ff97a372ac
audit-repo-health.md: - All 9 steps now read from repos.txt dynamically - Repos without relevant files (package.json, .dockerignore, etc.) are skipped - No more manual maintenance when repos are added/removed verify-all-backends.md: - Remove duplicate learning_ai_notes entry - Add learning_ai_efforise backend - Add learning_ai_efforise client + learning_ai_local_llms dashboard to web checks gitea-ci.md: - Steps 3+4 read from repos.txt, skip repos without gitea remote - Handle oss/ subdirectory repos via basename for Gitea API
3.2 KiB
3.2 KiB
| description |
|---|
| Start Gitea local CI, check status, or push all repos to trigger CI runs |
Gitea Local CI Workflow
Manages the self-hosted Gitea + act_runner CI infrastructure for all ByteLyst repos.
See docs/devops/GITEA_LOCAL_CI.md for full documentation.
1. Ensure Gitea and act_runner are running
// turbo
brew services start gitea && brew services start act_runner && sleep 2 && brew services list | grep -E "gitea|act_runner"
2. Verify Gitea is responding
// turbo
curl -s http://localhost:3300/api/v1/version | python3 -c "import sys,json; print('Gitea', json.load(sys.stdin)['version'])"
3. Push all workspace repos to Gitea to trigger CI
This pushes main to the gitea remote for every repo that has one configured.
Each push triggers the .gitea/workflows/ci.yml workflow on the local runner.
REPOS_DIR="/Users/sd9235/code/mygh"
while IFS= read -r repo; do
[[ -z "$repo" || "$repo" =~ ^# ]] && continue
cd "$REPOS_DIR/$repo" 2>/dev/null || continue
# Skip repos without a gitea remote
git remote get-url gitea &>/dev/null || continue
echo "=== $repo ==="
git push gitea main 2>&1 | tail -2
done < /Users/sd9235/code/mygh/learning_ai_common_plat/.windsurf/workflows/repos.txt
4. Wait for jobs to process, then check results
Wait ~2 minutes per repo for the runner (capacity=1) to process the queue, then check results.
REPOS_DIR="/Users/sd9235/code/mygh"
while IFS= read -r repo; do
[[ -z "$repo" || "$repo" =~ ^# ]] && continue
cd "$REPOS_DIR/$repo" 2>/dev/null || continue
git remote get-url gitea &>/dev/null || continue
# Use basename for Gitea API (oss/learning_ai_claw-cowork → learning_ai_claw-cowork)
repo_name=$(basename "$repo")
echo "=== $repo ==="
curl -s -u "bytelyst:bytelyst123" "http://localhost:3300/api/v1/repos/bytelyst/$repo_name/actions/jobs" | python3 -c "
import sys, json
jobs = json.load(sys.stdin).get('jobs', [])
if not jobs:
print(' (no jobs)')
else:
max_run = max(j['run_id'] for j in jobs)
for j in jobs:
if j['run_id'] == max_run:
c = j.get('conclusion','pending')
icon = '✅' if c == 'success' else '❌' if c == 'failure' else '⏳'
print(f' {icon} {c:12} {j[\"name\"]}')
" 2>/dev/null
done < /Users/sd9235/code/mygh/learning_ai_common_plat/.windsurf/workflows/repos.txt
5. (Optional) View logs for a failing job
Replace REPO and JOB_ID with the repo name and job number from step 4.
REPO="learning_ai_clock"
JOB_ID="76"
curl -s -u "bytelyst:bytelyst123" "http://localhost:3300/api/v1/repos/bytelyst/$REPO/actions/jobs/$JOB_ID/logs" | tail -30
6. (Optional) Stop Gitea services
brew services stop act_runner && brew services stop gitea
Troubleshooting
- Runner not picking up jobs:
brew services restart act_runner - Runner still queued after restart: re-register against
http://127.0.0.1:3300instead ofhttp://localhost:3300 - Stale .next/lock:
rm -f /Users/sd9235/code/mygh/learning_ai_common_plat/dashboards/*-web/.next/lock - Permission denied on tsc:
chmod +x /Users/sd9235/code/mygh/learning_ai_common_plat/node_modules/.bin/* - Check runner log:
tail -30 /opt/homebrew/var/log/act_runner.err