manticoresoftware/manticoresearch release-29.0.2

Manticore Search 29.0.2

on GitHub

Version 29.0.2

Released: August 14th 2026

This release adds sharded-table settings inspection, distributed and sharded table optimization, Unix-socket Buddy callbacks, more reliable conversational search, configurable file access, and faster columnar KNN rescoring, with fixes across backups, JDBC compatibility, highlighting, query inspection, KNN compaction, distributed-query memory use, template tables, and concurrent RT-table optimization.

New Features and Improvements

• 🆕 v29.0.0 Issue #4802 Added synchronous OPTIMIZE TABLE support for distributed and sharded tables via OPTION sync=1, applying the requested cutoff to every local RT component and configured remote mirror.
• 🆕 v28.10.0 PR #4799 Added SHOW TABLE SETTINGS support for native sharded tables.
• 🆕 v28.8.0 Buddy PR #691 Improved conversational search reliability by retrying failed LLM route-selection tool calls up to three times before returning an error.
• 🆕 v28.7.0 Issue #3837 Added access_columnar_attrs and access_secondary settings for selecting buffered file or mmap access to columnar and secondary-index files, and improved columnar KNN search performance by batching distance calculations during rescoring.

Breaking Changes

• ⚠️ v29.0.0 Issue #4739 Added streamed heartbeat replies for long-running shard writes and replication SST, changing the SHARD_WRITE API and the cluster protocol. This matters only when upgrading or downgrading mixed-version deployments that use native sharded-table writes or replication; standalone nodes and deployments that use neither require no special action. Existing data and configuration remain compatible, with no migration required. During upgrades, update remote shard agents before their coordinators and all replication peers before initiating SST. Downgrade remains possible without rebuilding data, but coordinators should be downgraded before agents and SST must be avoided while peers run mixed versions.

Bug Fixes

• 🪲 v29.0.2 Backup PR #140 Updated manticore-backup to 1.10.3, fixing failed backups leaving RT tables frozen when an error occurs after a per-table FREEZE.
• 🪲 v29.0.2 Issue #4434 Updated Buddy to 4.4.1, restoring MySQL Connector/J initialization by supporting the @@query_cache_size system variable and numeric system-variable values.
• 🪲 v29.0.1 Issue #4780 Fixed searchd crashes during RT-table optimization when concurrent updates resized blob-backed attributes, including KNN tables, by preventing merge workers from reading stale source memory.
• 🪲 v29.0.0 Issue #4739 Fixed long-running shard writes timing out while a physical table was flushing or performing other expensive work; heartbeat replies now keep the connection alive without requiring an excessively large agent_query_timeout.
• 🪲 v28.9.3 Issue #4792 Fixed CALL SNIPPETS and HIGHLIGHT() producing invalid nested HTML with html_strip_mode=retain; highlight wrappers now preserve existing element boundaries and generate well-formed markup.
• 🪲 v28.9.2 Issue #4678 Fixed a use-after-free that could crash searchd when SHOW THREADS inspected concurrently running HTTP SQL queries.
• 🪲 v28.9.1 Issue #4738 Fixed explicit optimize_cutoff values not applying to the KNN compaction threshold, so server-wide and per-table settings now control automatic chunk merging for vector tables as expected.
• 🪲 v28.6.8 PR #4777 Fixed EXPLAIN QUERY hanging on template tables when a full-text query contains field selectors; valid selectors are now discovered by the full-text parser, while malformed selectors return syntax errors.
• 🪲 v28.6.7 Issue #4771 Fixed high master-daemon memory use after completed distributed-agent operations by releasing large request and reply buffers immediately after parsing instead of retaining them until agent_query_timeout expires.