# Cron Job Model Fix (Post-Update)

**Session:** 2026-05-08 | **Impact:** All cron jobs with `model: null` break after Hermes Agent update

## Problem

After a Hermes Agent update, cron jobs with `model: null` fail immediately:

```
Error code: 400 - {'message': 'model should be in provider/model format'}
```

The new codepath strictly validates the `model` field — it no longer auto-resolves from config.

## Fix Pattern

### Step 1: Identify affected jobs
```bash
hermes cron list
```
Look for jobs where `model: null`.

### Step 2: Apply explicit model
```bash
hermes cron update <job-id> --model '{"model": "<provider>/<model>"}'
```

Example for this setup:
```bash
hermes cron update <job-id> --model '{"model": "custom/llama/model.gguf"}'
```

### Step 3: Verify
```bash
hermes cron list --name <job-name>
```
Confirm `model` is now `custom/llama/model.gguf`.

### Step 4: Restart gateway (if running)
```bash
systemctl --user restart hermes-gateway
```
The gateway process caches job definitions in memory. Restart ensures it loads updated models.

## Prevention

When creating new cron jobs, **always** specify the model explicitly:
```bash
hermes cron create --name "..." --model '{"model": "custom/llama/model.gguf"}' --prompt "..."
```

Or for script-only jobs (no LLM needed), set `--no-agent`:
```bash
hermes cron create --name "..." --script "..." --no-agent
```

## Session-Specifics (2026-05-08)

Three jobs were affected by the update to v0.13.0 (commit 474d1e812):
- `SRE Network Reconnaissance` (4733a436d99a) → fixed
- `Rechnungen-Organizer` (f773f8c23230) → fixed
- `memory-sync-daily` (ce496a8d1062) → deleted and recreated as `no_agent: true` script-only