feast-dev/feast

Join us on Slack!

👋👋👋 Come say hi on Slack!

Check out our DeepWiki!

Overview

Feast (Feature Store) is an open source feature store for machine learning. Feast is the fastest path to manage existing infrastructure to productionize analytic data for model training and online inference.

Feast allows ML platform teams to:

• Make features consistently available for training and serving by managing an offline store (to process historical data for scale-out batch scoring or model training), a low-latency online store (to power real-time prediction), and a battle-tested feature server (to serve pre-computed features online).
• Avoid data leakage by generating point-in-time correct feature sets so data scientists can focus on feature engineering rather than debugging error-prone dataset joining logic. This ensure that future feature values do not leak to models during training.
• Decouple ML from data infrastructure by providing a single data access layer that abstracts feature storage from feature retrieval, ensuring models remain portable as you move from training models to serving models, from batch models to realtime models, and from one data infra system to another.

Please see our documentation for more information about the project.

📐 Architecture

The above architecture is the minimal Feast deployment. Want to run the full Feast on Snowflake/GCP/AWS? Click here.

🐣 Getting Started

1. Install Feast

pip install feast

2. Create a feature repository

feast init my_feature_repo
cd my_feature_repo/feature_repo

3. Register your feature definitions and set up your feature store

feast apply

4. Explore your data in the web UI (experimental)

feast ui

5. Build a training dataset

from feast import FeatureStore
import pandas as pd
from datetime import datetime

entity_df = pd.DataFrame.from_dict({
    "driver_id": [1001, 1002, 1003, 1004],
    "event_timestamp": [
        datetime(2021, 4, 12, 10, 59, 42),
        datetime(2021, 4, 12, 8,  12, 10),
        datetime(2021, 4, 12, 16, 40, 26),
        datetime(2021, 4, 12, 15, 1 , 12)
    ]
})

store = FeatureStore(repo_path=".")

training_df = store.get_historical_features(
    entity_df=entity_df,
    features = [
        'driver_hourly_stats:conv_rate',
        'driver_hourly_stats:acc_rate',
        'driver_hourly_stats:avg_daily_trips'
    ],
).to_df()

print(training_df.head())

# Train model
# model = ml.fit(training_df)

            event_timestamp  driver_id  conv_rate  acc_rate  avg_daily_trips
0 2021-04-12 08:12:10+00:00       1002   0.713465  0.597095              531
1 2021-04-12 10:59:42+00:00       1001   0.072752  0.044344               11
2 2021-04-12 15:01:12+00:00       1004   0.658182  0.079150              220
3 2021-04-12 16:40:26+00:00       1003   0.162092  0.309035              959

6. Load feature values into your online store

Option 1: Incremental materialization (recommended)

CURRENT_TIME=$(date -u +"%Y-%m-%dT%H:%M:%S")
feast materialize-incremental $CURRENT_TIME

Option 2: Full materialization with timestamps

CURRENT_TIME=$(date -u +"%Y-%m-%dT%H:%M:%S")
feast materialize 2021-04-12T00:00:00 $CURRENT_TIME

Option 3: Simple materialization without timestamps

feast materialize --disable-event-timestamp

The --disable-event-timestamp flag allows you to materialize all available data (from 1970-01-01 up to the current UTC time) without specifying start and end timestamps. Rows keep the event timestamps from the source.

Materializing feature view driver_hourly_stats from 2021-04-14 to 2021-04-15 done!

7. Read online features at low latency

from pprint import pprint
from feast import FeatureStore

store = FeatureStore(repo_path=".")

feature_vector = store.get_online_features(
    features=[
        'driver_hourly_stats:conv_rate',
        'driver_hourly_stats:acc_rate',
        'driver_hourly_stats:avg_daily_trips'
    ],
    entity_rows=[{"driver_id": 1001}]
).to_dict()

pprint(feature_vector)

# Make prediction
# model.predict(feature_vector)

{
    "driver_id": [1001],
    "driver_hourly_stats__conv_rate": [0.49274],
    "driver_hourly_stats__acc_rate": [0.92743],
    "driver_hourly_stats__avg_daily_trips": [72]
}

📦 Functionality and Roadmap

The list below contains the functionality that contributors are planning to develop for Feast.

