Meilisearch
Open-source, fast, typo-tolerant search engine API.
Meilisearch tightens sharding latency and index authorization in a flurry of maintenance releases.
◆Recent moves
- 8d ago
v1.53.2: security fix + correctness patches
v1.53.2 bundles three correctness fixes—stop-word highlighting, exactness ranking after settings changes, and 100ms task-replication penalty removal—plus a security fix tying authorization to index opening. Small on the surface, but the authorization fix closes a potential access-control gap in multi-index setups.
View source ↗ - 13d ago
prototype-v1.50.0-reuse-http-client.3
Internal prototype iteration adding error context to CorruptedTaskQueue messages. No user-visible change; part of the reuse-http-client experiment cycle.
View source ↗ - 13d ago
prototype-v1.50.0-reuse-http-client.2
Another iteration of the reuse-http-client prototype branch with the same error-context change. Pure internal scaffolding; no shipped feature.
View source ↗ - 13d ago
Allow deleting non-deserializable enqueued tasks
Adds the ability to delete enqueued tasks that can't be deserialized, preventing a stuck-queue scenario. A reliability guard that operators running long-lived Meilisearch instances will notice if they've ever hit a corrupted task edge case.
View source ↗ - 14d ago
Prototype: reuse HTTP client for sharding, saves ~100ms per proxied task
The reuse-http-client.0 prototype demonstrates ~100ms latency reduction per proxied task in sharded deployments by keeping the index-scheduler web client alive across requests. This fits the pattern of Meilisearch investing in clustered-scale performance ahead of a likely stable release.
View source ↗ - 22d ago
prototype-v1.52.3-fix-crashes.6
Swaps mimalloc to a temporary repository during the crash-fix prototype cycle—an internal dependency management change with no user-visible effect.
View source ↗