Add --publish flag for Render publish-only deployment

[IamJasonBian]

Summary

• Add --publish flag to decouple blob uploads from live trading, enabling Render deployments that update the site without placing trades
• Skip broker initialization in publish-only mode to allow headless operation
• Add render.yaml configuration for allocation-engine-2.0 Background Worker

Changes

• trading_system/main.py: Add --publish parameter, skip broker init when publish-only, decouple blob upload from --live flag
• render.yaml (new): Worker configuration with start command --publish --dashboard --continuous
• public/dashboard/ (new): Dashboard HTML and GIF assets

Blob uploads now gated by live=not self.dry_run or self.publish, allowing site updates without live trading.

[IamJasonBian]

Code Review — PR #52

Overall: Clean implementation of publish-only mode. The broker skip logic and null guards are well-placed.

Issues

1. render.yaml exposes ticker strategy in start command — The startCommand hard-codes --ticker BTC SPY QQQ AMZN. Consider using an env var (TICKER_LIST) to avoid redeploying just to change symbols.

2. Publish + live interaction is unclear — The condition not (self.publish and self.dry_run) means --publish --live still initializes the broker. The --publish flag description says "Upload blobs in dry-run mode" but nothing enforces mutual exclusivity with --live. Consider adding a guard:

   if args.publish and args.live:
       parser.error("--publish and --live are mutually exclusive")

3. portfolio_data can be None downstream — At line ~699, portfolio_data is set to None when self.trading_bot is None, but log_state_to_blob and _send_oncall_summary receive it. Verify those functions handle portfolio=None gracefully.

Nits

• The Slack deploy message correctly shows "PUBLISH" mode — nice touch.
• render.yaml env vars all have sync: false — confirm this is intentional (secrets won't auto-sync across Render services).

---

Generated by Claude Code

[IamJasonBian]

Follow-up Review — PR #52

Additional observations not covered in the initial review:

1. Broker-skip condition + render.yaml = latent bug — The condition not (self.publish and self.dry_run) requires both flags to skip broker init. But render.yaml's start command uses --publish without --dry-run, meaning the Render worker will still attempt to initialize the broker. Either the condition should be not self.publish (if publish is always headless), or render.yaml needs --dry-run added.

2. render.yaml declares unused RH credentials — RH_USERNAME, RH_PASSWORD, RH_TOTP_CODE are declared as env vars for a publish-only worker that shouldn't need broker access. This creates confusion about what secrets are actually required for this service.

3. Merge conflicts exist — The PR's mergeable state is dirty. Needs a rebase onto current main before it can be merged.

---

Generated by Claude Code

[IamJasonBian]

Stale PR notice — This PR has been open for 18 days with no new commits since Mar 24. Two code reviews flagged a latent bug (broker init not skipped without --dry-run), unused RH credentials in render.yaml, and merge conflicts with main. Please rebase and address review feedback or close if no longer needed.

---

Generated by Claude Code

[IamJasonBian]

Stale — 4 weeks, no activity since Mar 24. Two prior reviews flagged a latent bug (broker init not actually skipped without --dry-run), unused RH credentials in render.yaml, and merge conflicts with main. The mergeable state was already dirty as of Apr 4. Please rebase and fix the broker-skip condition, or close if the Render deployment approach has changed.

---

Generated by Claude Code