# Model Switching (Paid vs Free)

This setup adds two model profiles:

- `paid` -> `openrouter/moonshotai/kimi-k2.5` with OpenRouter fallbacks
- `free` -> `openrouter/qwen/qwen3-coder:free` with OpenRouter fallback

## Recommended workflow (prevents legacy drift)

```bash
cd /Volumes/Data/openclaw-setups/openclaw-setup-max
bash ./scripts/model_hygiene_workflow.sh paid
```

This command:
- Applies the selected profile
- Restages/restarts guard LaunchAgents
- Prunes legacy models/providers/auth entries not referenced by current config

## On-demand switch (immediate)

```bash
cd /Volumes/Data/openclaw-setups/openclaw-setup-max
bash ./scripts/model_profile_switch.sh free
```

Switch back:

```bash
bash ./scripts/model_profile_switch.sh paid
```

Notes:
- This updates OpenClaw default + fallbacks.
- It also pushes `/model ...` to the active `agent:main:main` session by default.

## Show status and available profiles

```bash
bash ./scripts/model_profile_switch.sh status
```

## Schedule auto-switching (example: free from 10pm-7am)

1. Edit config:

```bash
cd /Volumes/Data/openclaw-setups/openclaw-setup-max
open config/model-schedule.config.json
```

Ensure:

```json
{
  "enabled": true,
  "dayProfile": "paid",
  "nightProfile": "free",
  "dayStartHour": 7,
  "nightStartHour": 22
}
```

2. Install schedule launchd job:

```bash
bash ./scripts/install_model_schedule_guard_launchd.sh
```

Note:
- Installer stages runtime files in `~/Library/Application Support/openclaw-local-model-guard`.
- If you edit configs/scripts in this repo, rerun installer scripts to sync stage files.

3. Verify:

```bash
launchctl print gui/$(id -u)/ai.openclaw.local.model-schedule-guard
tail -n 50 /tmp/openclaw-model-schedule-guard.log /tmp/openclaw-model-schedule-guard.err.log
```

If schedule is enabled, it will override manual model switches on the next run.

## Budget guard (optional but recommended)

This warns when high-cost models are left on and can auto-revert.

```bash
bash ./scripts/install_model_budget_guard_launchd.sh
```

Or install both schedule + budget:

```bash
bash ./scripts/install_local_model_guardrails.sh
```

## Disable / remove launchd jobs

```bash
launchctl bootout gui/$(id -u)/ai.openclaw.local.model-schedule-guard 2>/dev/null || true
launchctl bootout gui/$(id -u)/ai.openclaw.local.model-budget-guard 2>/dev/null || true
```

## After OpenClaw upgrades

Use the update workflow so guards/profile state are reapplied automatically:

```bash
cd /Volumes/Data/openclaw-setups/openclaw-setup-max
bash ./scripts/update_openclaw.sh
```

The update script reinstalls both guard LaunchAgents and kickstarts them immediately, so budget/profile logic is re-evaluated in the same run.