# JARVIS Silent Auto-Fix System

**Date:** 2026-02-17
**Status:** ✅ ACTIVE

## Problem

User getting spammed with notifications about issues instead of automatic fixes.

## Solution

**Silent Auto-Fix System:**

- Auto-fix problems WITHOUT notifications
- Only notify on CRITICAL failures (cannot auto-fix)

## Changes Applied

### 1. Health Check → Auto-Fix Engine

- **Before:** Check and log problems
- **After:** Check, auto-fix, only notify if fix FAILS

### 2. Cron Jobs → Silent Mode

- All output redirected to `/dev/null`
- Only critical failures logged
- Success = silence

### 3. Daily Summary

- Runs at 20:00
- Only sends notification if critical issues found
- Silent on success

### 4. On-Demand Status

- New command: `jarvis-status-quiet.sh`
- Shows auto-fixes and critical issues
- No spam, only when you ask

## Auto-Fix Capabilities

✅ **Can Auto-Fix:**

- Gateway down (restart service)
- Kanban down (restart service)
- Any systemd service failure

❌ **Cannot Auto-Fix (Alerts):**

- MariaDB Synology unreachable (external)
- Network down
- Hardware failure

## Logs

| File                      | Purpose                      |
| ------------------------- | ---------------------------- |
| `jarvis-silent-fixes.log` | Successful auto-repairs      |
| `jarvis-critical.log`     | Failed fixes (manual needed) |
| `jarvis-delegation.log`   | Agent task delegation        |
| `github-import.log`       | GitHub scanner activity      |

## Cron Schedule

```
Every hour:    Health check + auto-fix (silent)
Every 2 hours: Agent delegation (logged)
Daily 16:00:   GitHub scanner (logged)
Daily 20:00:   Summary (only if critical)
```

## Usage

```bash
# Check status (on-demand, no spam)
/home/alex/clawd/tools/jarvis-status-quiet.sh

# View silent fixes
tail /home/alex/clawd/logs/jarvis-silent-fixes.log

# View critical issues
tail /home/alex/clawd/logs/jarvis-critical.log

# View delegation activity
tail /home/alex/clawd/logs/jarvis-delegation.log
```

## Example Day

**08:00:** Gateway crashes → Auto-restarted silently ✅
**10:00:** Kanban hangs → Auto-restarted silently ✅
**14:00:** All services healthy ✅
**20:00:** Daily summary: "No critical issues" (silent) ✅

**User sees:** Nothing! (everything auto-fixed)

## Example Critical Failure

**12:00:** Gateway crashes
**12:00:** Auto-fix attempts restart
**12:00:** Restart FAILS (disk full, etc.)
**12:00:** Logged to critical.log ❌
**20:00:** Daily summary: "🚨 1 critical issue"

**User sees:** Notification to check critical issue

## Philosophy

> "The best notification is no notification"

- Fix what can be fixed automatically
- Only bother the human when necessary
- Provide visibility on-demand

## Verification

✅ Cron jobs silent
✅ Health check auto-fixes
✅ Critical-only notifications
✅ On-demand status available
✅ No spam

## No More Spam! 🎉

You'll only hear from JARVIS when:

1. Something CANNOT be auto-fixed
2. You explicitly run status check

Otherwise: **Silence = Success**