Meilisearch — Search, Database product updates and releases, tracked on megachangelog.
Added support for inline dynamic search rules that preview rule effects on individual queries, and enhanced performance details to show facet distribution computation timing. The inlineRule parameter allows applying temporary rule overrides during search without persisting them, useful for preview UIs.
Replaced unsafe bracket indexing with get method and added None recovery in the highlighting logic to prevent panics. This improves stability when processing edge cases in search result highlighting.
Added support for an edge case where the DSR index contains four or more rules, improving stability for complex rule configurations.
This release addresses an important stability issue. Users on v1.53.2 are recommended to update.
Meilisearch v1.52.4 addresses an important stability issue. Users on v1.52.3 are recommended to update.
This release addresses an important stability issue. Update from v1.54.2 is recommended.
Additional logging has been added to improve debugging and observability.
Fixed a potential panic condition where facet count amount could exceed length bounds. The fix ensures stability when sampling facet counts.
Optimized the sample candidate computation in facet counting to run once and produce better results, improving performance and reliability of faceted search operations.
Added visibility into the time required to compute a sample of candidates during facet counting operations. This helps developers understand performance characteristics of faceted search.
The sampling candidate count for facet operations is now configurable via an environment parameter, allowing users to tune performance characteristics without code changes.
Fixed internal errors when processing, displaying, deleting, or canceling dynamic search rule update tasks created in Meilisearch v1.53 or earlier versions.
Removed debug configuration from jemalloc to fix crashes that occurred when debug mode was enabled.
Prototype implementation using CboRoaringBitmaps to improve facet count sampling through optimized bitmap intersection operations.
Prototype release with enhanced debugging capabilities for MCP responses to aid in troubleshooting and development.
Adds implementation of the algorithm for calculating facet numbers, enabling accurate count statistics for faceted search results.
Adds a new algorithm for improved facet count estimation using sampling. This enhances facet functionality for large datasets.
This release fixes issues with expired embedder API keys preventing setting changes, resolves errors when updating or deleting documents with geo data, and aggregates analytics from the MCP route for better tracking.
Removed debug mode from jemalloc to fix crashes that occurred when debug was enabled.
Switched the default memory allocator from mimalloc to jemalloc to fix crashes that occurred in debug builds.
Fixed crashes that occurred in debug mode and removed the secure flag in favor of v2 approach.
Updated heed dependency and removed non_free_page_size to fix crashes when debug mode was enabled.
Resolves crashes that occurred in debug builds by using an experimental patched version of the heed dependency with unique transaction handling.
Fixed an issue where geolocation validation errors would prevent updates or deletions of existing documents. These operations now succeed even if geolocation data becomes invalid.
Meilisearch now detects and removes missing or corrupted tasks from the task queue, improving the reliability and stability of task processing.
Meilisearch v1.54.0 introduces breaking changes to dynamic search rules with a restructured actions API that now supports document boosting, demotion, and hiding via a new "scale" action. POST and PATCH endpoints for dynamic-search-rules now expect actions as an object with pin and scale arrays instead of a list, and score details now expose applied rule UIDs and scaling factors.
Fixed crashes that occurred in debug mode by tracking mimalloc with callbacks and properly registering callbacks to prevent memory management issues.
Symbols are now included in the release build to aid in debugging and crash analysis.
Fixed crashes that occurred when running Meilisearch in secure debug mode with mimalloc memory allocator. This update improves stability for developers using debug builds with enhanced security settings.
Work in progress prototype release focusing on scaling improvements.
Bug fixes and security improvements including removal of task replication penalty, fixes to stop-word highlighting and exactness ranking, and authorization hardening for index operations.
Enhanced error messaging for CorruptedTaskQueue errors by adding contextual information to help users better understand and debug task queue corruption issues.
Improved error messaging for CorruptedTaskQueue errors by adding context information to help developers better understand and debug task queue issues.
Users can now delete enqueued tasks that fail to deserialize, preventing stuck or corrupted tasks from blocking the system.
The index-scheduler web client is now always reused when proxying tasks, reducing latency by approximately 100ms per proxied task for users with sharding.
Applied code formatting and linting fixes to address crashes in the codebase.
Changed the mimalloc dependency to use a temporary repository to fix stability issues and prevent crashes.
Added MEILI_EXPERIMENTAL_TASK_QUEUE_MAX_READERS environment variable to control the maximum number of simultaneous LMDB read transactions for the task queue, improving concurrency management.
Meilisearch has been updated to version 1.52.3 with fixes for crashes.
Version 1.52.3 includes fixes for crash issues encountered in the previous release.
Version bump from v1.51.0 to v1.52.3 addresses and fixes crash issues.
