v0.67.0

0.67.0 (2026-10-13)

Bug Fixes

• Allow dill 0.4 so UDF serialization works on Python 3.14 (#6858) (139704f)
• Allow serving features while a feature view is MATERIALIZING (#6789) (e2d7fe2), closes #6780
• Allowlist the pinned commit SHA flagged by detect-secrets (4b81d18)
• Apply point-in-time filter correctly with the Polars backend (#6992) (01224d0)
• Authorize registry server writes before deserializing UDFs (e0ef16e)
• Automate versioned documentation publication (#6763) (5ad5592), closes #6762
• Avoid NumPy generic-unit deprecation when importing feast.type_map (5fea698)
• Bind metrics, REST registry, ui, and lineage servers dual-stack (#6886) (8d2773c)
• Bound Arrow Flight batch size in the offline server (#6896) (d169238)
• Compute std and var correctly for tiled aggregations (213a130)
• Convert PdfBytes values to protos like ImageBytes (#6968) (af20fe4)
• Decode fixed_size_list as the matching Feast list value type (#6944) (8cfe891)
• Do not log /metrics requests from the pre-fork metrics server thread (7087f1f), closes #6928
• Don't return from finally blocks in REST registry helpers (d7cb4d0)
• Encode non-finite floats in feature server JSON responses (#6930) (211ecb8)
• Feature freshness logic updated (#6765) (f0bc070)
• Fix monitoring auth (015cc84)
• Handle non-nanosecond timestamps in tiled aggregations (b8989cb)
• Improve feature list column sizing (#6793) (2573693)
• Include vector attributes in Field equality (#6934) (298c3f6)
• Increase feast container probe timeouts to avoid false restarts (e8dbd50)
• Infer the tensor type from retrieved values in to_tensor (a03e323)
• Infer timestamps for datetime columns in any unit or timezone (810391f)
• Infer vector width when vector_length is undeclared (0056565), closes #6907
• Initialize auth managers in standalone lineage server (#6873) (ecd0d38)
• Keep a zero watermark delay threshold on a Kafka source (e2e28c7)
• Keep entity rows apart in the ibis point-in-time join (a87c070)
• Keep the BatchFeatureView type and settings when copying it (#6989) (013fc57)
• Keep tiling, offline and stream engine settings when copying a StreamFeatureView (fac0f88)
• Limit tiled max and min aggregations to the window (2583c45)
• Load Milvus collections once instead of on every query (#6882) (f9e91fa)
• Make vector length validation reachable, schema-driven and vectorized (#6909) (29af8c0), closes #6907
• Map INT32_LIST, PDF_BYTES and IMAGE_BYTES type names to their ValueType (8c1e825)
• Map narrow and unsigned Arrow integer types to Feast value types (#6960) (ae74360)
• Map Spark smallint and tinyint columns to INT32 (9d42729)
• Map SQL Server bigint to INT64 and float to DOUBLE (6d15533)
• Map Trino DECIMAL columns to pyarrow decimal128 (155bc24)
• Normalize Arrow-backed dtypes from Ray Dataset.to_pandas() for pyarrow >=25 compatibility (#6916) (374f1d7)
• operator: Set grpc appProtocol on the offline Service port (f9dcfb5)
• Pin expected tensor to same device in online retrieval test (#6777) (1f2584d)
• Preserve an unset aggregation time window across a proto round-trip (2000379)
• Preserve entity row order in precomputed reads (ce37029)
• Pull the MinIO test image from pgsty/minio (fe27230)
• Pull the MinIO test image from quay (#6838) (367001b)
• Raise Trino monitoring query errors other than a missing table (6161766)
• Read Kafka and Kinesis sources without a batch source from proto (a620f5c)
• Read naive datetimes as UTC in UnixTimestamp values (bc703d9)
• Record which storage class wrote a custom_storage payload (#6988) (c0545f7)
• Refresh the SQL registry cache after deleting a data source (66c55e3)
• Refresh the SQL registry cache after deleting a permission (6b212f5)
• Reject unknown resource types and actions in the REST permission API (a5eaa36)
• Remove pip install command from Redis GLIDE docstring (66ff990)
• Remove runtime pip install hints from mlflow errors (dd9f099)
• Remove unsupported vector_len from PostgreSQL tutorial config (fcc9a85)
• Repair Chronon operator validation and Pixi CI setup (#6947) (3710f78)
• Report Dask monitoring read errors instead of no data (81673d9)
• Resolve feature view batch source name against registry (f8883d0)
• Restrict UI server CORS to trusted origins (#6986) (5e7ef3e)
• Return feature_view_metadata from the REST feature server (#6927) (bc5aeef)
• Serve aliased on demand feature views in feature services (823b92e)
• Snapshot tag filters in registry cache keys (81e1546)
• Stabilize materialization checks across hour boundaries (#6948) (acadb17)
• Stop the Couchbase online store reporting failed writes as written (b3f3c98)
• Support SQL registry dump (b60f9c7)
• Treat datetimes with a None utcoffset as naive (687bee8)
• Unwrap SecretStr tokens when building Trino auth objects (#6771) (922db5c)
• Use a valid ShuffleStrategy for the Ray DatasetContext (6259f5c)
• Use do_exchange for offline server reads to support HPA (bae07fc)
• Use group_by in Polars backend aggregation (#6965) (7b9fad2)
• Use upstream kubectl image as feast-operator CronJob default (d6cc1dc)
• Use UTC end bound and correct docs for disable_event_timestamp (#6956) (cc07c82), closes #6936

