diff --git a/content/develop/ai/_index.md b/content/develop/ai/_index.md index 5acffe0dad..5b560770b8 100644 --- a/content/develop/ai/_index.md +++ b/content/develop/ai/_index.md @@ -19,9 +19,9 @@ Redis stores and indexes vector embeddings that semantically represent unstructu {{< image-card image="images/ai-brain.svg" alt="AI Redis icon" title="Use LangCache to store LLM responses" url="/develop/ai/langcache/" >}} -## Featureform +## Redis Feature Form -Use [Featureform]({{< relref "/develop/ai/featureform/" >}}) to define, manage, and serve machine learning features on top of your existing data systems. The Featureform docs cover the Python SDK workflow from provider registration through feature serving. +Use [Redis Feature Form]({{< relref "/develop/ai/featureform/" >}}) to define, manage, and serve machine learning features on top of your existing data systems. The Feature Form docs cover the Python SDK workflow from provider registration through feature serving. #### Overview diff --git a/content/develop/ai/featureform/_index.md b/content/develop/ai/featureform/_index.md index 1636912fb6..fe8f156ab0 100644 --- a/content/develop/ai/featureform/_index.md +++ b/content/develop/ai/featureform/_index.md @@ -1,21 +1,21 @@ --- -Title: Featureform +Title: Redis Feature Form alwaysopen: false categories: - docs - develop - ai -description: Build feature engineering workflows with Featureform and Redis. -linkTitle: Featureform +description: Build feature engineering workflows with Redis Feature Form. +linkTitle: Redis Feature Form hideListLinks: true weight: 60 -bannerText: Featureform is currently in preview and subject to change. To request access to the Featureform Docker image, contact your Redis account team. +bannerText: Redis Feature Form is currently in preview and subject to change. To request access to the Feature Form Docker image, contact your Redis account team. bannerChildren: true --- -Featureform helps data teams define, materialize, and serve machine learning features by using a declarative Python SDK on top of existing data systems. +Redis Feature Form helps data teams define, materialize, and serve machine learning features by using a declarative Python SDK on top of existing data systems. -Featureform works with offline systems such as Snowflake, BigQuery, and Databricks or Spark, then uses Redis as the low-latency online store for feature serving. +Feature Form works with offline systems such as Snowflake, BigQuery, and Databricks or Spark, then uses Redis as the low-latency online store for feature serving. ## Get started @@ -26,6 +26,18 @@ Featureform works with offline systems such as Snowflake, BigQuery, and Databric - [Define features and labels]({{< relref "/develop/ai/featureform/features-and-labels" >}}) - [Work with training sets and feature views]({{< relref "/develop/ai/featureform/training-sets-and-feature-views" >}}) +## Latest updates + +The latest release adds enterprise-oriented capabilities: + +- **Unified batch and streaming pipelines**: Support for tiling, backfills, and incremental updates reduces custom pipeline work. +- **Workspaces for multi-tenancy**: Isolate providers, data, authentication, and observability at the workspace level. +- **Fine-grained job control**: Planning, impact analysis, split materializations, and queue-based job management provide visibility into changes before they affect production systems. +- **Atomic DAG updates**: Manage graph-level changes atomically instead of versioning individual resources, which simplifies rollback and change history. +- **Enhanced RBAC and security**: Workspace-scoped access controls, API key pairs, a granular role model, audit logs, secret-provider improvements, mTLS, and encrypted internal transport. +- **Two-service deployment model**: A simplified deployment architecture that reduces operational complexity. +- **Redesigned dashboard**: Configure workspaces and providers directly from the UI. + ## Next steps - [Streaming features]({{< relref "/develop/ai/featureform/streaming" >}}) diff --git a/content/develop/ai/featureform/datasets-and-transformations.md b/content/develop/ai/featureform/datasets-and-transformations.md index b8a289c7b8..d025ee642b 100644 --- a/content/develop/ai/featureform/datasets-and-transformations.md +++ b/content/develop/ai/featureform/datasets-and-transformations.md @@ -1,18 +1,18 @@ --- -title: Datasets and transformations in Featureform -description: Register datasets and define SQL or DataFrame transformations for Featureform features. +title: Datasets and transformations in Redis Feature Form +description: Register datasets and define SQL or DataFrame transformations for Feature Form features. linkTitle: Datasets and transformations weight: 40 --- Datasets and transformations are the bridge between your raw source data and your feature definitions. -- datasets point Featureform at existing tables or files +- datasets point Feature Form at existing tables or files - transformations create reusable feature engineering logic on top of those datasets ## Register datasets -Register the