# Quick Start — Use the Fork TODAY

## 1. Install Extension

```bash
pi install npm:@kylebrodeur/pi-model-router
```

## 2. Initialize Config

```
/router init
```

This scaffolds a default configuration at `~/.pi/agent/model-router.json`.

## 3. Use in Pi

In Pi, run:

```
/reload
```

You should see in console:

```
[router] Feature sync complete - ollama: true rate-limit: true
[router] ollama-sync: enabled
[router] rate-limit-fallback: enabled
```

## 4. Verify Features

### Auto-Sync Test

```bash
# In another terminal
ollama pull llama3.2:3b
```

```
# In Pi
/router ollama-sync
```

Expected: `Added: llama3.2:3b`

### Fallback Test

```
# In Pi — switch to a cloud model first
/model anthropic/claude-sonnet-4

# Then trigger fallback
/router fallback
```

Expected: switches to best available model matching your `fallbackSequence` (e.g., an Ollama model)

```
# Restore
/router restore
```

## 5. New Commands

| Command               | What it does                               |
| --------------------- | ------------------------------------------ |
| `/router ollama-sync` | Manually sync Ollama models                |
| `/router fallback`    | Switch to fallback model sequence (manual) |
| `/router restore`     | Restore original cloud model               |
| `/router init`        | Scaffold default config file               |
| `/router scope apply` | Sync router profiles to Pi enabled models  |
| `/router scope reset` | Clear router profiles from Pi enabled list |
| `/router scope show`  | Show current Pi scoped models settings     |

All existing `/router *` commands still work unchanged.

## 6. Daily Workflow

```bash
# Pull new Ollama model
ollama pull qwen3.5:7b
```

```
# In Pi — auto-sync triggers on session start
/new
# → [Router] Added 1 model(s): qwen3.5:7b
# → Run /reload to see: qwen3.5:7b

/reload
# → Model now available in /model
```

## Troubleshooting

**"Ollama not available"**
→ Check `ollama list` works in terminal

**Models not in registry**
→ `/reload` — extension reload refreshes model registry

**TypeScript errors**
→ Requires Pi 0.67+ for rate-limit fallback (uses `after_provider_response` event)
→ Ollama sync works with current Pi version

**Config not applying**
→ Verify `~/.pi/agent/model-router.json` exists
→ Check with `cat ~/.pi/agent/model-router.json`