Makes Ollama a first-class part of the experience instead of a silent dependency.
- Startup gate: the pad waits until Ollama and the embedding model are ready, so notes are never saved unindexable. Reads 'blurt needs Ollama to think', with a one-click install link.
- Resilient after that: if Ollama stops mid-session, you keep typing (capture + exact search work) and the peek resumes when it's back.
- Self-heal: blurt pulls the model if Ollama appears without it, and re-indexes any notes saved while it was down, no restart needed.
- Settings now shows live engine status (Ollama, the embedding model, catch-up indexing).
- Earlier fix included: exact search no longer breaks when Ollama is down.
Update: Settings > check for updates, or pipx upgrade blurt.