Skip to content

Various Sanity Studio plugins for integrating with the embeddings index API

License

Notifications You must be signed in to change notification settings

sanity-io/embeddings-index-ui

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Sanity Embeddings Index UI

Using this feature requires Sanity to send data to OpenAI.com, and Pinecone.io for storing vector interpretations of documents.

Sanity Studio v3 plugins that interact with the /embeddings-index HTTP API.

The Embeddings Index API enables the creation, management, and search of named embeddings vector indexes.

An embeddings index contains embeddings for all Sanity documents matching a configured GROQ filter in a dataset. A GROQ projection is applied to matching documents before vectorization.

You can query indexes using semantic text search to obtain a list of matching document IDs sorted by relevance.

When an index is first created, all documents matching the configured filter are synced into the index. Creating an index can take time, depending on the number of existing documents and the indexer load.

For a CLI alternative, check out the Embeddings Index CLI package.

Installation

npm install @sanity/embeddings-index-ui

@sanity/embeddings-index-ui contains the following Sanity Studio plugins:

For more information about using the plugins, see the relevant sections below.

Semantic reference search input

image

The Embeddings Index UI ships with a Semantic reference search input component. This enables you to search for references using natural language and to retrieve documents based on semantic meaning rather than exact string matches.

Usage

You can add the semantic reference search input by importing and adding embeddingsIndexReferenceInput as a plugin to sanity.config.ts (or .js):

import {defineConfig} from 'sanity'
import {embeddingsIndexReferenceInput} from '@sanity/embeddings-index-ui'

export default defineConfig({
  //...
  plugins: [embeddingsIndexReferenceInput()],
})

Then, enable semantic search using options.embeddingsIndex on reference fields. Example of a default configuration for a reference field:

defineField({
  name: 'myField',
  type: 'reference',
  to: [{type: 'myType'}], // The type(s) of document(s) to include
  options: {
    embeddingsIndex: {
      indexName: 'my-index', // Name of the embeddings index
      maxResults: 10, // Maximum number of returned results per request. Default: 10
      searchMode: 'embeddings' // Sets default search mode for the field. Enables toggling between 'embeddings' (semantic search) and 'default' (default search based on GROQ filter)
    }
  }
})

Setting options.embeddings.indexName on a reference field enables searching into the named index.

Note: the search uses to types as a filter for the index. Therefore, the types that the the reference field expects must exist in the index: the GROQ query specified in the embeddings index filter must include one or more documents that are relevant to the reference field.

Caveats: the semantic search functionality does not honor options.filter.

Default embeddings index configuration

You can enable a default configuration for the reference inputs through the plugin configuration.

Example:

import {defineConfig} from 'sanity'
import {embeddingsIndexReferenceInput} from '@sanity/embeddings-index-ui'

export default defineConfig({
  //...
  plugins: [embeddingsIndexReferenceInput({
    indexName: 'my-index', // Inputs use 'my-index' as the default index
    maxResults: 15, // Inputs return max. 15 results per request
    searchMode: 'embeddings' // Semantic search is the default search mode
  })],
})

If you assign a default indexName to the plugin, you can also enable embeddings search by setting options.embeddingsIndex: true for a reference field:

defineField({
  name: 'myField',
  type: 'reference',
  to: [{type: 'myType'}],
  options: {
    embeddingsIndex: true
  }
})

Embeddings Index API dashboard for Sanity Studio

A UI alternative to the Embeddings Index CLI to manage embeddings indexes in a Studio dashboard. It also lets you test semantic search on the indexes.

image

Usage

Add embeddingsIndexDashboard as a plugin to sanity.config.ts (or .js):

import {defineConfig} from 'sanity'
import {embeddingsIndexDashboard} from '@sanity/embeddings-index-ui'

export default defineConfig({
  //...
  plugins: [    
    process.env.NODE_ENV === 'development'
    ? embeddingsIndexDashboard()
    : {name: 'embeddings-index-dashboard-disabled'}
  ],
})

This adds the Embeddings Index API tool to the studio navigation bar, but only when the studio is running in developer mode (localhost).

If you want to enable the tool based on user access roles:

import {defineConfig} from 'sanity'
import {embeddingsIndexDashboard} from '@sanity/embeddings-index-ui'

export default defineConfig({
  //...
  plugins: [embeddingsIndexDashboard()],

  tools: (prev, context) => {
    const enabledForRoles = ['developer']
    const canManageEmbeddingsIndex = context.currentUser?.roles
      .map((role) => role.name)
      .some((roleName) => enabledForRoles.includes(roleName))
    return canManageEmbeddingsIndex ? prev : prev.filter((tool) => tool.name !== 'embeddings-index')
  },
})

License

MIT © Sanity

Develop and test

This plugin uses @sanity/plugin-kit with default configuration for build and watch scripts.

See Testing a plugin in Sanity Studio on how to run this plugin with hot reload in the studio.

Release new version

Run "CI & Release" workflow. Make sure to select the main branch and check "Release new version".

Semantic release will only release on configured branches, so it is safe to run release on any branch.