Sponsored Session: Observability and Agentic AI - From Observability to Answers...
Talkat Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit + Automotive Linux Summit 2025 · 2025-12-08..12-10·
summary
Demonstrates an agentic-AI shift from manual queries to grounded answers using ClickHouse + HyperDX + OpenTelemetry, running against a real observability store of roughly 100 PB with six-month TTL and a customer base exceeding 2,000 paying users. Shows agents auto-generating SQL to find root causes (cluster connectivity under 1,000 concurrent inserts, memory-limit failures, wrong partition key) but producing unstable queries and noisy investigations, so a multi‑MCP setup to stabilize SQL generation, scope data, and cap cost is built and partially successful.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:03ClickHouse AI plans
- 2:20What is ClickHouse
- 3:37ClickHouse in Japan
- 5:50ClickHouse ecosystem
- 6:54ClickHouse observability tools
- 9:17HyperDX platform
- 13:37AI agent capabilities
- 16:11ClickHouse for AI agents
- 17:42AI agent dataset challenges
- 19:13Testing AI agents
- 21:43Fixing AI agent instability
- 23:31Live AI agent error demo
- 27:27AI agent root cause analysis
- 28:53Future AI agent plans
- 30:35Q&A
related talks
34mThe State of SQL-Based Observability
19mSession: Game Data For All: How AI Is Democratizing Game Analytics
28mDO WE STILL NEED TO "OBSERVE" THE FUTURE OF AI & OBSERVABILITY
28mWhere are we going with observability?
19mDo We Still Need to “Observe”? the Future of AI & Observability
47mThomas in't Veld | AI Track | Compass AI & Data Summit 2025
13mObservability Where are We Headed?
30mAI Agents for Incident Management