From recording to note and knowledge
The animation explains the architecture; no live actions are executed.
This diagram belongs to the 5 October snapshot. My 18 October update describes the context correction I now use. A complete PLAUD trace is not shown here.
- PLAUD
-
- Capture audio
PLAUD is the recorder. The audio then needs to become available for import.
- CLI adapter
-
- PlaudCliProvider
PlaudCliProvider in bin/openclaw_meeting_memory.py is my source adapter. The wrapper provides transfer and status commands.
- Meeting memory
-
- Transcript and speakers
STAGES separates audio, transcript, diarization, identity, extraction, rag and kg. Each stage has its own status.
- Apple Notes
-
- Read the note
render_note(store, meeting_id) prepares the note. A saved iCloud note can then sync to my other Apple devices.
- RAG + graph
-
- Store · retrieval planned
New passages and relationships enter the knowledge layer after extraction. Retrieving earlier context before extraction is still development work.
Connections
- PLAUD → CLI adapter
- CLI adapter → Meeting memory
- Meeting memory → Apple Notes
- Meeting memory → RAG + graph
- RAG + graph → earlier context · planned → Meeting memory
How to read this
- PLAUD is the recorder. The audio then needs to become available for import.
- PlaudCliProvider in bin/openclaw_meeting_memory.py is my source adapter. The wrapper provides transfer and status commands.
- STAGES separates audio, transcript, diarization, identity, extraction, rag and kg. Each stage has its own status.
- render_note(store, meeting_id) prepares the note. A saved iCloud note can then sync to my other Apple devices. New passages and relationships enter the knowledge layer after extraction. Retrieving earlier context before extraction is still development work.