inspector-audio - Inspector Audio Subsystem: Bucket, Streaming, Peaks & Frontend Playback
Manages the full audio pipeline from offline extraction and bucket storage through CDN streaming, segment clipping, peaks generation, to browser-side <audio> playback.
Tags
Updated: 2026-06-30Capabilities
Typical Inputs
Typical Outputs
What this skill does
- resolve audio source across three tiers
- stream CDN audio same-origin
- proxy chapter MP3 audio
- generate segment clips via ffmpeg
- serve slim int8 peak envelopes
- serve audio metadata for dashboard
- extract audio offline via Katana
- upload audio and peaks to bucket
- reconcile reciter audio via auto_detect
- inject Xing TOC for VBR audio
- pack and unpack slim peak envelopes
- render audio via AudioPort in browser
Inputs
- Reciter slug
- Chapter audio URLs from CDN catalog
- Audio manifest JSON
- Bucket storage (NFS mount or CDN)
- Contributor source links
- ffmpeg binary
- Audio MP3 files
Outputs
- HTTP 200/206 audio stream responses
- Re-encoded segment clips
- Compressed int8 peak envelope files
- Dashboard audio metadata (URL, duration)
- Bucket content under reciters/<slug>/
- Reciter state transition to AWAITING_REVIEW
Requirements
- ffmpeg compiled with OpenSSL and protocol support
- Python runtime (inspector app or gunicorn)
- Bucket storage accessible via NFS mount or CDN
- Browser with Web Audio API support
- HTTP Range request support for audio streaming