Talk To Your Microservice Via a Chat Bot, not UI
summary
Built a GitHub‑issue chatbot called roer that reads issues and runs three commands (merge, deploy, release), cloning repos and running scripts inside Docker containers to package and push. It runs clients polling GitHub (hit limits like 5000 requests/hour), serves ~300 projects, and proved useful but fragile: duplicate‑post loops, silence when the bot or API is blocked, huge logs (link to external log pages), rate‑limit/backoff, named locks to serialize conflicting deploys, and a DB for tracking container state. Overall works but needs careful operational safety.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:27What is Ruler
- 3:28Architecture
- 11:21Problems
- 11:50Problem 1 Garbage
- 14:31Problem 2 Silence
- 18:57Problem 3 Logs
- 21:51Problem 4 Abuse Filters
- 24:54Problem 5 Conflicts
- 27:35Problem 6 Statefulness
- 32:11Problem 7 Thats annoying
- 36:15Conclusion
- 36:39How we solve conflicts
- 38:31The 7 problems
related talks
54mBuilding advanced ChatBots
13mExploring AI-Powered Chatbots and Their Transformative Role
01h 02mProductivity Bots
48mChatbots in Scala: meet your new robot overlords
51mCrafting intelligent GitHub Bots
55mEnterprise Chatbots (using messaging and AI to build novel user interfaces for work)
17mUsing bots to Improve Humanity
30mThe common bot - species, development and breeding