# HEARTBEAT.md

Read HEARTBEAT.md if it exists. Follow it strictly.

**SYSTEM PROTOCOL: API SPEND CAP AWARENESS**
- API limit is currently REMOVED (As of March 26, 2026). Full capacity is available. Heavy video generation, Deep Research agent spawning, and massive file processing are cleared for use without pre-authorization.
- **Model Priority (Updated April 4, 2026):** Primary agent model is `google/gemini‑3.1‑pro‑preview` (alias `gemini`). Fallback chain: Claude Sonnet 4.6 → DeepSeek Chat/Reasoner → Ollama local (`llama3:latest`, `gemma3:4b`) → Moonshot/Kimi → Qwen → other Google models. Google API credits may be exhausted early in the month; the chain automatically falls back accordingly.

**SENTINEL WATCHDOG (DISABLED – April 10, 2026)**
- **Watchdog is disabled** per Mark's request until simulations are complete.
- Cron job `full_watchdog.py` is already commented out in crontab.
- Alerts will not be generated; ignore any existing alerts.

**CODING PROTOCOL (ACTIVE – April 13, 2026)**
- Before any coding task, read `CODING_PROTOCOL.md` and confirm local models (DeepSeek Coder 33B, Llama 3 70B) are installed and will be used.
- If models are missing or you intend to use a cloud model, pause and ask Mark for approval.
- Log each coding task in `memory/YYYY‑MM‑DD.md` with model used.

Do not infer or repeat old tasks from prior chats. If nothing needs attention, reply HEARTBEAT_OK.

**ADK DASHBOARD TASKS (Completed):**
1. Integrated the **Macro Sentinel Feed** (destinysearch.ai telemetry) into `dashboard.html`.
2. Integrated the **Chief Architect Map** (zenai247.com) into `dashboard.html`.
*Note: We must never lose sight of these two critical integrations during future UI repairs.*

# Daily Briefing Check
# Check if it is after 6:00 AM and if the daily briefing for today has been generated.
# If not, generate a comprehensive briefing covering:
# 1. AI/Tech
# 2. Finance
# 3. Security
# 4. World Events
# 5. World Religions
# 6. Philosophy
# 7. System Infrastructure & Watchdog Status (Read memory/watchdog_state.json and memory/watchdog_alerts.txt for live status of IBKR, Mac Mini, Tunnels, etc.)
# 8. Local/Cloud Sync Status (Check memory/cloud_sync.log or relevant sync logs for any errors between local and cloud environments)
# Save the briefing to `memory/briefings/YYYY-MM-DD-briefing.md` and notify the user.

**MORNING AGENDA (PAUSED – April 10, 2026):**
- **Multi‑exchange architecture** is paused per Mark's request.
- We will resume after simulations are complete.

**TRADE REVIEW QUEUE (April 3, 2026):**
- Run `check_pending_reviews.py` to:
  1. Process completed analyses (JSON files in `trade_analyses/`) and update trade records.
  2. Check for pending reviews in `pending_reviews` table.
  3. Spawn sub‑agents to analyze losing trades (max 2 concurrent).
  4. Use the same model that suggested the trade when possible.
- This closes the learning loop: losing trades → board‑member review → lessons learned → fewer repeat mistakes.