Feast v0.4.0

v0.4.0 (2019-12-28)

This is a major release which changes the way features are ingested and retrieved, for the purpose of enhancing feature discoverability and ease of use.

Change summary

• We have introduced project namespacing, which now allows users to logically group features in a way that they will be used and shared, as opposed to how they are ingested. This is the first step towards the RFC published here.
• We have also introduced async feature set registration, which improves overall performance when attempting to register several feature sets at once.

Full Changelog

Implemented enhancements:

• Edit description in feature specification to also reflect in BigQuery schema description. #239
• Allow for disabling of metrics pushing #57

Merged pull requests:

• Java SDK release script #406 (davidheryanto)
• Use fixed 'dev' revision for test-e2e-batch #395 (davidheryanto)
• Project Namespacing #393 (woop)
• (docs) Change data types to upper case to match what works with implementation #389 (david30907d)
• Remove alpha v1 from java package name #387 (khorshuheng)
• Minor bug fixes for Python SDK #383 (voonhous)
• Allow user to override job options #377 (khorshuheng)
• Add documentation to default values.yaml in Feast chart #376 (davidheryanto)
• Add support for file paths for providing entity rows during batch retrieval #375 (voonhous)
• Update sync helm chart script to ensure requirements.lock in in sync with requirements.yaml #373 (davidheryanto)
• Catch errors thrown by BQ during entity table loading #371 (zhilingc)
• Async job management #361 (zhilingc)
• Infer schema of PyArrow table directly #355 (voonhous)
• Add readiness checks for Feast services in end to end test #337 (davidheryanto)
• Create CHANGELOG.md #321 (woop)

Feast v0.3.5

v0.3.5 (2019-12-26)

Full Changelog

Merged pull requests:

• Always set destination table in BigQuery query config in Feast Batch Serving so it can handle large results #392 (davidheryanto)

Feast v0.3.4

v0.3.4 (2019-12-23)

Full Changelog

Merged pull requests:

• Make redis key creation more determinisitic #380 (zhilingc)

Feast v0.3.3

v0.3.3 (2019-12-18)

Full Changelog

Implemented enhancements:

• Added Docker Compose for Feast #272
• Added ability to check import job status and cancel job through Python SDK #194
• Added basic customer transactions example #354 (woop)

Merged pull requests:

• Added Prow jobs to automate the release of Docker images and Python SDK #369 (davidheryanto)
• Fixed installation link in README.md #368 (Jeffwan)
• Fixed Java SDK tests not actually running (missing dependencies) #366 (woop)
• Added more batch retrieval tests #357 (zhilingc)
• Python SDK and Feast Core Bug Fixes #353 (woop)
• Updated buildFeatureSets method in Golang SDK #351 (davidheryanto)
• Python SDK cleanup #348 (woop)
• Broke up queries for point in time correctness joins #347 (zhilingc)
• Exports gRPC call metrics and Feast resource metrics in Core #345 (davidheryanto)
• Fixed broken Google Group link on Community page #343 (ches)
• Ensured ImportJobTest is not flaky by checking WriteToStore metric and requesting adequate resources for testing #332 (davidheryanto)
• Added docker-compose file with Jupyter notebook #328 (khorshuheng)
• Added minimal implementation of ingesting Parquet and CSV files #327 (voonhous)

Feast v0.3.2

v0.3.2 (2019-11-29)

Full Changelog

Merged pull requests:

• Fixed incorrect BigQuery schema creation from FeatureSetSpec #340 (davidheryanto)
• Filtered out feature sets that dont share the same source #339 (zhilingc)
• Changed latency calculation method to not use Timer #338 (zhilingc)
• Moved Prometheus annotations to pod template for serving #336 (zhilingc)
• Removed metrics windowing, cleaned up step names for metrics writing #334 (zhilingc)
• Set BigQuery table time partition inside get table function #333 (zhilingc)
• Added unit test in Redis to return values with no max age set #329 (smadarasmi)
• Consolidated jobs into single steps instead of branching out #326 (zhilingc)
• Pinned Python SDK to minor versions for dependencies #322 (woop)
• Added Auto format to Google style with Spotless #317 (ches)

Feast v0.3.1

v0.3.1 (2019-11-25)

Full Changelog

Merged pull requests:

• Added Prometheus metrics to serving #316 (zhilingc)
• Changed default job metrics sink to Statsd #315 (zhilingc)
• Fixed module import error in Feast CLI #314 (davidheryanto)

Feast v0.3.0

This release makes major changes to the Feast data model by introducing Feature Sets, along with all new APIs and SDKs for both online and batch serving.

Change summary:

• Introduced Feature Sets as a concept with a new Feast Core API and Feast Serving API.
• Upgraded Python SDK to support new Feast API. Allows for management of Feast as a library or through the command line.
• Implemented a Golang SDK and Java SDK to support the new Feast Core and Feast Serving APIs.
• Added support for multi-feature set retrieval and joins.
• Added point-in-time correct retrieval for both batch and online serving.
• Added support for an external source in Kafka.
• Added job management to Feast Core to manage ingestion/population jobs to remote Feast deployments.
• Added metrics support via Prometheus and Statsd

Feast v0.1.8

Merge pull request #280 from pradithya/remove_transaction

Remove redis transaction

Feast v0.2.0

This release makes major changes to data processing in the ingestion layer of Feast. By decoupling ingestion into two phases, Feast v0.2.0 allows Feast to support decentralized serving with centralized feature management.

Change summary:

• A centralized stream is introduced in Feast for managing state for all consuming Feast deployments.
• Ingestion is decoupled and now produces data to the stream instead of starting jobs that directly ingest to stores.
• Feast serving deployments consume from this central feature stream. This consumption is handled by long running jobs.

Feast v0.1.2

This release adds filtering capability when retrieving feature values from Feast Warehouse.

Given the following entity and features names:

entity: person
feature_names:
- age
- height
- address_city
- address_country

This release allows the retrieval of feature values for person entity from Warehouse where address_city = Bandung, as opposed to retrieving all the feature values for all address_city and performing the post-filtering at the client side.

Docker images:

gcr.io/kf-feast/feast-core:0.1.2
gcr.io/kf-feast/feast-serving:0.1.2

Python SDK:

pip install feast==0.1.2