抄作业 / 做成了什么 / Grok Bot

做了什么

把 TypeSafe Jev 接到 Grok Bot 作为廉价决策层：在昂贵的调研、浏览器、重试或子 agent 工作前先分类请求，从而复用新鲜产物、停止失败重试、限制调研或请求审批。附用量关卡、skill 模板和示例。

场景

调研监控 写代码上线

资料完整度

• ✓ 作者与原链接有
• ✓ 做了什么有
• ✓ 来源原文摘录有
• ✓ 截图 / 媒体有
• ✓ 步骤 / 章节有
• — 提示词 / 配置原文来源未提供
• — 结果 / 数字来源未提供
• ✓ 风险提示有
• ✓ 相关条目有

— 表示来源没有给出，本站不补写。

来源原文摘录

Connect TypeSafe Jev to Grok Bot as a cheap decision layer - usage gates, skill template, examples
Connect TypeSafe Jev to Grok Bot as a cheap decision layer. Jev classifies the request before expensive research, browser, retry, or subagent work, so Grok Bot can reuse a fresh artifact, stop a failing retry, cap research, or ask for approval. This package is a small open-source reference implementation. It does not change Grok Bot's foundation model and does not route Cursor models. - A Python router around the TypeSafe SDK's `system_one` call.

摘自项目 README，照录原文，未翻译。

效果截图

来源里的步骤 / 要点

1. `media/jev-grok-bot-demo.mp4`
2. A Python router around the TypeSafe SDK's `system_one` call.
3. Explicit actions such as `reuse_cache`, `stop_retry`, `run_deterministic`, `chat_only`, `research_capped`, `allow_subagent`, and `ask_human`.
4. `shadow` mode for measurement and `active` mode for a skill that honors the returned action.
5. A kill switch through `enabled: false` or a `bypass jev` marker.
6. Across five tasks, flight browser opens went from 1 to 0 when a cached artifact was reused; same-approach retries went from 3 to 0; flight-prep skills loaded we
7. The Grok Bot weekly usage meter moved 37% → 38% in the first phase and 39% after the second phase. Exact per-task Grok tokens were unavailable, so this is not a
8. In the separate 24-candidate timing comparison, the uncapped arm took 53.803 seconds and fetched 14 pages; the Jev-ranked top-five arm took 4.125 seconds and fe

照录来源原文，未翻译。

提示词

来源没有给出可直接复制的提示词原文。

怎么复制

1. 打开仓库：github.com/Bodila51/grok-bot-jev ↗
2. 读 README，把仓库链接发给你的 Bot，让它按 README 复现
3. 先在小范围跑通，再加 routine

风险与限制

• 用到浏览器或网页登录：最耗额度，能用 connector 就别开浏览器
• 作者的结果不保证能复现，以原帖为准

用量预估

用金属计算器估一下