v2.0.0: restructured, complete, agent-navigable journal

A major restructure of the Active Inference Journal into a uniform, machine- and agent-navigable corpus covering the entire Institute YouTube channel.

Highlights

• Complete channel coverage — all 729 videos from the @ActiveInference channel are represented, each with metadata and a transcript (YouTube captions, or local Whisper for caption-less videos).
• New v2 per-item schema — every item has metadata.json (video parts, titles, URLs, durations), transcript.txt + transcript.json, captions/, translations/, and assets/.
• Source-namespaced layout — data/video/activeinferenceinstitute/<Series>/<item>/, ready for other channels (data/video/<other>/) and non-video sources (data/<type>/<source>/).
• Entry points — INDEX.json (machine) and INDEX.md (human) at the repo root.
• Audio branch — the audio branch mirrors main and adds 64 kbps audio per item; main stays lightweight (audio-free).
• Complete documentation — docs/ (SCHEMA, ORGANIZATION, PIPELINE) plus README.md + AGENTS.md at every folder level.

Provenance

Generated and maintained by Journal-Utilities with verified zero data loss. Schema: docs/SCHEMA.md.