Version bump to v1.52.3 from v1.52.0, addressing crash issues.
Reverted a previous merge (PR #6543) that was causing crashes. The fix-pinning changes have been rolled back to restore stability.
Reverted a previous search speed optimization from version 1.52.2 that caused issues. This restores the previous search behavior.
Reverted the /tasks/stream and /batches/stream Server-Sent Events routes that were introduced in a previous release, addressing related issues.
Bug fixes including making the health route checks blocking and reverting some search progress traces.
Adds sharding support for foreign filters with increased document retrieval limit from 100 to 1000, and new indexSize and usedIndexSize metrics in stats endpoints. Includes security fixes for GitHub Actions workflows and dependency updates.
Introduces /tasks/stream and /batches/stream SSE routes for live engine connections, reducing network load through push-based task and batch status updates. Also improves search speed by converting O(n) operations to O(1) for large result sets, and fixes duplicate pins in federated search.
Version update for v1.52.0 release with crash fixes included in this maintenance release.
Version bump to v1.52.0 with crash fixes included in this release.
Meilisearch v1.51.0 adds filter conditions to Dynamic Search Rules, includes a new lastUpdatedAt field for tracking rule modifications, introduces new environment variables for DSR fuel control, and improves search performance. The release includes breaking changes: the --experimental-dumpless-upgrade flag is renamed to --upgrade-db, and several experimental features (replication parameters, import dumps fallback, and snapshot compression) are removed.
Optimized the deserialization time of the last fields IDs map to improve search request performance.
Optimized field IDs map deserialization to improve search request performance.
Meilisearch v1.50 revamps the Dynamic Search Rules API with simplified conditions and improved scaling up to 75K rules, adds support for federated document fetch in sharded configurations, and introduces several bug fixes and behavior improvements. Breaking changes affect the Dynamic Search Rules and network experimental features.
Prototype release implementing filter condition resolution logic.
Added missing route descriptions to OpenAPI specifications to improve API documentation completeness and clarity.
The synonyms storage system has been enhanced for better performance and efficiency.
Fixed a rare race condition in S3 snapshots that could cause upload failures and corrected federated remote search to prevent duplicate document returns. Also added missing parameters to chat workspace settings endpoints.
Meilisearch v1.47.1 fixes two authentication vulnerabilities: privilege escalation via index-scoped API keys with excessive permissions and information disclosure through search tenant tokens. Users on v1.47 or lower are strongly recommended to upgrade.
Meilisearch v1.48.2 and v1.47.1 fix two authentication vulnerabilities: CVE-2026-57824 allows index-scoped API keys to perform global actions and access global state, while CVE-2026-57823 allows search tenant tokens to disclose limited information about documents outside their scope. Upgrade recommended for all affected versions.
Introduces an experimental queueDocumentsFetch feature that forces document fetch routes to wait in the search queue when no processing threads are available, improving resource management under heavy load.
This is the latest stable release of Meilisearch, including all recent improvements and fixes.
This release fixes a regression where deletions by filter were incorrectly batched with other operations, and extends the new settings indexer to support more parameters for faster document indexing. Security patch included for tar dependency.
Fixed a bug where internal vector stores had mismatched quantization configuration, which could cause inconsistencies in vector storage and retrieval operations.
Reverted autobatch deletions by filter to fix a bug reported in the previous version. This restores the correct behavior for bulk deletion operations.
Meilisearch v1.45.0 improves indexing performance when changing settings and enhances document fetch performance. Key updates include autobatch deletions by filter, better settings indexing performance with experimental edition support, global facet search support, and fixes for document fetching blocking and binary quantization issues.
Meilisearch v1.44.0 adds remote federated facet search for sharded networks, improves indexing memory usage and GeoJSON performance, and introduces new stats query parameters for human-readable database sizes. Includes breaking changes to facet search defaults and embedder timeout behavior when using the network experimental feature.
This release addresses an authenticated SSRF vulnerability in Meilisearch. Self-hosting users who allow third parties to configure instances should upgrade, though no exploitation has been detected on Meilisearch Cloud.
Introduces a new, more efficient settings indexer supporting filterable, sortable, and facet search attributes, improves facet search indexing performance by reducing internal data structure generation time, and fixes bugs in lexicographic string filters and typo tolerance. The GET /health endpoint now returns HTTP 500 after task queue compaction to signal restart requirements.
Multiple bug fixes addressing issues with legacy settings indexer, multimodal fragment handling, embedder settings regeneration, and nested field indexing. Includes CI improvements for testing stability.
67 more changes from Meilisearch. Sign up to read the whole changelog.
Sign up freeGitHub or email — no card needed.