Author: Microsoft Developer - Bewertung: 0x - Views:6
AI agents need different data infrastructure than humans—and they forget everything between sessions. See how TiDB is built for agentic workloads, combining vector search, BM25, and SQL in a unified table to manage agent memory at scale. Learn how hybrid retrieval with RRF, Azure OpenAI embeddings, and ACID transactions enables reliable, scalable agent systems you can deploy in Azure environments.
𝗦𝗽𝗲𝗮𝗸𝗲𝗿𝘀:
* Ravish Patel
𝗦𝗲𝘀𝘀𝗶𝗼𝗻 𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻:
This is one of many sessions from the Microsoft Build 2026 event. View even more sessions on-demand and learn about Microsoft Build at https://build.microsoft.com
ODSP918 | English (US) | Cloud platform & data
Pre-recorded | (300) Advanced
#MSBuild
Chapters:
0:00 - Three main challenges: bursty workloads, massive concurrency, constant context recall
00:04:03 - Refund process and email failure scenario
00:05:15 - Introduction to TiDB
00:06:14 - TiDB features for agents: solving prior data problems
00:09:45 - Demo Step 2 – Inserting user memories and automatic embedding generation
00:11:57 - Demo Step 5 – Running hybrid search combining vector and keyword results
00:13:10 - Introduction to asset transactions across multiple tables in TiDB
00:13:38 - Transaction demo showing consistent multi-table insert and update operations
00:14:19 - Case study: Manus AI startup deploying millions of agent databases on TiDB