source objects that Featureform should treat as named inputs. +Register the source objects that Feature Form should treat as named inputs. ### Warehouse tables @@ -24,7 +24,7 @@ transactions = snowflake.register_table( ) ``` -For Snowflake providers, the provider-level `database`, `schema`, and `catalog` configuration determines where Featureform resolves the table unless you override those values during registration. +For Snowflake providers, the provider-level `database`, `schema`, and `catalog` configuration determines where Feature Form resolves the table unless you override those values during registration. ### Delta tables @@ -159,7 +159,7 @@ def high_value_users(features_df): ## Accessing transformation data -For development and validation, Featureform can retrieve the output of registered datasets or transformations as data frames: +For development and validation, Feature Form can retrieve the output of registered datasets or transformations as data frames: ```python df = client.dataframe(user_transaction_features) @@ -169,7 +169,7 @@ Use this sparingly in production workflows. Its main value is validation and ite ## Apply registrations -Once datasets and transformations are defined, register them with Featureform: +Once datasets and transformations are defined, register them with Feature Form: ```python client.apply() diff --git a/content/develop/ai/featureform/features-and-labels.md b/content/develop/ai/featureform/features-and-labels.md index ec95d4edf3..a49a2ba637 100644 --- a/content/develop/ai/featureform/features-and-labels.md +++ b/content/develop/ai/featureform/features-and-labels.md @@ -1,11 +1,11 @@ --- -title: Features and labels in Featureform -description: Define entities, features, labels, and aggregate windows in Featureform. +title: Features and labels in Redis Feature Form +description: Define entities, features, labels, and aggregate windows in Feature Form. linkTitle: Features and labels weight: 50 --- -Features and labels are the core semantic objects in Featureform. They describe what you want to predict, how feature values are keyed, and how those values should be computed over time. +Features and labels are the core semantic objects in Feature Form. They describe what you want to predict, how feature values are keyed, and how those values should be computed over time. ## Entities @@ -17,7 +17,7 @@ class User: pass ``` -Attach features and labels to the entity class so Featureform can reason about keys and lineage. +Attach features and labels to the entity class so Feature Form can reason about keys and lineage. ## Define features @@ -86,7 +86,7 @@ class User: ## Point-in-time correctness -Timestamps are central to correct ML training and serving workflows. When you include the relevant event timestamp in your feature and label definitions, Featureform can align historical examples correctly instead of leaking future information into the training data. +Timestamps are central to correct ML training and serving workflows. When you include the relevant event timestamp in your feature and label definitions, Feature Form can align historical examples correctly instead of leaking future information into the training data. Use timestamps consistently when: diff --git a/content/develop/ai/featureform/overview.md b/content/develop/ai/featureform/overview.md index 71ea3d2191..daf05d15c7 100644 --- a/content/develop/ai/featureform/overview.md +++ b/content/develop/ai/featureform/overview.md @@ -1,17 +1,17 @@ --- -title: Featureform overview -description: Learn what Featureform is, who it is for, and how it fits into Redis-based ML workflows. +title: Redis Feature Form overview +description: Learn what Redis Feature Form is, who it is for, and how it fits into Redis-based ML workflows. linkTitle: Overview weight: 10 --- -Featureform is a feature engineering and feature serving workflow built for teams that want to define machine learning features in code while keeping their existing data platforms. It gives application teams a declarative Python SDK for working with providers, datasets, transformations, entities, features, labels, training sets, and online feature serving. +Redis Feature Form is a feature engineering and feature serving workflow built for teams that want to define machine learning features in code while keeping their existing data platforms. It gives application teams a declarative Python SDK for working with providers, datasets, transformations, entities, features, labels, training sets, and online feature serving. -In a typical deployment, Featureform reads or computes feature data in an offline system such as Snowflake, BigQuery, or Databricks and materializes selected features to Redis for low-latency inference. +In a typical deployment, Feature Form reads or computes feature data in an offline system such as Snowflake, BigQuery, or Databricks and materializes selected features to Redis for low-latency inference. ## Core workflow -The Featureform workflow follows a consistent progression: +The Feature Form workflow follows a consistent progression: 1. Register providers for your offline systems and your Redis online store. 