# Switching Guide

Use this when the user wants to swap between hosted and local models.

## Commands

- Hosted ChatGPT-backed profile:
  ```bash
  chatgpt-mini
  ```

- Local Ollama-backed profile:
  ```bash
  local-qwen
  ```

## When to choose which

- Use **local-qwen** for:
  - coding work
  - shell automation
  - privacy-sensitive local tasks
  - lightweight daily use

- Use **chatgpt-mini** for:
  - higher-quality reasoning
  - tasks where the local model struggles
  - broader general conversation quality

## Rules

- Prefer explicit switching over mutating the default profile.
- Keep both profiles available.
- If the user asks for a "vrai switch", treat it as a request to enable both wrappers and a clear selection path.

## Verification

When switching to local:

1. Confirm Ollama is running.
2. Confirm `qwen3-coder:30b-a3b-q4_K_M` is present in `ollama list`.
3. Confirm the Hermes profile `local-ollama` is configured.
4. Run a small smoke test before trusting the model for real work.