stop covering the newest release the day
# after it ships, and would pin an "oldest" that nobody still runs.
#
# Triggers:
#   * every 12 hours, so upstream drift (a new uv, a Node bump, a PyPI change)
#     surfaces on a schedule rather than in someone's review cycle;
#   * when a release tag is created -- the moment the set of versions users can
#     update FROM changes, and the moment a broken updater would strand them;
#   * manually, where you can pick the route and how many releases to sample.
#
# Deliberately NOT on pull_request: a leg takes ~11 minutes of real toolchain
# installation, and the matrix multiplies that. Updating is release-shaped work,
# so it is gated on releases and the clock instead.

on:
  workflow_dispatch:
    inputs:
      route:
        description: 'Which combinations to run. all = every OS; both/update/installer = the linux legs; windows-desktop = the windows legs; macos-desktop = the macos legs.'
        required: false
        type: choice
        default: all
        options: [all, both, update, installer, windows-desktop, macos-desktop]
      tag-count:
        description: 'How many release tags to sample (newest, oldest, and a spread between).'
        required: false
        type: string
        default: '3'
  schedule:
    # Every 12 hours, off the hour to avoid the top-of-hour runner crunch.
    - cron: '20 7,19 * * *'
  push:
    tags:
      # Release tags only: the repo also carries backup/* and one-off tags.
      - 'v[0-9]+.[0-9]+.[0-9]+'
      - 'v[0-9]+.[0-9]+.[0-9]+.[0-9]+'

permissions:
  contents: read

concurrency:
  group: install-e2e-${{ github.ref }}
  cancel-in-progress: true

jobs:
  # Which released versions do we test updating FROM? Resolved once,
  # annotated with what each tag's own tree supports, and shared by every
  # OS's matrix so all combos cover the same set.
  pick-releases:
    name: Pick release tags
    runs-on: ubuntu-latest
    timeout-minutes: 5
    outputs:
      tags: ${{ steps.pick.outputs.tags }}
    steps:
      # This job only reads tag names and trees, so take the cheap
      # checkout: no blobs (filter), no other files (sparse), but DO fetch
      # tags -- they are the whole input, and the default shallow checkout
      # has none.
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          filter: blob:none
          fetch-tags: true
          sparse-checkout: scripts/sandbox/pick-release-tags.sh
          sparse-checkout-cone-mode: false
      - id: pick
        run: |
          set -euo pipefail
          tags="$(scripts/sandbox/pick-release-tags.sh --count '${{ inputs.tag-count || 2 }}')"
          echo "Testing updates from: $tags"
          # Annotate each tag with what its own tree supports, so run
          # workflows can natively skip surfaces the starting version does
          # not have. Today: does the release ship the desktop app
          # (apps/desktop, #20059)? Cheaper here -- the tags are already
          # fetched -- than a probe job per leg.
          enriched="$(for t in $(echo "$tags" | jq -r '.[]'); do
            if git ls-tree -d "$t" apps/desktop | grep -q .; then d=true; else d=false; fi
            echo "{\"ref\":\"$t\",\"desktop\":$d}"
          done | jq -sc .)"
          echo "Annotated: $enriched"
          echo "tags=$enriched" >> "$GITHUB_OUTPUT"

  # Expand the support matrix against the picked tags: one leg per
  # {os, install-method, update-method, tag}, split into a matrix per OS.
  generate-matrix:
    name: Expand combinations
    needs: pick-releases
    runs-on: ubuntu-latest
    timeout-minutes: 5
    outputs:
      linux: ${{ steps.gen.outputs.linux }}
      windows: ${{ steps.gen.outputs.windows }}
      macos: ${{ steps.gen.outputs.macos }}
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          sparse-checkout: scripts/sandbox/generate-e2e-matrix.mjs
          sparse-checkout-cone-mode: false
      - id: gen
        run: |
          set -euo pipefail
          matrices="$(node scripts/sandbox/generate-e2e-matrix.mjs \
            --tags '${{ needs.pick-releases.outputs.tags }}')"
          echo "$matrices"
          for key in linux windows macos; do
            echo "$key=$(echo "$matrices" | node -e 'let d="";process.stdin.on("data",c=>d+=c).on("end",()=>console.log(JSON.stringify(JSON.parse(d)[process.argv[1]])))' "$key")" >> "$GITHUB_OUTPUT"
          done
          # The plan, human-readable: a combination x starting-tag chart on
          # the run's summary page.
          node scripts/sandbox/generate-e2e-matrix.mjs \
            --tags '${{ needs.pick-releases.outputs.tags }}' \
            --format markdown >> "$GITHUB_STEP_SUMMARY"

