questions: { requested: boolean, issued: boolean, route: choice, // +1 }
NanoJev
NanoJev 是 Jev 的微型版本,实现了并行决策、动态候选和端到端训练流程。

每一个项目,源码里都真的在调用 Jev · 每日自动更新
questions: { requested: boolean, issued: boolean, route: choice, // +1 }
NanoJev 是 Jev 的微型版本,实现了并行决策、动态候选和端到端训练流程。
questions: { department: choice, return_reason: choice, requested_resolution: choice, // +2 }
基于 Qwen2.5-0.5B 构建的微型 Jev-like 模型,可在 MacBook 上训练和运行。
questions: { queue: choice, escalate: boolean, urgency: score, // +1 }
一个小型开放决策模型:状态加类型化问题输入,输出校准概率,基于Qwen3.5重现Jev和System One。
questions: { intent: choice, needs_reasoning: score, }
开源的 System One 决策模型,非自回归,响应时间低于15毫秒,可本地替代 TypeSafe Jev。
questions: { department: choice, urgent: boolean, frustration: score, }
OpenDecision 是一个开源语义决策引擎,类似于 TypeSafe 的 Jev。
questions: { binary: boolean, rating: score, }
使用本地LLM复现Jev功能
questions: { department: choice, frustration: score, is_urgent: boolean, }
兼容Jev的系统,开源实现Jev功能。
questions: { enabled: boolean, color: choice, progress: score, }
在 Apple Silicon 上使用 DiffusionGemma 实现本地 JEV 风格决策,包含基准测试和编码 agent 示例。
questions: { composition: choice, legibility: boolean, }
基于共享视觉编码和 Qwen3-VL,实现图像原生的类型化决策。
questions: { billing: boolean, urgent: boolean, team: choice, // +1 }
Jev风格的模型推理引擎,提供与Jev兼容的API。