• We welcome contribution to all items in the roadmap!

• Natural Language Processing

  • Vector Search (Alpha release. See RFC)
  • Enhanced Feature Server and SDK for native support for NLP

• Data Sources

  • Chronon source (contrib)
  • Snowflake source
  • Redshift source
  • BigQuery source
  • Parquet file source
  • Azure Synapse + Azure SQL source (contrib plugin)
  • Hive (community plugin)
  • Postgres (contrib plugin)
  • Spark (contrib plugin)
  • Couchbase (contrib plugin)
  • Athena (contrib plugin)
  • Clickhouse (contrib plugin)
  • Oracle (contrib plugin)
  • MongoDB (contrib plugin)
  • Ray source (contrib plugin)
  • Kafka / Kinesis sources (via push support into the online store)

• Offline Stores

  • Chronon (contrib)
  • Snowflake
  • Redshift
  • BigQuery
  • DuckDB
  • Dask
  • Remote
  • Azure Synapse + Azure SQL (contrib plugin)
  • Hive (community plugin)
  • Postgres (contrib plugin)
  • Trino (contrib plugin)
  • Spark (contrib plugin)
  • Couchbase (contrib plugin)
  • Athena (contrib plugin)
  • Clickhouse (contrib plugin)
  • Ray (contrib plugin)
  • Oracle (contrib plugin)
  • MongoDB (contrib plugin)
  • Hybrid
  • Custom offline store support

• Online Stores

  • Chronon
  • Snowflake
  • DynamoDB
  • Redis
  • Dragonfly
  • Datastore
  • Bigtable
  • SQLite
  • Remote
  • Postgres
  • HBase
  • Cassandra / AstraDB
  • ScyllaDB
  • MySQL
  • Hazelcast
  • Elasticsearch
  • SingleStore
  • Couchbase
  • MongoDB
  • Aerospike
  • Qdrant (vector store)
  • Milvus (vector store)
  • Faiss (vector store)
  • Hybrid
  • Azure Cache for Redis (community plugin)
  • Custom online store support

• Feature Engineering

  • On-demand Transformations (On Read) (Beta release. See RFC)
  • Streaming Transformations (Alpha release. See RFC)
  • Batch transformation (In progress. See RFC)
  • On-demand Transformations (On Write) (Beta release. See GitHub Issue)

• Streaming

  • Custom streaming ingestion job support
  • Push based streaming data ingestion to online store
  • Push based streaming data ingestion to offline store

• Deployments

  • AWS Lambda (Alpha release. See RFC)
  • Kubernetes (See guide)

• Feature Serving

  • Python Client
  • Python feature server
  • Feast Operator (alpha)
  • Java feature server (alpha)
  • Go feature server (alpha)
  • Offline Feature Server (alpha)
  • Registry server (alpha)

• Data Quality Management (See RFC)

  • Feature Quality Monitoring — built-in metrics, drift detection, serving log monitoring, and UI dashboard

• Feature Discovery and Governance

  • Python SDK for browsing feature registry
  • CLI for browsing feature registry
  • Model-centric feature tracking (feature services)
  • Amundsen integration (see Feast extractor)
  • DataHub integration (see DataHub Feast docs)
  • Feast Web UI (Beta release. See docs)
  • Feast Lineage Explorer

🎓 Important Resources

Please refer to the official documentation at Documentation

• Quickstart
• Tutorials
• Examples
• Running Feast with Snowflake/GCP/AWS
• Change Log

👋 Contributing

Feast is a community project and is still under active development. Please have a look at our contributing and development guides if you want to contribute to the project:

• Contribution Process for Feast
• Development Guide for Feast
• Development Guide for the Main Feast Repository

🌟 GitHub Star History

✨ Contributors

Thanks goes to these incredible people: