Breaking Changes

none

New Features

none

Fixes

• Make sure shard is properly cleaned up on partial load by @etiennedi in #5102
• async indexing: validate input before push by @asdine in #5090
• Fix nil-pointer deref panic on partial shard cleanup by @etiennedi in #5109
• VoyageAI new embedding and rerank models by @fzowl in #5108
• Update backup related dependencies MinIO, Azure, Google Storage to latest versions by @antas-marcin in #5120
• Add trivy vulnerability check on every pull request by @jfrancoa in #5133
• fix obj del no local existence by @jeroiraz in #5124
• fix batch obj deletion by @jeroiraz in #5126

Full Changelog: v1.24.17...v1.24.18

Breaking Changes

none

New Features

none

Fixes

• Only lock while updating mappings by @dirkkul in #5059
• Reenable custom linter by @dirkkul in #5063
• Increase time between mapping refreshs by @dirkkul in #5062
• Better error handling during LSM store initialization by @etiennedi in #5065
• Fix score extraction by @dirkkul in #5066
• Fix shard loading on t activation, speed up BQ cache loading by @etiennedi in #4977
• Add slow query reporter to ObjectSearch and ObjectVectorSearch by @sebneira in #5069
• fix fsync segment directory after compaction by @jeroiraz in #5075
• Add support for getting the Google auth tokens using Google's service account by @antas-marcin in #5073
• Improve error message: Annotate failing segment with segment path on load by @etiennedi in #5087
• scalar quantization: add SIMD dot product by @asdine in #5077
• revert: coordinator reply channel length on pulling by @moogacs in #5089
• fix expired context on dimension tracking by @etiennedi in #5092
• Fix unbounded concurrency for batch delete by @etiennedi in #5094
• Make sure shard is properly cleaned up on partial load by @etiennedi in #5102
• async indexing: validate input before push by @asdine in #5090
• Fix nil-pointer deref panic on partial shard cleanup by @etiennedi in #5109
• VoyageAI new embedding and rerank models by @fzowl in #5108
• Modify buf-checks job to accept github.token by @antas-marcin in #5103
• Bump goreleaser/goreleaser-action from 5 to 6 by @antas-marcin in #5117
• Update backup related dependencies MinIO, Azure, Google Storage to latest versions by @antas-marcin in #5120
• Add trivy vulnerability check on every pull request by @jfrancoa in #5133

New Contributors

• @sebneira made their first contribution in #5069

Full Changelog: v1.25.2...v1.25.3

Breaking Changes

none

New Features

none

Fixes

• Improve error message: Annotate failing segment with segment path on load by @etiennedi in #5087
• Revert: coordinator reply channel length on pulling by @moogacs in #5089
• Fix expired context on dimension tracking by @etiennedi in #5092
• Fix unbounded concurrency for batch delete by @etiennedi in #5094

Full Changelog: v1.24.16...v1.24.17

Breaking Changes

none

New Features

none

Fixes

• Fix HNSW visited list leak and improve error handling of already deleted nodes by @etiennedi in #5013
• Only lock while updating mappings by @dirkkul in #5059
• Increase time between mapping refreshs by @dirkkul in #5062
• Better error handling during LSM store initialization by @etiennedi in #5065
• Fix score extraction by @dirkkul in #5066
• Add slow query reporter to ObjectSearch and ObjectVectorSearch by @sebneira in #5069
• fix fsync segment directory after compaction by @jeroiraz in #5075
• Add support for getting the Google auth tokens using Google's service account by @antas-marcin in #5073

New Contributors

• @sebneira made their first contribution in #5069

Full Changelog: v1.24.15...v1.24.16

Breaking Changes

none

New Features

none

Fixes

• Do not recover from panics in integration tests by @dirkkul in #4970
• Support Implicit Tenant Activation (targeting v1.25.1) by @etiennedi in #4826
• RAFT: avoid rebuilding GQL twice by @moogacs in #4969
• fix: add update class description by @moogacs in #4982
• Refactor cluster client/server RPC implementation by @reyreaud-l in #4983
• Fixed documentation reference in meta.go by @Maurits-de-Groot in #4990
• RAFT: MT use RLock for reads by @moogacs in #4991
• Fix extracting near object vector for named vector classes by @tsmith023 in #4994
• fix: response of replication.CommitReplication() bubble up error by @moogacs in #5001
• various fixes to indexing queue by @asdine in #4854
• More accurate module usage telemetry data by @antas-marcin in #4996
• Remove superfluous findEntrypoint in delete logic by @etiennedi in #5010
• Fix broken index.Contains logic in flat index by @etiennedi in #4971
• RAFT: always load db on start for snapshot by @moogacs in #5014
• fix: local CommitReplication response if local shard Not found by @moogacs in #5017
• replication: replica fail early on getters with no-retry errors by @moogacs in #5031
• Fix race in tombstone cleanup for stable/v1.24 by @trengrj in #5030
• Fix handling of OpenAI headers by @antas-marcin in #4997
• Add support for X-Google-Vertex-Api-Key header in multi2vec-palm module by @antas-marcin in #5033
• Keep track of deleted tombstones across commit logs by @trengrj in #5028
• Delete bloom tmp files before compaction by @amourao in #4992
• Introduce Tombstone Cleanup Limits by @etiennedi in #5012
• chore wrapped allow list by @jeroiraz in #5008
• Proplen tracker fix by @donomii in #5022
• backup:scheduler: compare and error based on major versions only by @antas-marcin in #5039
• chore: context-aware wand by @jeroiraz in #4907
• Replication: bubble up error on replication writes if shard init errors by @moogacs in #5044
• Fix Azure OpenAI batch endpoint handling by @dirkkul in #5035
• refactor cluster implementation to have clear separation of raft/schema by @reyreaud-l in #4993
• Cluster package housekeeping by @reyreaud-l in #5045
• Fix handling of OctoAI headers by @antas-marcin in #5050
• Support auto tenant creation for single object inserts by @parkerduckworth in #5047
• finish refactor of raft & cluster config by @reyreaud-l in #5053
• Fix HNSW visited list leak and improve error handling of already deleted nodes by @etiennedi in #5013

New Contributors

• @Maurits-de-Groot made their first contribution in #4990

Full Changelog: v1.25.1...v1.25.2

Breaking Changes

none

New Features

none

Fixes

• Do not recover from panics in integration tests by @dirkkul in #4970
• Fixed documentation reference in meta.go by @Maurits-de-Groot in #4990
• Fix extracting near object vector for named vector classes by @tsmith023 in #4994
• fix: response of replication.CommitReplication() bubble up error by @moogacs in #5001
• various fixes to indexing queue by @asdine in #4854
• More accurate module usage telemetry data by @antas-marcin in #4996
• Remove superfluous findEntrypoint in delete logic by @etiennedi in #5010
• Fix broken index.Contains logic in flat index by @etiennedi in #4971
• fix: local CommitReplication response if local shard Not found by @moogacs in #5017
• replication: replica fail early on getters with no-retry errors by @moogacs in #5031
• Fix race in tombstone cleanup for stable/v1.24 by @trengrj in #5030
• Fix handling of OpenAI headers by @antas-marcin in #4997
• Add support for X-Google-Vertex-Api-Key header in multi2vec-palm module by @antas-marcin in #5033
• Keep track of deleted tombstones across commit logs by @trengrj in #5028
• Delete bloom tmp files before compaction by @amourao in #4992
• Introduce Tombstone Cleanup Limits by @etiennedi in #5012
• chore wrapped allow list by @jeroiraz in #5008
• Proplen tracker fix by @donomii in #5022
• chore: context-aware wand by @jeroiraz in #4907