2. Register datasets that represent raw inputs or curated tables. @@ -23,18 +23,30 @@ The Featureform workflow follows a consistent progression: ## Where Redis fits -Redis is the online inference store in the Featureform workflow. After you materialize a feature view, applications can retrieve feature values with low latency for prediction requests, personalization, fraud detection, recommendation systems, and similar ML use cases. +Redis is the online inference store in the Feature Form workflow. After you materialize a feature view, applications can retrieve feature values with low latency for prediction requests, personalization, fraud detection, recommendation systems, and similar ML use cases. ## Main interfaces The main user-facing interface is the Python SDK: -- `ff.Client(...)` to connect to Featureform +- `ff.Client(...)` to connect to Feature Form - provider registration methods such as `register_snowflake`, `register_spark`, and `register_redis` - decorators and builder APIs for transformations, features, and labels - materialization and serving methods such as `materialize_feature_view(...)` and `serve_feature_view(...)` -Featureform also includes a dashboard for viewing registered resources and tasks, but the Python SDK is the primary authoring interface. +Feature Form also includes a dashboard for viewing registered resources and tasks, but the Python SDK is the primary authoring interface. + +## Latest updates + +The latest release adds enterprise-oriented capabilities: + +- **Unified batch and streaming pipelines**: Support for tiling, backfills, and incremental updates reduces custom pipeline work. +- **Workspaces for multi-tenancy**: Isolate providers, data, authentication, and observability at the workspace level. +- **Fine-grained job control**: Planning, impact analysis, split materializations, and queue-based job management provide visibility into changes before they affect production systems. +- **Atomic DAG updates**: Manage graph-level changes atomically instead of versioning individual resources, which simplifies rollback and change history. +- **Enhanced RBAC and security**: Workspace-scoped access controls, API key pairs, a granular role model, audit logs, secret-provider improvements, mTLS, and encrypted internal transport. +- **Two-service deployment model**: A simplified deployment architecture that reduces operational complexity. +- **Redesigned dashboard**: Configure workspaces and providers directly from the UI. ## What to read next diff --git a/content/develop/ai/featureform/providers.md b/content/develop/ai/featureform/providers.md index e3ae6af6a3..8d99264ed4 100644 --- a/content/develop/ai/featureform/providers.md +++ b/content/develop/ai/featureform/providers.md @@ -1,33 +1,33 @@ --- -title: Connect providers in Featureform -description: Register offline providers and Redis online stores for Featureform workflows. +title: Connect providers in Redis Feature Form +description: Register offline providers and Redis online stores for Feature Form workflows. linkTitle: Providers weight: 30 --- -Use providers to connect Featureform to the systems where your data already lives. +Use providers to connect Feature Form to the systems where your data already lives. In most deployments, you'll register: - offline providers store or compute source data and transformed data - online providers serve low-latency feature values to applications -Redis is the online provider in the Featureform workflow. +Redis is the online provider in the Feature Form workflow. ## Before you begin Before you register providers, make sure you have: -- network access from Featureform to the systems you want to register +- network access from Feature Form to the systems you want to register - credentials for each provider - a clear decision about which system is authoritative for offline feature computation - a Redis deployment for online serving -Featureform's current provider APIs rely heavily on catalog configuration. For Snowflake and Databricks, make sure you choose the catalog model before you register providers or datasets. +Feature Form's current provider APIs rely heavily on catalog configuration. For Snowflake and Databricks, make sure you choose the catalog model before you register providers or datasets. ## Offline providers -Featureform supports several offline systems, including: +Feature Form supports several offline systems, including: - Snowflake - BigQuery @@ -48,7 +48,7 @@ redis = client.register_redis( ) ``` -If your deployment requires TLS or ACL configuration, use the connection options supported by your current Featureform release and align them with your Redis security standards. +If your deployment requires TLS or ACL configuration, use the connection options supported by your current Feature Form release and align them with your Redis security standards. For secured Redis deployments, you can also configure ACL usernames and TLS settings explicitly: @@ -196,7 +196,7 @@ Use either `cluster_id` or `cluster_config`, not