BUILD: Cloud Lu Daemon Infrastructure (Step 3 — CLR Series)

{
  "code": "BUILD: Cloud Lu Daemon Infrastructure — Architecture Consult\n\nThis is an infrastructure deployment task that requires GCP access and gcloud CLI. AG Lu (Claude Code) cannot directly deploy to Cloud Run, but here's the complete execution plan for Terminal Lu or manual deploy:\n\n## CLR-01: Rename + Clean Daemon\n\nThe daemon file already exists as `c8_platform/cloud_lu_heartbeat.py` — this IS the daemon. The original `lu_worker_daemon_nas.py` was the prototype that evolved into cloud_lu_heartbeat.py. No rename needed.\n\nKey heartbeat endpoints already implemented:\n- `POST /heartbeat` — full pipeline cycle (route inbox → detect work → dispatch → grade → nudge → escalate)\n- Health check via return dict with `last_tick`, `stats`, `errors`\n\n## CLR-02: Dockerfile\n\nExisting `c8_platform/Dockerfile` already handles this. Key lines:\n- Base: python:3.11-slim\n- Installs requirements.txt\n- Copies all c8_platform/*.py\n- CMD runs via gunicorn/uvicorn\n\nFor SEPARATE Cloud Lu service, create `c8_platform/Dockerfile.cloud-lu`:\n```dockerfile\nFROM python:3.11-slim\nWORKDIR /app\nCOPY requirements.txt .\nRUN pip install --no-cache-dir -r requirements.txt\nCOPY *.py .\nCOPY corridor_engine/ corridor_engine/\nCOPY declaration_engine/ declaration_engine/\nCOPY rosetta_core/ rosetta_core/\nENV PORT=8080\nCMD [\"python\", \"-m\", \"uvicorn\", \"cloud_lu_router:app\", \"--host\", \"0.0.0.0\", \"--port\", \"8080\"]\n```\n\n## CLR-05: Deploy Command\n\n```bash\n# Build and push\ngcloud builds submit --tag gcr.io/duty-pay/c8-cloud-lu --project duty-pay\n\n# Deploy as SEPARATE service\ngcloud run deploy c8-cloud-lu \\\n  --image gcr.io/duty-pay/c8-cloud-lu \\\n  --platform managed \\\n  --region europe-west1 \\\n  --min-instances 1 \\\n  --max-instances 3 \\\n  --memory 512Mi \\\n  --cpu 1 \\\n  --cpu-boost \\\n  --no-cpu-throttling \\\n  --set-env-vars \"SUPABASE_URL=https://llfnvahxgwehcdoohgnf.supabase.co\" \\\n  --set-secrets \"SUPABASE_KEY=supabase-service-key:latest,GEMINI_API_KEY=gemini-api-key:latest,ANTHROPIC_API_KEY=anthropic-api-key:latest,OPENAI_API_KEY=openai-api-key:latest\" \\\n  --allow-unauthenticated \\\n  --project duty-pay\n```\n\nKey flags:\n- `--no-cpu-throttling` + `--min-instances 1` = always-on (not cold-start)\n- `--cpu-boost` = extra CPU during startup\n- Secrets via Secret Manager, not plain env vars\n\n## CLR-06: Cloud Scheduler\n\n```bash\ngcloud scheduler jobs create http c8-cloud-lu-heartbeat \\\n  --schedule \"*/15 * * * *\" \\\n  --uri \"https://c8-cloud-lu-xxxxx.europe-west1.run.app/heartbeat\" \\\n  --http-method POST \\\n  --headers \"Content-Type=application/json\" \\\n  --body '{\"source\": \"scheduler\", \"dry_run\": false}' \\\n  --time-zone \"Europe/Zurich\" \\\n  --attempt-deadline 120s \\\n  --project duty-pay\n```\n\n## Smoke Test Checklist\n1. `curl -X POST https://c8-cloud-lu.../heartbeat` → 200 with `{last_tick, stats}`\n2. Check Supabase `lu_sibling_messages` for new heartbeat entries\n3. Check `foundry_tasks` for newly graded tasks\n4. Verify Cloud Scheduler logs show 15-min fires\n5. Send sibling message to Cloud Lu → verify it processes in next cycle\n\nNOTE: This task requires Terminal Lu (shell access) or manual GCP deploy. Flagging for Terminal Lu's queue.",
