Merge branch 'master' into docs/add-castai-contributor-logo

Author: Lukas Valatka

CHANGELOG.md

@@ -1,5 +1,33 @@
 # Changelog
 
+# [0.56.0](https://github.com/feast-dev/feast/compare/v0.55.0...v0.56.0) (2025-10-27)
+
+
+### Bug Fixes
+
+* Add mode field to Transformation proto for proper serialization ([2390d2e](https://github.com/feast-dev/feast/commit/2390d2ea654e299fc74f697212542b755f3b4938))
+* Date wise remote offline store historical data retrieval  ([#5686](https://github.com/feast-dev/feast/issues/5686)) ([949ba3d](https://github.com/feast-dev/feast/commit/949ba3dae420f82f493018113d1fd6de9e130a56))
+* Fix STRING type handling in on-demand feature views ([#5669](https://github.com/feast-dev/feast/issues/5669)) ([dfbb743](https://github.com/feast-dev/feast/commit/dfbb7433f059e6f0d1d4ef6c0ef65b63dac1c1ff))
+* Fixed torch install issue in CI ([366e5a8](https://github.com/feast-dev/feast/commit/366e5a8c8f8093eda840b667849c6d2e45fa56bb))
+* ODFV not getting counted in resource count ([1d640b6](https://github.com/feast-dev/feast/commit/1d640b6c8136c47a78887e4490a5b7ae677b7c99))
+* Skip tag updates if user do not have permissions ([#5673](https://github.com/feast-dev/feast/issues/5673)) ([0a951ce](https://github.com/feast-dev/feast/commit/0a951ce8d7f9b31490fa279339eacd444d2d2434))
+
+
+### Features
+
+* Add document of Go feature server. ([#5697](https://github.com/feast-dev/feast/issues/5697)) ([cbd1dde](https://github.com/feast-dev/feast/commit/cbd1dde9a0a6e5a3ec7e3520b6613d3818bcd842))
+* Add flexible commandArgs support for complete Feast CLI control ([#5678](https://github.com/feast-dev/feast/issues/5678)) ([6414924](https://github.com/feast-dev/feast/commit/64149246c1925e9f3dcac60d9ab629225c232261))
+* Add HDFS as a feature registry ([#5655](https://github.com/feast-dev/feast/issues/5655)) ([4c65872](https://github.com/feast-dev/feast/commit/4c65872ee6cf7e14ed14c8a8a7e141126027e575))
+* Add nodeSelector to service config ([#5675](https://github.com/feast-dev/feast/issues/5675)) ([9728cde](https://github.com/feast-dev/feast/commit/9728cde4d3cf4d22a790d3a3af2eba705b7a56d3))
+* Add OTEL based observability to the Go Feature Server ([#5685](https://github.com/feast-dev/feast/issues/5685)) ([f4afdad](https://github.com/feast-dev/feast/commit/f4afdad27c7fe92e9778e29ad08e4b227a3c17a4))
+* Added health endpoint for the UI ([#5665](https://github.com/feast-dev/feast/issues/5665)) ([3aec5d5](https://github.com/feast-dev/feast/commit/3aec5d5fd24540d10f07e79c081c2658ca35678c))
+* Added kuberay support ([e0b698d](https://github.com/feast-dev/feast/commit/e0b698d7b8733c8177ca053bc89defb01ebeb538))
+* Added support for filtering multi-projects ([#5688](https://github.com/feast-dev/feast/issues/5688)) ([eb0a86e](https://github.com/feast-dev/feast/commit/eb0a86eb81defb5ccb2407a0d5f2b2425bcb61c1))
+* Batch Embedding at scale for RAG with Ray ([cc2a46d](https://github.com/feast-dev/feast/commit/cc2a46d54c413ed52a9bf568588dd06096592c1f))
+* Optimize SQL entity handling without creating temporary tables ([#5695](https://github.com/feast-dev/feast/issues/5695)) ([aa2c838](https://github.com/feast-dev/feast/commit/aa2c8386253181145c4f314187e0873b96b2be59))
+* Support aggregation in odfv ([#5666](https://github.com/feast-dev/feast/issues/5666)) ([564e965](https://github.com/feast-dev/feast/commit/564e9651dabea5458a77a8889920749cb1a6a5ed))
+* Support cache_mode for registries ([021e9ea](https://github.com/feast-dev/feast/commit/021e9ea759bfee0292c5f7c804119ed9a15d6a58))
+
 # [0.55.0](https://github.com/feast-dev/feast/compare/v0.54.0...v0.55.0) (2025-10-14)
 
 

docs/README.md

@@ -52,6 +52,11 @@ Feast helps ML platform/MLOps teams with DevOps experience productionize real-ti
 
 * *For AI Engineers*: Feast provides a platform designed to scale your AI applications by enabling seamless integration of richer data and facilitating fine-tuning. With Feast, you can optimize the performance of your AI models while ensuring a scalable and efficient data pipeline.
 