New Contributors

• @Maurits-de-Groot made their first contribution in #4990

Full Changelog: v1.24.14...v1.24.15

Breaking Changes

none

New Features

none

Fixes

• fix: response of replication.CommitReplication() bubble up error by @moogacs in #5001
• fix: local CommitReplication response if local shard Not found by @moogacs in #5017
• replication: replica fail early on getters with no-retry errors by @moogacs in #5031
• backup:scheduler: compare and error based on major versions only by @antas-marcin in #5039

Full Changelog: v1.23.15...v1.23.16

Breaking Changes

none

New Features

none

Fixes

• fix prevent empty segment generation by @jeroiraz in #4893
• raft: refactor db opening & logging by @moogacs in #4875
• raft: always apply local schema before db by @moogacs in #4876
• raft: open db on service.Open instead on apply by @moogacs in #4878
• raft: remove schemaOnly and rename lastAppliedIndexOnStart by @moogacs in #4882
• improve log messages to apply to also dump cmd related output by @reyreaud-l in #4895
• Ensure classCache is added to the context before validate objects call by @tsmith023 in #4902
• Fix batch delete proto Java option setting by @antas-marcin in #4903
• Octoai header fixes + image generation by @dirkkul in #4906
• chore: warn with older partially written compacted segments by @jeroiraz in #4908
• Re-add schemaOnly parameter when applying raft log entries and ensure we reloadDB on catching up by @reyreaud-l in #4897
• Support raft rf scale (part 2) by @aliszka in #4913
• RAFT: fix data race on registering call backs and atomic lastAppliedI… by @moogacs in #4914
• RAFT: support back RF scale increase by @moogacs in #4856
• Replace namedLocks with KeyLocker by @aliszka in #4919
• Fix generative-aws module, add support for all Bedrock models by @antas-marcin in #4918
• Migrate text2vec-huggingface module to batching by @dirkkul in #4894
• backup:scheduler: compare and error based on major versions only by @moogacs in #4925
• Improve error messages on failed get/init shard by @aliszka in #4923
• fix race in tombstone cleanup by @asdine in #4910
• fix: skipping deleted nodes when finding entry point for node by @jeroiraz in #4807
• Add support for Google Gemini 1.5 models by @antas-marcin in #4929
• Remove unsearchable properties by @donomii in #4930
• Fix cluster remote aggregate hybrid neartext by @donomii in #4926
• Update VoyageAI vectorizer available model names by @fzowl in #4933
• Optimize locking for HNSW flat search by @trengrj in #4938
• Improved error message when autotenant expected to work with single object insertion by @parkerduckworth in #4948
• fix: special case when overwriting object which is locally deleted by @jeroiraz in #4950
• Do not validate the model for openai with a different base url by @dirkkul in #4951
• RAFT: fix race between apply for classes in the schema and report ready only after db reloaded by @moogacs in #4946
• Fix govulnchecks by @antas-marcin in #4958
• Add retry on failure in CI for tests instability by @antas-marcin in #4961
• fix deadlock by avoiding recursive locking in meta_class by @reyreaud-l in #4959
• Add support for separate headers for Google Studio and Vertex AI by @antas-marcin in #4962
• add missing wait for schema version on vector.PutObject by @moogacs in #4963

Full Changelog: v1.25.0...v1.25.1

Breaking Changes

none

New Features

none

Fixes

• Update VoyageAI vectorizer available model names by @fzowl in #4933
• fix: special case when overwriting object which is locally deleted by @jeroiraz in #4950
• Fix govulnchecks by @antas-marcin in #4958
• Add support for separate headers for Google Studio and Vertex AI by @antas-marcin in #4962

Full Changelog: v1.24.13...v1.24.14

Breaking Changes

none

New Features

none

Fixes

• fix: special case when overwriting object which is locally deleted by @jeroiraz in #4950
• Fix govulnchecks by @antas-marcin in #4958

Full Changelog: v1.23.14...v1.23.15