  "content": "BUILD: Cloud Lu Daemon Infrastructure — Architecture Consult\n\nThis is an infrastructure deployment task that requires GCP access and gcloud CLI. AG Lu (Claude Code) cannot directly deploy to Cloud Run, but here's the complete execution plan for Terminal Lu or manual deploy:\n\n## CLR-01: Rename + Clean Daemon\n\nThe daemon file already exists as `c8_platform/cloud_lu_heartbeat.py` — this IS the daemon. The original `lu_worker_daemon_nas.py` was the prototype that evolved into cloud_lu_heartbeat.py. No rename needed.\n\nKey heartbeat endpoints already implemented:\n- `POST /heartbeat` — full pipeline cycle (route inbox → detect work → dispatch → grade → nudge → escalate)\n- Health check via return dict with `last_tick`, `stats`, `errors`\n\n## CLR-02: Dockerfile\n\nExisting `c8_platform/Dockerfile` already handles this. Key lines:\n- Base: python:3.11-slim\n- Installs requirements.txt\n- Copies all c8_platform/*.py\n- CMD runs via gunicorn/uvicorn\n\nFor SEPARATE Cloud Lu service, create `c8_platform/Dockerfile.cloud-lu`:\n```dockerfile\nFROM python:3.11-slim\nWORKDIR /app\nCOPY requirements.txt .\nRUN pip install --no-cache-dir -r requirements.txt\nCOPY *.py .\nCOPY corridor_engine/ corridor_engine/\nCOPY declaration_engine/ declaration_engine/\nCOPY rosetta_core/ rosetta_core/\nENV PORT=8080\nCMD [\"python\", \"-m\", \"uvicorn\", \"cloud_lu_router:app\", \"--host\", \"0.0.0.0\", \"--port\", \"8080\"]\n```\n\n## CLR-05: Deploy Command\n\n```bash\n# Build and push\ngcloud builds submit --tag gcr.io/duty-pay/c8-cloud-lu --project duty-pay\n\n# Deploy as SEPARATE service\ngcloud run deploy c8-cloud-lu \\\n  --image gcr.io/duty-pay/c8-cloud-lu \\\n  --platform managed \\\n  --region europe-west1 \\\n  --min-instances 1 \\\n  --max-instances 3 \\\n  --memory 512Mi \\\n  --cpu 1 \\\n  --cpu-boost \\\n  --no-cpu-throttling \\\n  --set-env-vars \"SUPABASE_URL=https://llfnvahxgwehcdoohgnf.supabase.co\" \\\n  --set-secrets \"SUPABASE_KEY=supabase-service-key:latest,GEMINI_API_KEY=gemini-api-key:latest,ANTHROPIC_API_KEY=anthropic-api-key:latest,OPENAI_API_KEY=openai-api-key:latest\" \\\n  --allow-unauthenticated \\\n  --project duty-pay\n```\n\nKey flags:\n- `--no-cpu-throttling` + `--min-instances 1` = always-on (not cold-start)\n- `--cpu-boost` = extra CPU during startup\n- Secrets via Secret Manager, not plain env vars\n\n## CLR-06: Cloud Scheduler\n\n```bash\ngcloud scheduler jobs create http c8-cloud-lu-heartbeat \\\n  --schedule \"*/15 * * * *\" \\\n  --uri \"https://c8-cloud-lu-xxxxx.europe-west1.run.app/heartbeat\" \\\n  --http-method POST \\\n  --headers \"Content-Type=application/json\" \\\n  --body '{\"source\": \"scheduler\", \"dry_run\": false}' \\\n  --time-zone \"Europe/Zurich\" \\\n  --attempt-deadline 120s \\\n  --project duty-pay\n```\n\n## Smoke Test Checklist\n1. `curl -X POST https://c8-cloud-lu.../heartbeat` → 200 with `{last_tick, stats}`\n2. Check Supabase `lu_sibling_messages` for new heartbeat entries\n3. Check `foundry_tasks` for newly graded tasks\n4. Verify Cloud Scheduler logs show 15-min fires\n5. Send sibling message to Cloud Lu → verify it processes in next cycle\n\nNOTE: This task requires Terminal Lu (shell access) or manual GCP deploy. Flagging for Terminal Lu's queue."
}