Show HN: Vons – Jev-inspired local decisions for AI agents, in the browser

AI 导读

Vons 是一个用于AI代理工作流的本地决策模型，强调主机控制执行与策略。项目提供研究代码、TypeScript SDK及论文，目标是实现轻量级、可浏览器运行的决策组件。

Vons introduces compact, host-controlled decision models for AI agents, offering research code, SDKs, and papers aimed at browser-based, lightweight decision-making components.

重点速览

• Vons 通过本地决策模型实现AI代理工作流的可控执行 Vons enables controlled execution of AI agent workflows via local decision models
• 项目包含直接评分器与条件扩散评分器等核心算法 Includes direct scorer and conditional diffusion scorer algorithms
• 支持CPU/WASM及WebGPU执行的轻量化设计目标 Targets lightweight execution with CPU/WASM and WebGPU support
• 提供Python合同及TypeScript SDK的完整开发工具链 Provides full development toolchain with Python contract and TypeScript SDK
• 研究代码不含预训练权重且无npm包发布 Research code lacks pretrained weights and has no npm package

一句话：Vons 为浏览器端AI代理决策提供了模块化研究框架 / Vons offers a modular research framework for browser-based AI agent decision-making

推荐理由 项目为AI代理在边缘计算场景的部署提供了创新性技术方案