Repository navigation
feat(llm): upgrade web-llm to 0.2.86 and default to Gemma 4 E2B - #30
Merged
Merged
Conversation
- Pin web-llm 0.2.86 in llm-rag.js and llm-worker.js - Re-audit downloadBytes for all 22 models against the rebuilt v0_2_86 WASM libraries (weights and VRAM are unchanged) - Add gemma-4-E2B-it-q4f16_1-MLC as a direct, limited-RAG-tier option with its 4K prebuilt context; document the size formula - Test that the main thread and worker pin the same web-llm build
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
- Gemma 4 E2B moves to the recommended group and RAG tier as the only isDefault model; Llama 3.2 3B stays in the recommended group - New installs persist their config on first read. An install that already downloaded models but never saved a setting is pinned to the previous default (Llama 3.2 3B), and saved configs are not migrated, so no existing user gets a surprise 2.84 GB download
arminpasalic
changed the title
arminpasalic
marked this pull request as ready for review
October 24, 2026 13:33
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Upgrades web-llm 0.2.84 → 0.2.86 and makes Gemma 4 E2B (
gemma-4-E2B-it-q4f16_1-MLC) the default model for new installs.Upgrade
llm-rag.js,llm-worker.js) now pin 0.2.86. A new test keeps the two in step.v0_2_86), so I re-measured everydownloadBytes(each +0.7–1.0 MB). The formula is now documented inmodel-constraints.jsand reproduces the previous audit byte for byte.Gemma 4 E2B as the default
recommendedRAG tier, the onlyisDefaultmodel.context_window_size.q4f32_1skipped).model_manifestpath.Existing users keep their model (new installs only)
STORAGE_VERSIONbump: saved configs on Llama 3.2 3B are not migrated.model_id, so it would have switched silently.getConfig()now handles the first read without a stored config. Ifvectoria_models_readyorvectoria_models_cachedis set, it pins Llama 3.2 3B. It always saves the result, so a new install can't later look like an old one.Verification
npm run check: 233 unit tests (new tests cover the default and the new-installs-only behaviour), 16 wasm tests, build, stamp checknpm audit(root and mcp-server): 0 vulnerabilitiesDEFAULT_CONFIGand the catalog agree on the defaultDraft until the browser test is done.