search

firebase / genkit

An open source framework for building AI-powered apps with familiar code-centric patterns. Genkit makes it easy to integrate, test, and deploy sophisticated AI features to Firebase or Google Cloud.
Apache License 2.0
655 stars 89 forks source link

[JS] Add support for Vertex Vector Search to the Vertex AI plugin #448

Closed chrisraygill closed 2 months ago

chrisraygill commented 3 months ago

Is your feature request related to a problem? Please describe.

Developers who want to use Vertex Vector Search as a scalable, production-ready solution don't have a way to do so through Genkit.

Describe the solution you'd like

Make Vertex Vector Search available through the Vertex AI plugin as indexer and retriever components.

Describe alternatives you've considered

Developers can define their own indexer and retriever components, but it's more convenient to have it exposed through the plugin as a first class experience.

Additional context

More info on Vertex Vector Search available here: https://cloud.google.com/vertex-ai/docs/vector-search/overview

cabljac commented 3 months ago

@chrisraygill I thought I'd discuss this here, let me know if it should be somewhere else:

Unless I've misunderstood. a Vertex AI index will store an id and the embedding, and users are expected to keep the content of the document and any metadata which corresponds to this id somewhere else (In another database, e.g BQ, to retrieve via the id)

see here for example

In Genkit we have a Document which has content and metadata, and a retriever responds with a list of documents.

Should we enforce BigQuery as the database in this feature?

cabljac commented 2 months ago

adding for reference https://github.com/firebase/genkit/pull/519