Strategic AI Agents for K8s SRE: Beyond Ad-hoc Prompting
summary
Built an internal AISR agent that starts from pseudo CLI tool calls and later moves to Agno tool calling, using parallel data querying and deterministic context (give pod name + namespace) to trigger reliable root‑cause analysis. A lab injects reproducible failures for back‑to‑back testing; 14 iterations of prompt and algorithm changes yield a measurable 14 iterations, 3% improvement. Uses transformed plain‑English context, cleans YAML inputs, and prefers narrow, tool‑driven steps over freeform chat to reduce hallucination and scale in production.
AI-written summary · about
chapters (13)
- 0:00Introduction to Strategic SSRE AI Agents
- 0:22Understanding AI in SRE
- 1:07History and Evolution of AI in SRE
- 1:59Current Trends and Open Source Players
- 4:02Challenges and Misconceptions in AI SRE
- 5:23Building Effective AI SRE Agents
- 8:56Optimizing AI SRE Performance
- 9:34System Prompts and Context Enrichment
- 13:03Deterministic Approaches in AI SRE
- 19:59Tools and Data Sources for AI SRE
- 23:15Production Deployment and Scaling
- 26:34Continuous Improvement and Testing
- 31:53Conclusion and Future Outlook
related talks
26mManufacturing Alerts To ReActive Agents: Self-Evolving AI Agents on K8S for Predictive…
30mAI Agents for Incident Management
17mFighting AI with AI
34mAutomating SRE Operations with Multi-Agent AI
40mWhen Incidents Fix Themselves: AI SRE in action
44mFrom 4 Hours to 8 Minutes with AI Agents That Transform SRE...
35mAgentic Incident Response
30mArchitecting Integrity: Responsible Software Evolution in the AI Era