Features

• Add a non-JVM read path for Lance data sources (#6943) (e2a53bb), closes #6925 #5782 #6909
• Add a non-JVM write path for Lance data sources (#6945) (f754323), closes #6943 #6909
• Add Chronon online and offline store integrations (#6188) (dd059c6)
• Add consistency_level to the Milvus online store config (97296c1)
• Add dualStack server option to the Feast operator (#6887) (725a147), closes #6977
• Add Lance optional dependencies (#6958) (20bc993)
• Add LanceFormat table format with version/tag pinning (1196e22), closes #5650
• Add MLflow as a first-class offline DataSource for Feast (#6702) (3667772)
• Add optional GLIDE client for Redis online reads (#6870) (6e9a532)
• Add quality monitoring features for Trino offline store (#6779) (9340545)
• Add SavedDatasetLanceStorage (1118d3e)
• Add standalone MCP server as a first-class component (0efab5c), closes python-keycloak#623
• Add structured audit logging for MCP feature server (#6456) (640ac70)
• Add Trino compute engine for batch retrieval and materialization (#6898) (11a47b4)
• Batch Postgres online reads across feature views (#6832) (514bee7), closes #3259
• bigtable: Make write batch size and concurrency configurable (#6847) (5388522)
• Bucket feature count labels in latency metrics (#6811) (d9ea120)
• Expose runtime version in Feast UI (#6843) (fa8f06b)
• Make Milvus index and search params configurable (ef743ff)
• Offline write batch for IcebergSource (5be5ed0)
• OpenLineage consumer improvements - dataset versioning, column lineage, ownership (69f294e)
• operator: Add tolerations and nodeSelector to the FeatureStore CR (#6741) (03dac6f)
• Serve the MCP server dual-stack on --host :: (4fc9726)
• Support Milvus partition keys in the Milvus online store (0ed3285)
• Support token, uri and db_name in the Milvus online store (c5efecd)
• Version-pinning for FeatureService (online + offline) (#6718) (787845a)

Performance Improvements

• Avoid pandas copy in Spark reads (#6878) (4771156)
• Decouple metrics HTTP server into a dedicated child process (#6855) (0b621de)
• Reduce online serving hot-path overhead from metrics and audit logging (5ee7ee3)