Automate Feast feature management and retrieval workflows with CLI operations.
npx skills add https://github.com/feast-dev/feast --skill feast-user-guide
Please help me install this Agent Skill. Skill: feast-user-guide Source: https://github.com/feast-dev/feast/tree/main/skills Command: npx skills add https://github.com/feast-dev/feast --skill feast-user-guide
Feast users need a clear, structured guide to define features, configure feature_store.yaml, retrieve features online/offline, and use the CLI, including building RAG pipelines.
Create a Feast project, configure feature_store.yaml, and run feast apply to register your definitions.
High-intent search queries and answers about installing and using this skill.
To configure feature_store.yaml for a Feast feature store project, you define your offline and online store providers, registry path, and data sources. This configuration file is required to register definitions and run feast apply.
Feast enables online and offline feature retrieval through its CLI and SDK. You use historical retrieval for batch training datasets and online retrieval to fetch low-latency feature vectors for inference.
You define Feast entities, data sources, and feature views in Python files within your project repository. These definitions map raw data columns to structured features before running feast apply to register them.
Yes, Feast supports vector-based document retrieval pipelines. You can use on-demand feature views to process and retrieve feature vectors for Retrieval-Augmented Generation applications.
A feature view maps features directly from a defined data source, while an on-demand feature view applies custom transformations to existing features. On-demand views are useful for generating derived features during retrieval.
Yes, you must run feast materialize to load features from the offline data source into the online store. This step ensures that your latest feature values are available for low-latency online retrieval.
ivanshamaevBuild a Feast feature repository with Entities, FeatureViews, and FeatureServices.
belokonmDeploy ML models to production with infrastructure-as-code templates and monitoring.
scanadyDesign end-to-end ML pipelines with data ingestion, feature engineering, training, and deployment.
robotijnAudit feature store configurations for training-serving skew and data leakage.
