Add the Gemma worker with the OPFS model cache

main

Commit

927tanmay2d ago
  • Web Worker loads only the chosen mode's Gemma through the text-generation pipeline, warms up with one token, streams replies and can be stopped
  • GemmaClient wraps it for the page: load, generate, stop, dispose
  • Model files kept in OPFS via react-ai-voice-avatar/model-cache: Light reloads in 2.9 s instead of a 270 s download (MODEL-TESTS.md run 6)
  • Dev-only test panel at ?dev=gemma, not in the production build
  • Tasks: Light first; Heavy and the deep review move to phase H

Entire-Checkpoint: 01M40PS4KPVQ3Z8AAM6BGAWHEM

Checkpoints

Mock Interview Platform for Job Hunting

Claude CodeOpus 5.5
View session
Checkpoint 1