# Drive Monitor Refactor - Task Status

## Summary
Unify DriveMonitor file state tracking from JSON to PostgreSQL database table `drive_files`.

## Status: COMPLETED

---

## Completed ✅

### 1. Migration
- Created `migrations/6.3.1-01-drive-files/up.sql` with table schema
- Indexes: bot_id, file_type, indexed, fail_count
- Migration data from kb_documents for backwards compatibility

### 2. Schema  
- Created `src/core/shared/schema/drive.rs` with:
  - `DriveFile` (Queryable)
  - `NewDriveFile` (Insertable)  
  - `DriveFileUpdate` (AsChangeset)
  - diesel table! macro

### 3. Repository
- Created `src/drive/drive_files.rs` with `DriveFileRepository`
- Methods: get_file_state, upsert_file, mark_indexed, mark_failed, get_max_fail_count, get_files_to_index, delete_file, get_all_files_for_bot, get_files_by_type

---

## Working Environment

```bash
# Access production logs
ssh administrator@63.141.255.9
sudo incus exec system -- tail -f /opt/gbo/logs/err.log | grep -i "drive\|error"

# Check CI logs  
ssh administrator@63.141.255.9
sudo incus exec alm-ci -- tail -f /opt/gbo/logs/*.log

# Health endpoint
https://system.pragmatismo.com.br/api/health
```

---

## Testing Playbook

1. **Rendering**
   - Navigate: https://chat.pragmatismo.com.br/salesianos
   - Send "oi", verify HTML response

2. **Direct Tool Execution**
   - Click "cartas" button
   - Ask about cartas

3. **KB Injection**
   - Ask about cartas without clicking button
   - Verify KB influences response

4. **Parallel LLM**
   - Start conversation in one window
   - Verify start.bas doesn't lock other windows

5. **Config Settings**
   - Verify history-limit=3 in config.csv respected
   - Toggle Table switch button
   - Verify prompt injection

---

## Critical Rules

- Never stop - YOLO loop
- Push to ALM after each fix
- Check health endpoint after deploy
- Monitor err.log continuously
- Work only in ~/src/gb
- Push all submodules to ALM

---

## Database Access

```bash
# Get credentials from Vault
ssh administrator@63.141.255.9 "sudo incus exec system -- curl -s --cacert /opt/gbo/conf/system/certificates/ca/ca.crt -H 'X-Vault-Token: <SECRET>' https://10.157.134.250:8200/v1/secret/data/gbo/tables 2>/dev/null"
```

---

## Container Layout
- `system` - BotServer (port 8080)
- `tables` - PostgreSQL (port 5432)  
- `vault` - Vault (port 8200)
- `directory` - Zitadel (port 8080/9000)
- `drive` - MinIO (port 9100)
- `cache` - Valkey (port 6379)
- `llm` - llama.cpp (port 8081)
- `vector_db` - Qdrant (port 6333)
- `alm-ci` - CI runner

---

## Current Status

- DriveMonitor refatorado para usar database (JSON removido)
- FileState struct removido
- Métodos load/save_file_states removidos
- Repository pattern implementado
- Compilação verificada: OK

---

## Next Steps

1. Push para ALM todos os submódulos
2. Aguardar CI build
3. Testar com Playwright
4. Verificar health endpoint
5. Monitorar err.log
6. Reportar status e continuar YOLO

fala da carta de 2021


FINLA: