Databend released 182 changes across June and July 2026 spanning query optimizer statistics (KLL Histograms, Count-Min Sketches, Top-N stats), spatial index joins for R-Tree-based geometry queries, a new community-contributed Paimon Catalog for lakehouse interoperability, expanded data export formats (Avro, ORC, Arrow), SQL and pipeline improvements (ILIKE, Materialized CTEs, Stream Backlog), distributed multi-insert and clustering work, tighter security controls (endpoint egress policy, stage path traversal protection, credential redaction), Lua transactions in Databend Meta, and 80 bug fixes covering crashes, correctness, and upgrade compatibility. It also walks through building a model evaluation observability pipeline using Stage, Stream, and Task features, announces a private-preview AI assistant called Databend Cloud Agent, and lists bendsql and JDBC driver updates.
Table of contents
Two Months, 182 UpdatesBetter Statistics for Better Query PlansSpatial Indexes Move Into Join ExecutionPaimon Joins the Open Lakehouse EcosystemMore Ways to Export and Trace DataSQL and Pipeline ImprovementsSmoother Distributed Writes and Storage MaintenanceTighter Security and Tenant BoundariesDatabend Meta Adds Lua TransactionsWhat the 80 Bug Fixes AddressedBuilding a Model Eval Observability PipelineDatabend Cloud Agent Enters Private PreviewClient and Driver UpdatesWhat This Means for Agent Trace WorkloadsQuestions this post answers
What is Spatial Index Join in Databend and how does it speed up spatial queries?
Spatial Index Join is a new join execution feature in Databend that uses an R-Tree structure to narrow candidate row pairs before applying the exact spatial predicate, avoiding full pairwise comparisons. It includes a Local Spatial Index Join for single-node execution and a Distributed Broadcast Spatial Join for joining a large dataset with a smaller spatial table across a cluster, plus faster geometry processing via streaming bounding-box extraction and reading SRID directly from EWKB headers. Teams building geospatial analytics can track database engine updates like this on daily.dev.
Can Databend query Apache Paimon tables directly without migrating data?
Yes, Databend added a community-contributed Paimon Catalog that lets it read Paimon tables directly and supports distributed writes, so teams already storing data in Paimon can connect it to their analytics workflow without first copying data into another format or storage layer. Iceberg compatibility was also improved, including more reliable predicate pushdown and CI coverage for Iceberg Variant Metadata. Engineers weighing lakehouse catalog options can follow interoperability updates like this on daily.dev.
What is Stream Backlog in Databend and why does it matter for pipeline monitoring?
Stream Backlog is a new API and table function in Databend that estimates pending changes in a Stream, letting operators check whether a downstream cleaning Task is keeping up with incoming data. In a typical pipeline, Stage metadata provides lineage, a Stream tracks incremental changes from a raw table, and Stream Backlog reveals if processing needs to run more frequently or with more compute. Anyone operating unattended data pipelines can keep tabs on operational features like this via daily.dev.