Aha, If so, maybe tsfile can take a place hhhhhh. On Sun, Nov 23, 2025 at 10:56 AM Colin Shi <[email protected]> wrote:
> Have thought about that plan. But PR data volume is fairly small. Maybe > it's better to replace MySQL with SQLite, so there's no need to set up a > database environment. > > On Sat, 22 Nov 2025 at 11:21, Yuan Tian <[email protected]> wrote: > > > Hi zhenyu, > > > > So nice to see that! Just want to ask, why not directly save the pr info > > into IoTDB instead of MySQL? Each pr should have timestamp with it, IoTDB > > is suitable for such case. > > > > Best regards, > > Yuan Tian > > > > Colin Shi <[email protected]>于2025年11月22日 周六10:57写道: > > > > > Hi all, > > > > > > First of all, I'm sorry I would write in Chinese next, as the tool's > READ > > > and comments are all in Chinese. > > > > > > 一个运维的小工具(https://github.com/shizy818/iotdb_prs_ai),用自然语言搜索IoTDB > > PR。基本流程是: > > > 1. 抓取IoTDB PR信息存入MySQL > > > 2. 结合本地IoTDB源码和工具(git, read, grep,glob),通过LLM大模型分析PR解决的问题,存入向量数据库 > > > 3. 聊天机器人通过LLM大模型调用向量数据库封装的工具 > > > > > > 存在的问题: > > > 1. PR的评论中的图片目前没有处理 > > > 2. PR中对问题的描述比较简略。比如大多数没有附带原始错误的stack trace信息,因此用报错信息不一定能找到PR。 > > > 如果第二步分析PR的时候能够带上原始问题的分析文档,搜索可能更准确。 > > > 3. PR的tag中没有区分版本,feat和bug。从运维角度,其实只需要分析Bug PR。 > > > 4. 优化LLM的提示词 > > > > > >