+
+![](assets/feast_persona_diagram.png)
+
+
+
 ## What Feast is not?
 
 ### Feast is not

docs/assets/feast_persona_diagram.png

@@ -35,10 +35,40 @@ When defining an ODFV, you can specify the transformation mode using the `mode`
 
 ### Singleton Transformations in Native Python Mode
 
-Native Python mode supports transformations on singleton dictionaries by setting `singleton=True`. This allows you to 
-write transformation functions that operate on a single row at a time, making the code more intuitive and aligning with 
+Native Python mode supports transformations on singleton dictionaries by setting `singleton=True`. This allows you to
+write transformation functions that operate on a single row at a time, making the code more intuitive and aligning with
 how data scientists typically think about data transformations.
 
+## Aggregations
+
+On Demand Feature Views support aggregations that compute aggregate statistics over groups of rows. When using aggregations, data is grouped by entity columns (e.g., `driver_id`) and aggregated before being passed to the transformation function.
+
+**Important**: Aggregations and transformations are mutually exclusive. When aggregations are specified, they replace the transformation function.
+
+### Usage
+
+```python
+from feast import Aggregation
+from datetime import timedelta
+
+@on_demand_feature_view(
+    sources=[driver_hourly_stats_view],
+    schema=[
+        Field(name="total_trips", dtype=Int64),
+        Field(name="avg_rating", dtype=Float64),
+    ],
+    aggregations=[
+        Aggregation(column="trips", function="sum"),
+        Aggregation(column="rating", function="mean"),
+    ],
+)
+def driver_aggregated_stats(inputs):
+    # No transformation function needed when using aggregations
+    pass
+```
+
+Aggregated columns are automatically named using the pattern `{function}_{column}` (e.g., `sum_trips`, `mean_rating`).
+
 ## Example
 See [https://github.com/feast-dev/on-demand-feature-views-demo](https://github.com/feast-dev/on-demand-feature-views-demo) for an example on how to use on demand feature views.
 

docs/reference/beta-on-demand-feature-view.md

@@ -62,11 +62,22 @@ batch_engine:
 | `max_parallelism_multiplier` | int | 2 | Parallelism as multiple of CPU cores |
 | `target_partition_size_mb` | int | 64 | Target partition size (MB) |
 | `window_size_for_joins` | string | "1H" | Time window for distributed joins |
-| `ray_address` | string | None | Ray cluster address (None = local Ray) |
+| `ray_address` | string | None | Ray cluster address (triggers REMOTE mode) |
+| `use_kuberay` | boolean | None | Enable KubeRay mode (overrides ray_address) |
+| `kuberay_conf` | dict | None | **KubeRay configuration dict** with keys: `cluster_name` (required), `namespace` (default: "default"), `auth_token`, `auth_server`, `skip_tls` (default: false) |
+| `enable_ray_logging` | boolean | false | Enable Ray progress bars and logging |
 | `enable_distributed_joins` | boolean | true | Enable distributed joins for large datasets |
 | `staging_location` | string | None | Remote path for batch materialization jobs |
-| `ray_conf` | dict | None | Ray configuration parameters |
-| `execution_timeout_seconds` | int | None | Timeout for job execution in seconds |
+| `ray_conf` | dict | None | Ray configuration parameters (memory, CPU limits) |
+
+### Mode Detection Precedence
+
+The Ray compute engine automatically detects the execution mode:
+
+1. **Environment Variables** → KubeRay mode (if `FEAST_RAY_USE_KUBERAY=true`)
+2. **Config `kuberay_conf`** → KubeRay mode
+3. **Config `ray_address`** → Remote mode
+4. **Default** → Local mode
 
