GitHub Trending
← 返回首页

GitHub Trending 2026-09-24

今日热榜以类型化决策模型与 Agent 编排为主:Laya 系列非自回归决策引擎、Z.ai ZCode 编程智能体、jev-chat 手机对话副驾 等项目受关注。

  • NandhaKishorM/laya ⭐20,471 (Python) — Non-autoregressive System 1 decision engine. Typed choice / score / yes-no decisions.
  • zai-org/ZCode ⭐6,536 (TypeScript) — Z.ai’s coding agent harness. Powerful, intelligent, extensible.
  • mizorewww/laya-mlx ⭐5,960 (Python) — Native MLX runtime for Laya typed decision models — 7–14 ms short decisions on M-series chips.
  • jev-chat/jev-chat-jarvis ⭐5,251 (Kotlin) — 手机端对话副驾:在微信/QQ/X/飞书读懂对方、给候选回复、一键填入输入框。
  • unreallabsai/unreal-agent ⭐1,719 (Go) — Async-first agent harness.
  • bespokelabsai/nimble ⭐1,685 (Python) — Local typed decisions, contrastive data curation, and model evaluation.
  • mizorewww/laya-coreml ⭐1,391 (Python) — Local Laya typed decisions on Apple Core ML and Neural Engine.
  • newliver666/apk-reverse ⭐1,278 (Python) — Suitable for Android APK reverse engineering analysis.
  • Lumid-Off/AirCard-Windows ⭐785 (Rust) — AirCard Windows client.
  • heyjunpenn/awesome-jev ⭐772 (Astro) — Verified community-maintained catalog of 834 open-source projects in the jev ecosystem.
  • githubnext/localjev ⭐736 (TypeScript) — Local jev inference runtime.
  • SewCabinSpout/cleanupper ⭐723 (JavaScript) — Free, open-source macOS disk cleanup CLI.
  • volotat/mini-AGI ⭐719 (Python) — Continual learning model trained from scratch on 8GB VRAM laptops.

注:GitHub数据来源:GitHub Search API(api.github.com/search/repositories,sort=stars&created>2026-09-17)。