Cedar 0.1.0 is the first public build. The capture-and-notes workflow is
feature-complete for solo meetings:
- Audio + visual recording. Cedar captures the conversation AND the
screen at the same time. Slides, demos, browsers, code editors — every
shown thing becomes a timestamped frame in the meeting.
- Smart visual dedupe. A perceptual-hash gate skips near-duplicate
frames so you keep only the moments where the screen actually changed.
- OCR on every kept frame. Vision's text recognition runs after each
frame is saved; the recognised text feeds the cross-modal search layer.
- Transcription. Local Whisper-derived model. Audio doesn't leave
your computer.
- Structured memos. The transcript and the captured frames become a
searchable, organisable memo you can edit and share.
- Calendar. Cedar reads your system Calendar so the right meeting is
at the top.
- System audio. Cedar can tap the audio of other apps (Zoom, Google
Meet, etc.) so the other side of the call is captured too.
Known caveats:
- macOS 13+ for this initial build. Linux and Windows builds are in
active development.
- The on-device LLM sidecar wants ~8 GB of RAM. It still runs on less
but you'll see the fan spin during heavy generation.
This is the release the website shipped against. Subsequent release notes
will land in this folder as new files.