both. ## Catalog options -Featureform supports several catalog models for offline data: +Feature Form supports several catalog models for offline data: ```python from featureform import UnityCatalog, GlueCatalog, SnowflakeCatalog, TableFormat diff --git a/content/develop/ai/featureform/quickstart.md b/content/develop/ai/featureform/quickstart.md index efd1238562..b30491f598 100644 --- a/content/develop/ai/featureform/quickstart.md +++ b/content/develop/ai/featureform/quickstart.md @@ -1,15 +1,15 @@ --- -title: Featureform quickstart +title: Redis Feature Form quickstart description: Register providers, define a feature, materialize it to Redis, and serve it. linkTitle: Quickstart weight: 20 --- -This quickstart helps you register a simple Featureform workflow and serve feature values from Redis. +This quickstart helps you register a simple Redis Feature Form workflow and serve feature values from Redis. You'll do the following: -1. Connect to Featureform. +1. Connect to Feature Form. 2. Register an offline provider and a Redis online store. 3. Register a dataset and define a transformation. 4. Define an entity and features. @@ -17,18 +17,18 @@ You'll do the following: 6. Materialize a feature view to Redis. 7. Serve feature values. -At the end, you'll have a Featureform feature view backed by Redis and a working `serve_feature_view(...)` call. +At the end, you'll have a Feature Form feature view backed by Redis and a working `serve_feature_view(...)` call. ## Before you begin Before you start, make sure you have: -- a running Featureform deployment +- a running Feature Form deployment - access to an offline system, such as Databricks or Spark - access to a Redis deployment for online serving - credentials for the systems you register in this guide -## Connect to Featureform +## Connect to Feature Form ```python import featureform as ff diff --git a/content/develop/ai/featureform/streaming.md b/content/develop/ai/featureform/streaming.md index 2f871139a2..b0baac4470 100644 --- a/content/develop/ai/featureform/streaming.md +++ b/content/develop/ai/featureform/streaming.md @@ -1,11 +1,11 @@ --- -title: Streaming features in Featureform +title: Streaming features in Redis Feature Form description: Build stream-backed features with Kafka, streaming transformations, and Redis serving. linkTitle: Streaming weight: 70 --- -Featureform supports stream-backed feature workflows for use cases that need continuously updated online values while preserving historical correctness for training. +Redis Feature Form supports stream-backed feature workflows for use cases that need continuously updated online values while preserving historical correctness for training. This page focuses on the documented SDK surface: register Kafka, define a streaming transformation, define features with `from_stream(...)`, optionally backfill from a batch source, materialize a feature view, and serve it from Redis. @@ -13,7 +13,7 @@ This page focuses on the documented SDK surface: register Kafka, define a stream Before you build streaming features, you need: -- a running Featureform deployment +- a running Feature Form deployment - a registered Redis online store - a stream-capable compute provider such as Spark or Databricks - a Kafka topic that contains the events you want to transform into features diff --git a/content/develop/ai/featureform/training-sets-and-feature-views.md b/content/develop/ai/featureform/training-sets-and-feature-views.md index 87d5b27dce..c9879ffe65 100644 --- a/content/develop/ai/featureform/training-sets-and-feature-views.md +++ b/content/develop/ai/featureform/training-sets-and-feature-views.md @@ -1,11 +1,11 @@ --- -title: Training sets and feature views in Featureform -description: Build training sets and materialize feature views to Redis with Featureform. +title: Training sets and feature views in Redis Feature Form +description: Build training sets and materialize feature views to Redis with Feature Form. linkTitle: Training sets and feature views weight: 60 --- -Training sets and feature views are where Featureform turns definitions into outputs that model builders and applications can use. +Training sets and feature views are where Feature Form turns definitions into outputs that model builders and applications can use. - training sets support model development - feature views support online inference diff --git a/content/develop/whats-new/_index.md b/content/develop/whats-new/_index.md index c66c1308e1..545a436665 100644 --- a/content/develop/whats-new/_index.md +++ b/content/develop/whats-new/_index.md @@ -25,7 +25,7 @@ weight: 10 ### Redis AI & Vectors -- Added [Featureform documentation]({{< relref "/develop/ai/featureform" >}}) for feature engineering workflows +- Added [Redis Feature Form documentation]({{< relref "/develop/ai/featureform" >}}) for feature engineering workflows - [Overview]({{< relref "/develop/ai/featureform/overview" >}}) - [Quickstart guide]({{< relref "/develop/ai/featureform/quickstart" >}}) - [Provider connections]({{< relref "/develop/ai/featureform/providers" >}})