fix: ensure first-time setup token reaches docker logs (#858)

[DamageShadow]

Summary

• Add ENV PYTHONUNBUFFERED=1 to docker/backend/Dockerfile so lifespan print() output is not block-buffered on the Docker log pipe
• Switch the first-time setup token block in main.py from print() to logger.warning() so the token flushes reliably and appears in structured logs / Vector

Fixes #858

Root cause

Without PYTHONUNBUFFERED=1, Python block-buffers stdout (~8KB) when connected to a Docker pipe (not a TTY). The setup token and ~30 other lifespan print() calls never reached docker compose logs backend, blocking fresh installs.

Test plan

• Rebuilt backend image — docker exec trinity-backend env | grep PYTHONUNBUFFERED returns PYTHONUNBUFFERED=1
• With setup_completed=false, restarted backend — docker compose logs backend | grep "Setup token:" returns the token as a JSON WARNING log
• Lifespan status prints (e.g. Using dedicated scheduler service, Log archive service started) now visible in docker logs
• CI passes

Made with Cursor

[vybe]

Thanks for fixing this! 🙏 One process note before it can be merged:

Per Trinity's SDLC (CLAUDE.md → SDLC, and .claude/DEVELOPMENT_WORKFLOW.md), contributor PRs land on dev first — work flows feature → dev → main, and main only ever receives release PRs cut from dev. This PR is currently based on main.

Could you retarget the base branch from main to dev? You can do it in place: click Edit next to the PR title and change the base dropdown to dev — no need to close/reopen. (You may need to rebase onto dev afterward.)

Heads-up: #1168 also addresses #858 (already approved, targeting dev) — worth a look to make sure these don't collide.

Thanks again!

[github-actions]

⚠️ Nightly unit-suite check skipped — merge conflict against dev.

Resolve by running git merge dev locally and pushing the result. The next nightly run will re-test once the conflict is gone.