/deploy

Safe, repeatable deployment pipeline for Frood. Handles the full workflow: dev push → main merge → production deploy → health verification.

Pre-flight Checks (MANDATORY)

Before ANY git operations, run these checks and STOP if any fail:

1. Clean Working Tree

bash
1git status --short

If unstaged/staged changes exist: Ask the user:

"You have uncommitted changes in [list files]. Would you like to commit them first, or stash them for the deploy?"

Do NOT silently stash or proceed with dirty working tree. The user must explicitly choose.

2. Commits Ahead of Remote

bash
1git rev-list --count origin/dev..HEAD

If 0 commits ahead: Nothing to deploy. Tell the user and stop.

3. Test Suite

bash
1python -m pytest tests/ -x -q

If failures: Report them. Ask: "Tests have failures — deploy anyway?" Only known pre-existing failures (like test_auth_flow.py:156) can be safely ignored. New failures should block deployment.

4. Multi-Workstream Audit (if 20+ commits ahead)

When there are many accumulated commits, group them by workstream and verify each is in a complete state:

bash
1git log --oneline origin/dev..HEAD

For each workstream prefix in commit messages (e.g., [custom-claude-code-ui], [per-project-task-memories]):

• Check that the latest phase has a SUMMARY.md (plan complete)
• Check that the phase has a VERIFICATION.md with status: passed
• Flag any workstream with incomplete plans

Present a summary table:

If any workstream is incomplete: Warn the user and ask whether to proceed.

Deployment Steps

Only proceed after all pre-flight checks pass (or user explicitly approves warnings).

Step 1: Push dev

bash
1git push origin dev

Step 2: Merge to main

bash
1git checkout main
2git merge dev
3git push origin main
4git checkout dev

Step 3: Deploy to production

bash
1ssh frood-prod "cd ~/frood && git pull origin main && sudo systemctl restart frood"

Step 4: Verify (wait 5 seconds for service startup)

bash
1ssh frood-prod "sleep 5 && sudo systemctl is-active frood && curl -s -o /dev/null -w '%{http_code}' http://localhost:8000/api/health"

Expected: active + 200

Output

After deployment, show:

## Deployment Complete

| Step | Status |
|------|--------|
| Push dev | OK (N commits) |
| Merge to main | OK |
| Push main | OK |
| SSH deploy | OK |
| Service health | OK/FAIL |

**Commit range:** abc1234..def5678
**Files changed:** N files (+X/-Y lines)

If the health check fails, immediately run /prod-check for detailed diagnostics.

What NOT to Do

• Do NOT proceed with dirty working tree without user consent
• Do NOT silently stash changes — ask the user
• Do NOT skip the test suite check
• Do NOT force-push anything
• Do NOT attempt to fix production issues automatically — report and recommend