  linux:
    name: ${{ matrix.name }}
    # The update/installer route choices map to the linux update methods;
    # either way the whole linux matrix runs (legs are cheap and the
    # distinction wasn't worth a filter layer in the generator).
    if: github.event_name != 'workflow_dispatch' || contains(fromJSON('["all", "both", "update", "installer"]'), inputs.route)
    needs: generate-matrix
    strategy:
      # One leg breaking is worth knowing about even if another already
      # failed, so let every leg report.
      fail-fast: false
      matrix: ${{ fromJSON(needs.generate-matrix.outputs.linux) }}
    uses: ./.github/workflows/install-e2e-run.yml
    with:
      install-method: ${{ matrix.install_method }}
      update-method: ${{ matrix.update_method }}
      install-ref: ${{ matrix.install_ref }}
      tag-has-desktop: ${{ matrix.tag_has_desktop }}
      leg-id: ${{ matrix.leg_id }}

  windows:
    name: ${{ matrix.name }}
    if: github.event_name != 'workflow_dispatch' || contains(fromJSON('["all", "windows-desktop"]'), inputs.route)
    needs: generate-matrix
    strategy:
      fail-fast: false
      matrix: ${{ fromJSON(needs.generate-matrix.outputs.windows) }}
    uses: ./.github/workflows/install-e2e-windows-run.yml
    with:
      install-method: ${{ matrix.install_method }}
      update-method: ${{ matrix.update_method }}
      install-ref: ${{ matrix.install_ref }}
      tag-has-desktop: ${{ matrix.tag_has_desktop }}
      leg-id: ${{ matrix.leg_id }}

  macos:
    name: ${{ matrix.name }}
    if: github.event_name != 'workflow_dispatch' || contains(fromJSON('["all", "macos-desktop"]'), inputs.route)
    needs: generate-matrix
    strategy:
      fail-fast: false
      matrix: ${{ fromJSON(needs.generate-matrix.outputs.macos) }}
    # Two driver arms: the OS-agnostic script driver (shared with linux)
    # and the published-dmg GUI driver; the run workflow routes.
    uses: ./.github/workflows/install-e2e-macos-run.yml
    with:
      install-method: ${{ matrix.install_method }}
      update-method: ${{ matrix.update_method }}
      install-ref: ${{ matrix.install_ref }}
      tag-has-desktop: ${{ matrix.tag_has_desktop }}
      leg-id: ${{ matrix.leg_id }}

  # The leg player: one static HTML for the whole run. Uploaded BEFORE the
  # matrix legs so it exists even when every leg dies; the report job links
  # every ran leg to it with that leg's logs zip as a #zip= hash param
  # (hash survives the artifact URL's server-side redirect, the query does
  # not). archive: false makes GitHub name the artifact after the FILE
  # (playback.html), ignoring the name: input -- harmless, the renderer
  # looks it up by that name.
  leg-player:
    name: Upload leg player
    runs-on: ubuntu-latest
    timeout-minutes: 5
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          sparse-checkout: tests/install/e2e-assets/playback.html
          sparse-checkout-cone-mode: false
      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: install-e2e-player
          path: tests/install/e2e-assets/playback.html
          archive: false
          retention-days: 14
          if-no-files-found: error

  # The outcome, human-readable: the plan chart again, with each cell
  # replaced by how that leg actually concluded. Per-leg conclusions are
  # NOT reachable through `needs` (a matrix job's result collapses to one
  # aggregate), so the table body comes from the run's own job list; the
  # `needs` results only sequence this job after every leg and provide
  # the per-OS aggregates.
  report:
    name: Result chart
    if: always()
    needs: [leg-player, pick-releases, linux, windows, macos]
    runs-on: ubuntu-latest
    timeout-minutes: 5
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          sparse-checkout: scripts/sandbox/generate-e2e-matrix.mjs
          sparse-checkout-cone-mode: false
      - env:
          GH_TOKEN: ${{ github.token }}
        run: |
          set -euo pipefail
          {
            echo "OS jobs: linux ${{ needs.linux.result }}, windows ${{ needs.windows.result }}, macos ${{ needs.macos.result }}"
            echo
            # The tag annotations let the chart say WHY a cell skipped
            # (pre-desktop vs declared TODO) instead of a flat "skip".
            gh api "repos/${{ github.repository }}/actions/runs/${{ github.run_id }}/jobs?per_page=100" \
              --paginate --jq '.jobs[] | {name, conclusion}' > /tmp/e2e-jobs.ndjson
            gh api "repos/${{ github.repository }}/actions/runs/${{ github.run_id }}/artifacts?per_page=100" \
              --paginate --jq '.artifacts[] | {name, id}' > /tmp/e2e-artifacts.ndjson
            echo 'Legend: ✅ ran green · ❌ ran red · pre-desktop / TODO = why a leg skipped · 📼 opens the leg player (recording + synced logs)'
            echo
            node scripts/sandbox/generate-e2e-matrix.mjs --format results \
              --tags '${{ needs.pick-releases.outputs.tags }}' \
              --artifacts /tmp/e2e-artifacts.ndjson < /tmp/e2e-jobs.ndjson
          } >> "$GITHUB_STEP_SUMMARY"