GitHub Repository

O(n) streaming JSON parser for LLM tool calls. Agents act sooner, abort bad outputs early. WASM SIMD, up to 2000× faster than stock AI SDK parsers.

16 starsJavaScript

VectorJSON – O(n) streaming parser to handle LLM JSON outputs

Name: VectorJSON – O(n) streaming parser to handle LLM JSON outputs
Availability: InStock
Author: teamchong

by teamchong·Feb 18, 2026·1 point·1 comment

AI Analysis

●●●BangerBig BrainSolve My ProblemWizardry

Replaces O(n²) token re-parsing with true O(n) streaming; Vercel SDK does 4K re-parses on 50KB payloads.

Strengths

•Genuine algorithmic improvement; WASM SIMD streaming detection means each byte scanned exactly once, not thousands.
•Schema-driven field filtering skips unwanted JSON at byte level before JS allocation; saves 100MB+ throwaway strings per response.
•Zero-config WASM embedding; handles WasmGC interop internally, users just import and call `feed()`.

Weaknesses

•Niche audience; only relevant to AI SDK authors and agent frameworks, not mainstream web development.
•Bundle size 47KB gzipped limits suitability for ultra-light browser deployments; WASM overhead may not justify gains for small payloads.