AI in SRE: Unlocking Prometheus Insights with Natural Language
summary
Built a web and Slack tool called PromptChart that turns questions into PromQL, runs them against Prometheus, and translates results back to natural language. It uses the Prometheus metadata API and LLM function calls as tools, handled counters and group sum across three service instances to report ~1.36 million items shipped, and shows simple models fail on complex queries while Gemini 2.0 produces correct PromQL. Source is open source and usable today.
AI-written summary · about
chapters (8)
- 0:00Introduction to AI in SRE
- 0:44Understanding the Problem and Solution
- 4:04Approach and Architecture
- 10:34Exploring the Prompt Chart Tool
- 14:19Case Studies and Examples
- 27:56Summary and Lessons Learned
- 33:00Future Improvements and Contributions
- 36:15Conclusion and Final Thoughts
