Accelerating OpenSearch Service Operations With an Operation Agent
summary
Operates ~1300 OpenSearch clusters and reports monthly ~200–400 support tickets and over 200 incident alerts; builds an AI-driven Operation Agent that queries OpenSearch, Prometheus and Kubernetes via MCP servers, moves from a single-agent to a multi-agent with supervisor design, and adds short/long-term memory to reuse successful queries. Results: tool confusion and huge token costs (example: one list-index call ≈20k tokens) shrink, accuracy and cost efficiency improve, but human oversight remains required.
AI-written summary · about
chapters (9)
- 0:00Operation agent intro
- 4:17OpenSearch ops challenges
- 8:35First single-agent system
- 13:12Single-agent limits
- 17:47Multi-agent architecture
- 21:19Agent memory improvements
- 24:12Cost and context insights
- 27:25Future plans and questions
- 27:46Q&A
related talks
28mAI Agents in Action: Harnessing MCP and OpenSearch for Modern Use Cases
42mBuilding Self-Healing Infrastructure: Agentic AI & Vector Search in OpenSearch
28mChatOps Using OpenSearch MCP Server
37mAgentic AI Meets Search: Real-World Use Cases With...
43mOpenSearch & AI Driven Cloud Native Platforms: Adaptive & Agentic Microserv...
40mAgentic Relevance Tuning With OpenSearch: Faster, Easier, and Higher Quality Search at Scale
36mBuilding Next-gen Search With OpenSearch
46mOperational Logs and Event Management for HPC S...