How SelectDB Tackles Speed, Unification, Agent‑Native and Cloud Elasticity for the Agent Era

Event Overview

The SelectDB AI product launch in 2026 attracted over 30,000 developers, architects, and enterprise users who streamed the event to discuss data‑infrastructure trends in the Agent era.

Speed: Sub‑second Response as a Baseline

CEO Ma Ruyue emphasized that "sub‑second response is no longer an optimization, it is a baseline" because an Agent‑driven query may trigger dozens of SQL statements, amplifying latency 10‑ to 50‑fold compared with traditional BI.

SelectDB (built on Apache Doris) demonstrated leading query performance in authoritative benchmarks such as ClickBench, SSB, TPC‑H and TPC‑DS (see chart).

Unified Multi‑modal Data Management

Internal and external data‑source unification enables cross‑Lakehouse, relational and object‑storage access.

Multi‑modal unification supports structured, semi‑structured (JSON), full‑text and vector data within a single analysis engine.

Ma Ruyue stated that future analysis engines must support both cross‑source queries and unified multi‑modal analysis.

Agent Native: Making Data Friendly to Agents

Agent Native is presented as a core highlight: the engine and its stored data become Agent‑friendly, not merely allowing Agents to connect to a database.

Key quote: "The biggest challenge in the Agent era is not generating SQL, but understanding enterprise business semantics." To address this, SelectDB released the MCP Server and a semantic layer that let agents define metrics, dimensions and business logic centrally.

Live demos showed agents completing metric discovery, dimension queries and full analyses via natural language.

Agent Observability with Litefuse

As Agent applications move to production, new observability needs arise: hallucination detection, token consumption tracking, and tool‑selection diagnostics.

SelectDB launched Litefuse, an observability platform that integrates trace collection, storage, analysis and evaluation. Compared with Langfuse, Litefuse saves up to 88% of storage space; text retrieval is 5‑10× faster than traditional LIKE queries, and it remains compatible with over 100 AI‑ecosystem SDKs.

Litefuse was also open‑sourced at https://litefuse.ai/, lowering the barrier for building Agent observability pipelines.

Cloud Elasticity and Serverless Architecture

Agent workloads are bursty and unpredictable, requiring elastic scaling. SelectDB’s storage‑compute separation allows independent scaling, with automatic seconds‑level expansion and contraction and pay‑as‑you‑go billing.

Alibaba Cloud announced the commercial availability of SelectDB Serverless in March 2026, offering the same elastic capabilities and reducing operational complexity for Agent‑driven analytics.

Overall Vision

The launch tied together MCP Server, semantic layer, multi‑modal management, CLI/Skill tooling, Litefuse observability and Serverless architecture under a single goal: evolving the database into an Agent‑native data infrastructure for the AI era.