Repository navigation
Reject BQ on packed binary indexes and fix runnable onboarding - #15
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Calling
enableBQ()on a packed binary index could trap inside WASM. Both native BQ entry points now reject that storage before mutation; the existing structured error shape is preserved, and binary insertion/search/persistence remain usable after rejection. Float and SQ8 behavior is retained.The public onboarding examples also contained invalid constructors, implicit BQ enablement, unsupported WASM SQ8 configuration, and browser-only persistence in the Node benchmark. This fixes those paths, clarifies binary-code versus total-memory usage, and executes the root/package/API README quickstarts against real installed tarballs. Node examples verify real EVWB file roundtrips, including restored nearest neighbors.
Validation: 9 targeted native regressions, 17 release-WASM SDK cases, 10 WASM persistence tests, 33 Jest tests, 11 IndexedDB shim tests, and 5 package/demo persistence checks each pass. Strict native/WASM lint, formatting, TypeScript, installed ESM/CommonJS declarations and browser bundling pass. Installed consumers pass on Node 16 and 22; all three README quickstarts execute successfully.
The full 100k-vector/128d normalized-cosine Node benchmark runs 1,000 queries with P99 1.143 ms and verifies a file reload in 125.92 ms. Six fresh-process 100k-vector loads take 121–139 ms. WASM is 698,412 bytes; combined gzip WASM/JS is 295,502 bytes.
Evidence:
docs/reviews/2026-10-20_onboarding_validation.mdanddocs/benchmarks/2026-10-20_onboarding.json. Merge is gated on hosted checks and actual mobile-device results. This is a GitHub source/demo correction; package versions are unchanged.The checked-in package/demo intentionally use the documented scalar compatibility profile for older Safari; the default source build still enables SIMD. All measurements above use the frozen scalar artifacts.