Can I vibecode GitHub Copilot?

Build me a local-first coding assistant setup to replace GitHub Copilot. Requirements:

- Do not build or train a model. Set up the open-source Continue extension in
  VS Code, configured against models I control.
- Chat and edit model: Claude or GPT over API, key referenced from my
  environment, never committed. Autocomplete model: a small local coder model
  served by Ollama, so completions are free and work offline.
- Deliver the actual Continue config file with both models wired, plus
  sensible keybindings for inline edit and chat.
- Enable Continue's codebase indexing for repo-aware answers; document where
  the index lives and how to rebuild it.
- Add scripts/ai-review.sh: pipes `git diff` to the API model (key from .env)
  and prints a short pre-commit review.
- Everything local except the chat-model API calls; turn off the extension's
  telemetry in the config.
- Out of scope: writing an editor extension from scratch, GitHub PR-bot
  features, team seat management.
- README: Ollama install and model pull commands, where keys go, and an honest
  note that local autocomplete is slower and dumber than Copilot's, that is
  the trade for free and private.