KeewanoDB improves query speed by 45% with separate tables for users

KeewanoDB has announced a new architecture that assigns a separate table to every user, device and agent, eliminating the need for cross‑table joins in queries. The change, unveiled at the company’s virtual launch event on Thursday, promises to streamline data pipelines for AI agents that rely on detailed event histories.

Traditional relational databases flatten event streams into rows and pre‑computed aggregates, discarding the sequence and contextual links that agents need to explain why something happened. When a retrieval or analytics pipeline must reconstruct that history, it pulls data from a warehouse or search index at query time, adding latency and complexity. KeewanoDB’s per‑entity tables keep the full event context intact, allowing an agent to read its own history in a single, contiguous table.

Early benchmarks from Keewano show that queries on the new schema run up to 45 % faster than equivalent joins on a conventional design, while storage overhead drops by roughly 25 % because duplicate metadata is no longer replicated across tables. “We’re giving data engineers a way to build pipelines that feel more like reading a log than joining tables,” said lead engineer Maya Patel. “Agents can now access their own history in milliseconds, which is critical for real‑time decision making.”

The move positions KeewanoDB as a key tool for teams building conversational AI, recommendation engines and autonomous agents that must reason from past events. The company plans to release an SDK for popular data‑engineering frameworks next month and is exploring integration with major cloud data warehouses. If widely adopted, the architecture could change how developers architect event‑driven systems, making AI pipelines faster, simpler, and more faithful to the raw data that fuels them.