 ## Usage Examples
 

docs/reference/compute-engine/ray.md

@@ -20,8 +20,8 @@ from feast.infra.offline_stores.contrib.trino_offline_store.trino_source import
 )
 
 driver_hourly_stats = TrinoSource(
-    event_timestamp_column="event_timestamp",
-    table_ref="feast.driver_stats",
+    timestamp_field="event_timestamp",
+    table="feast.driver_stats",
     created_timestamp_column="created",
 )
 ```

docs/reference/data-sources/trino.md

@@ -0,0 +1,54 @@
+# Go feature server (Alpha)
+
+## Overview
+The Go feature server is an HTTP/gRPC endpoint that serves features. It is written in Go.
+
+## Configuration of `feature_store.yaml`
+The current Go feature server needs a Python based feature Transformation service support. Please refer to the following code as an example:
+```
+# -*- coding: utf-8 -*-
+from feast.feature_store import FeatureStore
+
+
+def main():
+    # Init the Feature Store
+    store = FeatureStore(repo_path="./feature_repo/")
+
+    # Start the feature transformation server
+    # default port is 6569
+    store.serve_transformations(6569)
+
+if __name__ == "__main__":
+    main()
+```
+At the same time, we need to configure the `feature_store.yaml` as following:
+
+```
+...
+entity_key_serialization_version: 3
+feature_server:
+    type: local
+    transformation_service_endpoint: "localhost:6569"
+...
+```
+## Supported APIs
+Here is the list of supported APIs:
+| Method | API | Comment |
+|:---: | :---: | :---: |
+| POST   | /get-online-features | Retrieve features of one or many entities |
+| GET    | /health              | Status of the Go Feature Server |
+
+## OTEL based Observability
+The Go feature server support [OTEL](https://opentelemetry.io/) based Observabilities.
+To enable it, we need to set the global env `ENABLE_OTEL_TRACING` to `"true"` (as a string type!) in the container or your local OS.
+```
+export ENABLE_OTEL_TRACING='true'
+```
+There are example OTEL infra setup under the `/go/infra/docker/otel` folder.
+
+## Demo
+Please check the Reference[2] for a local demo of Go feature server. If you want to see a real world example of applying Go feature server in Production, please check Reference[1].
+
+## Reference
+1. [Expedia Group's Go Feature Server Implementation (in Production)](https://github.com/EXPEbdodla/feast)
+2. [A Go Feature server demo from Feast](https://github.com/feast-dev/feast-credit-score-local-tutorial)

docs/reference/feature-servers/go-feature-server.md

@@ -9,10 +9,11 @@ The Ray offline store is a data I/O implementation that leverages [Ray](https://
 
 The Ray offline store provides:
 - Ray-based data reading from file sources (Parquet, CSV, etc.)
-- Support for both local and distributed Ray clusters
+- Support for local, remote, and KubeRay (Kubernetes-managed) clusters
 - Integration with various storage backends (local files, S3, GCS, HDFS)
 - Efficient data filtering and column selection
 - Timestamp-based data processing with timezone awareness
+- Enterprise-ready KubeRay cluster support via CodeFlare SDK
 
 
 ## Functionality Matrix
@@ -59,9 +60,15 @@ For complex feature processing, historical feature retrieval, and distributed jo
 
 ## Configuration
 
-The Ray offline store can be configured in your `feature_store.yaml` file. Below are two main configuration patterns:
+The Ray offline store can be configured in your `feature_store.yaml` file. It supports **three execution modes**:
 
-### Basic Ray Offline Store
+1. **LOCAL**: Ray runs locally on the same machine (default)
+2. **REMOTE**: Connects to a remote Ray cluster via `ray_address`
+3. **KUBERAY**: Connects to Ray clusters on Kubernetes via CodeFlare SDK
+
+### Execution Modes
+
+#### Local Mode (Default)
 
 For simple data I/O operations without distributed processing:
 
@@ -72,7 +79,44 @@ provider: local
 offline_store:
     type: ray
     storage_path: data/ray_storage        # Optional: Path for storing datasets
-    ray_address: localhost:10001          # Optional: Ray cluster address
+```
+
+#### Remote Ray Cluster
+
+Connect to an existing Ray cluster:
+
+```yaml
+offline_store:
+    type: ray
+    storage_path: s3://my-bucket/feast-data
+    ray_address: "ray://my-cluster.example.com:10001"
+```
+
+#### KubeRay Cluster (Kubernetes)
+
+Connect to Ray clusters on Kubernetes using CodeFlare SDK:
+
+```yaml
+offline_store:
+    type: ray
+    storage_path: s3://my-bucket/feast-data
+    use_kuberay: true
+    kuberay_conf:
+        cluster_name: "feast-ray-cluster"
+        namespace: "feast-system"
+        auth_token: "${RAY_AUTH_TOKEN}"
+        auth_server: "https://api.openshift.com:6443"
+        skip_tls: false
+    enable_ray_logging: false
+```
+
+**Environment Variables** (alternative to config file):
+```bash
+export FEAST_RAY_USE_KUBERAY=true
+export FEAST_RAY_CLUSTER_NAME=feast-ray-cluster
+export FEAST_RAY_AUTH_TOKEN=your-token
+export FEAST_RAY_AUTH_SERVER=https://api.openshift.com:6443
+export FEAST_RAY_NAMESPACE=feast-system
 ```
 
