Feast Web UI

[adchia]

Is your feature request related to a problem? Please describe. Several things are harder than they should be today:

• Tag searching / filtering
• Discovery of feature views / data sources / models
• Understanding lineage of models are difficult.
• Understanding what features are available / materialized to the online store
• Managing features across projects

Describe the solution you'd like A Feast Web UI that enables discovery of features, data sources, and models across projects.

• [x] Tests checking that changes to the feast registry protos don't break the web UI (i.e. an integration test that starts up the Feast UI)
• [x] a feast cli command like feast ui that scans subdirectories for feature_store.yaml and calls feast registry-dump and pipes it to a known location for the web UI to reference. It also should produce the project-list.jsonthat we use for the landing page. This should also start the local web ui server.
• [x] Improved search bar experience that does type-ahead suggestions on tag names
• [x] release process that automatically publishes the npm package.

After initial launch

• [ ] A project level description config (e.g. in feature_store.yaml which populates the landing page of a given project in the UI and is output in the "registry" dump)
• [ ] Working across all types
• [ ] Links to e.g. feast-dev documentation (e.g a ? that shows details on hover)
• [ ] Flagging system for different UI components
• [x] Custom plugin system
• [x] Having a UI that dynamically updates + re-syncs against the registry file
• [x] Migration towards perhaps https://github.com/protobufjs/protobuf.js/ instead of zod to parse data
• [ ] There's a "statistics" page in feature views that needs data to be populated. It's hidden by default behind a flag, but once we start pulling sample data and statistics in generating the json dump, this can be re-enabled
• [ ] Cleaned up UI for the data sets that gives information about the declared expectations
• [ ] UI for the data sets that highlights validation errors in a saved dataset

Describe alternatives you've considered A clear and concise description of any alternative solutions or features you've considered.

Additional context Add any other context or screenshots about the feature request here.

[woop]

There's a "statistics" page in feature views that needs data to be populated. It's hidden by default behind a flag, but once we start pulling sample data and statistics in generating the json dump, this can be re-enabled

Have you spent time thinking about the initial stats that you want to trend here, and whether we can start producing those as part of DQM efforts?

[adchia]

Yeah cc @pyalex and @kevjumba

Simple statistics I think will go a long way. e.g. for numeric features min / max / % missing / # unique values etc. For DQM, probably also want a way to compare two data sets side by side.

I think ultimately that the best solution would actually be to enable users to arbitrarily run queries and explore the data within this UI. I'd envision that we also would want users to be able to plug in data they're interested in to this UI and/or do cool data visualizations.

[woop]

Yeah cc @pyalex and @kevjumba

I think ultimately that the best solution would actually be to enable users to arbitrarily run queries and explore the data within this UI. I'd envision that we also would want users to be able to plug in data they're interested in to this UI and/or do cool data visualizations.

That's a good point. I attribute at least a part of the success of the likes of Grafana to its ability in dealing with the rendering of charts, but not standing in the way of whatever query you want to run. Some level of "batteries included" makes sense, but ultimately we'll see the best ideas from the community.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.