Add video transcripts, embeddings, catalog + update documentation

- 5,407 Whisper transcripts (44MB) from Khan Academy videos
- 5,044 sliding-window embedding .npy files (247MB)
- Video catalog.json (5,044 videos, 77K windows)
- AGENTS.md: complete rewrite for current Vite/ES6 modular architecture
- README.md: updated features (50 domains, 2,450 questions, 5,000+ videos)
- Session notes: pipeline refresh status and key insights
- .gitignore: un-ignore video data, ignore duplicate transcripts_raw/

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: jeremymanning

.gitignore

@@ -297,17 +297,15 @@ checkpoints.zip
 wikipedia_articles_level_0.json.zip
 level_0_concepts.json.zip
 data.zip
-.gitignore
 backups/embeddings.zip
 backups/large_checkpoints/level_0_final.json
 backups/large_checkpoints/level_1_after_download.json
 
-# Video pipeline working files (large binary data)
-data/videos/.working/embeddings/
+# Video pipeline working files
 data/videos/.working/coordinates/
-
-# Video catalog (generated by export_video_catalog.py)
-data/videos/catalog.json
+data/videos/.working/audio_cache/
+data/videos/transcripts_raw/
+data/videos/transcripts/
 
 # Python virtual environments
 .venv/