Can I vibecode Granola?

Add full-text search to the meeting-notes app in this folder. Requirements:

- Index every note in ~/MeetingNotes (summary and transcript) with SQLite FTS5,
  in whatever language the app is already written in.
- Re-index a note when its .md file changes, and pick up new files automatically.
- A local search page on http://localhost:4321: one search box, results ranked by
  relevance, each showing the meeting date, title and a highlighted snippet.
- Clicking a result opens the .md file in my default editor.
- Everything stays local: no accounts, no network calls for search.
- Update the README with how to start the search page.

Add speaker labels to the transcripts in the meeting-notes app in this folder. Requirements:

- After transcription, run speaker diarization on the saved audio with pyannote.audio.
  The pyannote models need a Hugging Face token: read it from .env.
- Merge the speaker turns with the whisper.cpp segments so every transcript line starts
  with "Speaker 1:", "Speaker 2:" and so on.
- A "speakers:" block at the top of each .md file lets me rename them (Speaker 1: Anna),
  and the summary uses those names for action-item owners.
- If diarization fails or no token is set, fall back to the unlabeled transcript.
- Update the README with the Hugging Face steps: create a token and accept the model terms.

Make the meeting-notes app in this folder notice calls by itself. Requirements:

- Every 10 seconds, check whether zoom.us, Microsoft Teams or Webex is running, or
  whether a Chrome or Safari tab is open on meet.google.com.
- When a call starts, show a macOS notification "Record this meeting?" with Record and
  Ignore buttons. Never start recording without a click.
- Stop and process the recording 2 minutes after the call ends.
- A menu bar toggle turns detection off.
- Out of scope: reading my calendar or joining calls.
- List any new macOS permissions (Automation access to read browser tabs) in the README.

Add a "chat with my meetings" page to the meeting-notes app in this folder. Requirements:

- Split every note in ~/MeetingNotes into chunks, embed them with nomic-embed-text
  through Ollama, and store the vectors in SQLite next to the notes.
- A page on http://localhost:4322 with one question box. Answer with the LLM already
  configured in .env, using only the five most relevant chunks.
- Every answer lists its sources as links to the notes it used.
- Re-embed a note when its file changes.
- Out of scope: sharing, accounts and any cloud vector database.
- Update the README with the Ollama install step and the model pull command.

Add an importer for a Granola CSV export to the meeting-notes app in this folder. Requirements:

- A command that takes the path to the CSV Granola emails me (Settings > Profile > Generate CSV).
- Read the CSV header first and map the columns for title, date, note summary and transcript;
  show me the mapping and ask before importing anything.
- Write one .md file per meeting to ~/MeetingNotes/YYYY-MM-DD-HHMM.md in the same format as new
  notes, marked "imported from Granola" at the top, with the transcript below the divider when
  the CSV has one.
- Skip meetings that already have a file, so running it twice is safe.
- Print a summary at the end: imported, skipped and failed rows.
- Update the README with the export steps and the import command.