Add Model.listSearchIndexes()

[hlebegue]

Prerequisites

• I have written a descriptive issue title
• I have searched existing issues to ensure the bug has not already been reported

Mongoose version

8.2.2

Node.js version

16.20.2

MongoDB server version

Atlas latest

Typescript version (if applicable)

No response

Description

When I use model.createSearchIndex() with an index that already exists, my node application is going down with a SIGTERM

Steps to Reproduce

Use an existing test case to create a search index, but attempt to create it twice with the same name

//just there 
const vectorModelsSupported = [
		{
			name: 'text_embedding_3_small',
			size: 1536,
			enabled: true
		},
		{ 
			name: 'text_embedding_3_small', 
			size: 1536,
			enabled: true
		}
	]

const EmbeddingSchema = new Schema({
	chunkText: { type: String, maxLength: 32764, default: "" },
	embedding_text_embedding_3_small: { type: Array, maxLength: 1536 },
	embedding_text_embedding_3_large: { type: Array, },
	embedding_text_embedding_ada_002: { type: Array, maxLength: 1536 },
	chunkLen: Number,
	chunkIndex: Number,
	upload: {
		type: ObjectId,
		ref: "upload",
	},
	chunkUrl: { type: String, maxLength: 1000, }, /*ex: #page=2 for PDF or #id for HTML*/
	created: { type: Date, default: Date.now },
	usage: { type: Number },
	status: { type: String, default: "inactive" }, // inactive, active | error, selected, deleted
})

const createSearchIndexes = async () => { 
	
	//debug("createSearchIndexes", "This is the list of indexes:", await embedding.listIndexes())
	debug("createSearchIndexes", "This is the list of indexes:", await embedding.searchIndexes)
	if (!modelsSupported) return;
	
	modelsSupported.forEach((model) =>{
		if (model.enabled){
			
			//const field =
			
			const fields = {}
			fields[`embedding_${model.name}`]=  [
				{
					"dimensions": model.size, // 1536,
					"similarity": "euclidean",
					"type": "knnVector"
				}
			]

			try{
				embedding.createSearchIndex({
					name: model.name, // "text_embedding_3_small",
					definition: {
						"mappings": {
							"fields": fields
							}
						}
					}
				)
			}
			catch (err){
				debug("createSearchIndexes", "Failed to create index for model ",model, err)
			}
		
		}

	})
}

embedding.on('index',(event) => {
	debug("Embedding", "===> This is an index event", event)
	
	createSearchIndexes()

})

Expected Behavior

the createSearchIndex should throw an error I can catch, or there should be an API that returns the search index.

The Model.listIndexes does not return the search indexes.

[IslandRhythms]

const mongoose = require('mongoose');

const { Schema } = mongoose;

//just there 
const vectorModelsSupported = [
  {
    name: 'text_embedding_3_small',
    size: 1536,
    enabled: true
  },
  { 
    name: 'text_embedding_3_small', 
    size: 1536,
    enabled: true
  }
]

const EmbeddingSchema = new Schema({
chunkText: { type: String, maxLength: 32764, default: "" },
embedding_text_embedding_3_small: { type: Array, maxLength: 1536 },
embedding_text_embedding_3_large: { type: Array, },
embedding_text_embedding_ada_002: { type: Array, maxLength: 1536 },
chunkLen: Number,
chunkIndex: Number,
upload: {
  type: mongoose.Schema.Types.ObjectId,
  ref: "upload",
},
chunkUrl: { type: String, maxLength: 1000, }, /*ex: #page=2 for PDF or #id for HTML*/
created: { type: Date, default: Date.now },
usage: { type: Number },
status: { type: String, default: "inactive" }, // inactive, active | error, selected, deleted
});

const Test = mongoose.model('Test', EmbeddingSchema);


async function run() {
  await mongoose.connect('mongodb://localhost:27017');

  vectorModelsSupported.forEach((model) =>{
		if (model.enabled){
			
			//const field =
			
			const fields = {}
			fields[`embedding_${model.name}`]=  [
				{
					"dimensions": model.size, // 1536,
					"similarity": "euclidean",
					"type": "knnVector"
				}
			]

			try{
				Test.createSearchIndex({
					name: model.name, // "text_embedding_3_small",
					definition: {
						"mappings": {
							"fields": fields
							}
						}
					}
				)
			}
			catch (err){
				console.log('err', err);
			}
		
		}

	})
}

run();

[hlebegue]

With the latest Mongoose, I was able to catch error directly treating as a promise so it didn't send a SIG TERM

  embedding.createSearchIndex({
	  name: model.name, // "text_embedding_3_small",
	  definition: {
		  "mappings": {
			  "fields": fields
			  }
		  }
  }).then( (success)=> {
	  console.log("Succeeded!", success)
  }).catch((err) => {
	  //console.log("This is the part that preventsa SIGTERM")
  })

I suppose the real issue is that I don't get the list of search indexes when I use

await embedding.listIndexes()

or accessing this property

await embedding.searchIndexes

The list returned is always empty.

Question:
How can I determine if a searchIndex already exists so I don't have to rely on creating it even if it exists?
What API should I use? If none are available, can 1 be added?

Thank you

[IslandRhythms]

Please modify my script to demonstrate your issue. Also when using code blocks, use three ` to have it appear as a code block like so:
"```"
const name = 'Test'

"```"

[vkarpov15]

The issue in the original post is that you need to await on embedding.createSearchIndex. Since createSearchIndex is async, you can't handle errors with try/catch unless you use await.

In terms of listing existing search indexes, we're currently missing an implementation of listSearchIndexes(). We will add that.

[hlebegue]

Fantastic ! Thank you for adding the Model.listSearchIndex() and respond to this issue. Best wishes