跳到正文
OpenAI Developers· @OpenAIDevs · X·· 4 小时前AI 评分51

OpenAI 讲解 Codex 编排架构如何让任务在智能体与工程师间流转

AI 导读

OpenAI Developers 介绍 Codex 的编排器架构:任务在 task service 中保存状态与发现,dispatch loop 周期性检查就绪任务并启动 Codex turn 分配智能体。等待 PR 评审的任务可订阅评审评论,唤醒后由新智能体处理反馈;部署后另一智能体接手同一任务检查性能是否恢复基线。

正文 · 原文

Each task keeps its state and findings, so work can pass between agents and engineers and resume after a code review or deployment.

Engineers define the workflow in the task service. A dispatch loop periodically checks which tasks are ready for the next step. It assigns an agent to each task by starting a Codex turn with the instructions for the task’s current state.

For example, a task waiting on a pull request review can subscribe to review comments. The agent doesn’t need to keep running while an engineer reviews the fix. When a matching comment arrives, the task becomes ready again and the dispatch loop can assign an agent to address the feedback.

After deployment, another agent picks up the same task to check whether performance has returned to baseline.

来源:OpenAI Developers · x.com