Feast

Test and debug Feast, a feature store for managing machine learning data. Run focused unit and integration tests, add tests for components such as registries or online stores, inspect live registry state, and investigate common runtime and retrieval problems.

• Running a focused Feast test after a code change
• Writing unit tests for a new online store component
• Investigating stale registry metadata

Understand how Feast components fit together and how data moves through the system. Trace operations such as applying definitions, materializing features, and retrieving online data, and locate the relevant code when exploring or changing the codebase.

• Tracing what happens during `feast apply`
• Understanding how registry metadata is stored
• Following feature data into an online store

Prepare to contribute to Feast by setting up the development environment, running tests and linters, and finding your way around the project structure. The guide also explains documentation placement and practical pull request conventions.

• Setting up a local Feast development environment
• Running tests and lint checks before a change
• Finding where a new store implementation belongs

Use this guide to define Feast entities and feature views, configure a feature store, and retrieve online or historical features. It also explains CLI commands and vector search, which finds similar documents for RAG pipelines.

• Defining entities, data sources, and feature views
• Configuring feature_store.yaml and registering definitions
• Retrieving online features or historical training data