January 2026 focused on strengthening reliability and predictability of term-aggregation behavior in paradedb/paradedb. A critical consistency upgrade unified bucket limit handling across all execution paths by tying it to the paradedb.max_term_agg_buckets GUC. The change ensures TopN window aggregations, AggregateScan, and direct paradedb.aggregate calls all respect a single source of truth for bucket limits, reducing surprises for users with large term vocabularies. A GUC-based default path was added for paradedb.aggregate, while preserving explicit bucket_limit overrides. Local validation against Postgres 17 (using cargo pgrx) and a dedicated repro script confirmed expected behavior across bucket settings. This work was co-authored by Krishna Prasad and validated via targeted tests.

December 2025 monthly summary for cocoindex. Focused on delivering reliability, maintainability, and usability improvements across testing, documentation, and live data workflows. Key outcomes include stabilizing the testing workflow in environments with limited Python stdlib availability, reorganizing documentation to strengthen Query Support accessibility, and adding a proactive warning when live update passes exceed the configured refresh interval. These efforts reduce test flakiness, accelerate onboarding, and improve observability of potential delays in live updates.