Meeting annotation and speaker-attribution tooling for high-fidelity conversation intelligence.
This repo is the public, reusable core of a larger production workflow. It is designed to be clean, modular, and easy to adopt in other projects.
- Multi-speaker segmentation + attribution
- Meeting timeline and event annotation
- Speaker-aware summarization scaffolds
- Export-friendly structured outputs
- Release speaker-attribution baseline API
- Add meeting event taxonomy + annotation schema
- Add quality metrics for diarization/attribution
- Publish reproducible sample notebook + CLI
- Private implementation roots: caf-interview-analysis, mining-meetings, caf_meeting_podcast.
Active build-out. Initial public baseline is focused on clean APIs, examples, and strong docs.
MIT