Show HN: 我开发了一个能自动将工单转化为设计文档和 PR 的工具
1 分•作者: rajivprab•4 个月前
嗨，HN！我创建了 Code Prodigy ( <a href="https://codeprodigy.io/" rel="nofollow">https://codeprodigy.io/</a> )，一个自主的 AI 工程师，驻留在你的工单追踪器上。当有人在 Jira（或 Linear、Asana、Trello...）中提交工单时，它会：
1. 调查所有代码库中的工单，并创建设计文档
2. 打开一个或多个 PR（拉取请求）来实现所请求的更改
3. 回复其自身 PR 或原始工单上的评论。重新设置基线、处理反馈、推送更新
它不是一个需要你主动提示的聊天窗口。它只是监视你的工单追踪器，并在创建工单后立即开始工作。在你的工程师审查并批准之前，任何内容都不会被合并或部署。
工作原理：每个任务都在 AWS 中启动一个临时的 Docker 容器，其中包含 Claude Agent SDK。代理运行一个工作流程，该流程识别相关的代码库，调查现有的代码库，为工单的需求提出设计方案，实现它，并启动新的代理来审查其自身的工作。你的代码和工单仅供为你工作的代理访问，这些代理在私有子网中的全新 Docker 容器中运行。
与类似产品相比，它的不同之处在于：
* 完全由我们托管。你这边无需任何基础设施或设置
* 没有模型锁定。我们使用当前最先进的技术
* 与你的团队使用的任何工单系统或版本控制系统兼容。无需迁移
* 监视其 PR 或工单上的评论，并根据你的反馈进行迭代
* 使用行业最佳实践来获得最佳的 AI 结果。例如，研究、计划、实施、独立审查、独立修订等
* 无需手动步骤即可启动代理，从而大大加速了组织的采用
我们构建这个工具是因为我们厌倦了工单在积压中停留数周之久。拥有一个 AI 代理自动生成设计文档和 PR 极大地加速了流程。即使它的输出不完美，它仍然对克服惰性有很大帮助——允许工单在几天内解决，而不是几周。
如果你对此感兴趣，请与我们联系，我们很乐意为你提供演示并回答你的任何问题。我们很乐意为你的团队使用的任何工单系统和版本控制系统构建支持。
<a href="https://codeprodigy.io/" rel="nofollow">https://codeprodigy.io/</a>
查看原文
Hi HN! I built Code Prodigy ( <a href="https://codeprodigy.io/" rel="nofollow">https://codeprodigy.io/</a> ), an autonomous AI engineer that lives on your ticket tracker. When someone files a ticket in Jira (or Linear, Asana, Trello...), it:<p>1. Investigates the ticket across all your repos, and creates a design doc
2. Opens one or more PRs implementing the requested change
3. Responds to comments on its own PRs or on the original ticket. Rebases, addresses feedback, pushes updates<p>It's not a chat window you have to prompt. It just watches your ticket tracker and starts work as soon as a ticket is created. Nothing gets merged or deployed until one of your engineers has reviewed and signed-off on it.<p>How it works: Each job spins up an ephemeral docker container in AWS with the Claude Agent SDK. The agent runs a workflow that identifies the relevant repositories, investigates the existing codebase, comes up with a design plan for the ticket's requirements, implements it, and spins up fresh agents to review its own work. Your code and your tickets are only ever made accessible to agents that are working for you, in a fresh docker container in a private subnet.<p>What makes it different from somewhat-similar products:
- completely hosted by us. zero infra or setup needed on your end
- no model lock-in. we use whatever is SOTA at the present time
- works with whatever ticketing system or version-control system your team is using. no migration needed
- watches for comments on its PRs or tickets, and iterates based on your feedback
- uses industry best practices for getting optimal ai results. Eg, research, plan, implement, independent review, independent revision, etc
- zero manual steps needed to kick off the agent, thus massively accelerating organizational adoption<p>We built this tool because we got tired of tickets sitting in the backlog for weeks and weeks. Having an AI agent automatically produce a design doc and PR has been a huge accelerator. Even when its output isn’t perfect, it still helps immensely in overcoming inertia - allowing tickets to get resolved in days, instead of weeks.<p>If this sounds interesting to you, reach out to us and we’d love to give you a demo and answer any questions you may have. We’re happy to build support for any ticketing system and version-control system your team is using.<p><a href="https://codeprodigy.io/" rel="nofollow">https://codeprodigy.io/</a>