Apache Doris 4.1 on VeloDB: Unifying AI, Search, and Real-Time Analytics

Apache Doris 4.1 is now available on VeloDB Cloud and includes advances in both the AI and search features introduced in 4.0, as well as the real-time analytics core for which Apache Doris is known.

On the AI side, the new IVF index stores clustered inverted lists on disk and pairs with product quantization, taking vector search past memory limits. The search() function provides full-text search and a rich set of operators in plain SQL. And single JSON documents now reach 100MB, with DOC Mode for retrieving large text, so long-context data lives in one table.

On the analytics side, the core engine runs up to 22.6% faster than 4.0 on standard benchmarks and ranks first on ClickBench in cold query, with the new Condition Cache reusing results by filter attribute. Full Iceberg V3 support is also here, and real-time MySQL and PostgreSQL sync covers full load through incremental changes.

Join us for a walkthrough of the release and what each capability can replace in your stack.

Key Takeaways:

• Apache Doris 4.1 unifies vector search, full-text search, structured filtering, and long-context JSON for production AI and search workloads
• Compared to 4.0, Apache Doris 4.1 improved performance across the board: 22.6% faster in TPC-H, 19.1% in TPC-DS, 14.3% in SSB, and ranked first on ClickBench in cold query.

Speakers:

Kevin Shen, Principal Product Manager, VeloDB