 ### Ray Offline Store + Compute Engine
@@ -175,8 +219,29 @@ batch_engine:
 |--------|------|---------|-------------|
 | `type` | string | Required | Must be `feast.offline_stores.contrib.ray_offline_store.ray.RayOfflineStore` or `ray` |
 | `storage_path` | string | None | Path for storing temporary files and datasets |
-| `ray_address` | string | None | Address of the Ray cluster (e.g., "localhost:10001") |
+| `ray_address` | string | None | Ray cluster address (triggers REMOTE mode, e.g., "ray://host:10001") |
+| `use_kuberay` | boolean | None | Enable KubeRay mode (overrides ray_address) |
+| `kuberay_conf` | dict | None | **KubeRay configuration dict** with keys: `cluster_name` (required), `namespace` (default: "default"), `auth_token`, `auth_server`, `skip_tls` (default: false) |
+| `enable_ray_logging` | boolean | false | Enable Ray progress bars and verbose logging |
 | `ray_conf` | dict | None | Ray initialization parameters for resource management (e.g., memory, CPU limits) |
+| `broadcast_join_threshold_mb` | int | 100 | Size threshold for broadcast joins (MB) |
+| `enable_distributed_joins` | boolean | true | Enable distributed joins for large datasets |
+| `max_parallelism_multiplier` | int | 2 | Parallelism as multiple of CPU cores |
+| `target_partition_size_mb` | int | 64 | Target partition size (MB) |
+| `window_size_for_joins` | string | "1H" | Time window for distributed joins |
+
+#### Mode Detection Precedence
+
+The Ray offline store automatically detects the execution mode using the following precedence:
+
+1. **Environment Variables** (highest priority)
+   - `FEAST_RAY_USE_KUBERAY`, `FEAST_RAY_CLUSTER_NAME`, etc.
+2. **Config `kuberay_conf`**
+   - If present → KubeRay mode
+3. **Config `ray_address`**
+   - If present → Remote mode
+4. **Default**
+   - Local mode (lowest priority)
 
 #### Ray Compute Engine Options
 
@@ -385,6 +450,8 @@ job.persist(hdfs_storage, allow_overwrite=True)
 
 ### Using Ray Cluster
 
+#### Standard Ray Cluster
+
 To use Ray in cluster mode for distributed data access:
 
 1. Start a Ray cluster:
@@ -406,6 +473,53 @@ offline_store:
 ray start --address='head-node-ip:10001'
 ```
 
+#### KubeRay Cluster (Kubernetes)
+
+To use Feast with Ray clusters on Kubernetes via CodeFlare SDK:
+
+**Prerequisites:**
+- KubeRay cluster deployed on Kubernetes
+- CodeFlare SDK installed: `pip install codeflare-sdk`
+- Access credentials for the Kubernetes cluster
+
+**Configuration:**
+
+1. Using configuration file:
+```yaml
+offline_store:
+    type: ray
+    use_kuberay: true
+    storage_path: s3://my-bucket/feast-data
+    kuberay_conf:
+        cluster_name: "feast-ray-cluster"
+        namespace: "feast-system"
+        auth_token: "${RAY_AUTH_TOKEN}"
+        auth_server: "https://api.openshift.com:6443"
+        skip_tls: false
+    enable_ray_logging: false
+```
+
+2. Using environment variables:
+```bash
+export FEAST_RAY_USE_KUBERAY=true
+export FEAST_RAY_CLUSTER_NAME=feast-ray-cluster
+export FEAST_RAY_AUTH_TOKEN=your-k8s-token
+export FEAST_RAY_AUTH_SERVER=https://api.openshift.com:6443
+export FEAST_RAY_NAMESPACE=feast-system
+export FEAST_RAY_SKIP_TLS=false
+
+# Then use standard Feast code
+python your_feast_script.py
+```
+
+**Features:**
+- The CodeFlare SDK handles cluster connection and authentication
+- Automatic TLS certificate management
+- Authentication with Kubernetes clusters
+- Namespace isolation
+- Secure communication between client and Ray cluster
+- Automatic cluster discovery
+
 ### Data Source Validation
 
 The Ray offline store validates data sources to ensure compatibility: