Run Milvus using Node.js

This topic describes how to run Milvus using Node.js.

1. Initialize a Node.js Project

npm init

2. Install TypeScript and Node Milvus SDK and its dependencies

npm install @zilliz/milvus2-sdk-node typescript --save

3. Download sample code HelloMilvus.ts

$ wget https://raw.githubusercontent.com/milvus-io/milvus-sdk-node/main/example/HelloMilvus.ts

4. Scan HelloMilvus.ts

This sample code does the following:

• Imports the Node.js SDK package:

import { MilvusClient, DataType, InsertReq } from "@zilliz/milvus2-sdk-node"

• Connects to the Milvus server:

const milvusClient = new MilvusClient("localhost:19530");

• Creates a collection:

const collectionName = "hello_milvus";
    const dim = "4";
    const createRes = await milvusClient.createCollection(
        {
            collection_name: collectionName,
            fields: [
                {
                    name: "count",
                    data_type: DataType.VarChar,
                    is_primary_key: true,
                    type_params: {
                      max_length: '100',
                    },
                }, 
                {
                    name: "random_value",
                    data_type: DataType.Double,
                    description: "",
                }, 
                {
                    name: "float_vector",
                    data_type: DataType.FloatVector,
                    description: "",
                    type_params: {
                      dim
                    }
                }
            ]
          }
    );


    console.log("--- Create collection ---", createRes, collectionName);

• Inserts vectors in the new collection:

const generateInsertData = function generateInsertData(
  fields: { isVector: boolean; dim?: number; name: string; isBool?: boolean }[],
  count: number) {
    const results = [];
    while (count > 0) {
      let value: any = {};
  
      fields.forEach((v) => {
        const { isVector, dim, name, isBool } = v;
        value[name] = isVector
          ? [...Array(dim)].map(() => Math.random() * 10)
          : isBool
          ? count % 2 === 0
          : count;
      });

      value["count"] = count;
      results.push(value);
      count--;
    }
    return results;
}

    const fields = [
      {
        isVector: true,
        dim: 4,
        name: "float_vector",
      },
      {
        isVector: false,
        name: "random_value",
      },
    ];
    const vectorsData = generateInsertData(fields, 1000);
  
    const params: InsertReq = {
      collection_name: collectionName,
      fields_data: vectorsData,
      partition_name: "test",
    };
  
    await milvusClient.insert(params);
    console.log("--- Insert Data to Collection ---");

• Loads the collection and builds index on it:

    await milvusClient.createIndex({
      collection_name: collectionName,
      field_name: "float_vector",
      extra_params: {
        index_type: "IVF_FLAT",
        metric_type: "L2",
        params: JSON.stringify({ nlist: 10 }),
      },
    });
    console.log("--- Create Index in Collection ---");

• Searches the collection:

        // need load collection before search
    const loadCollectionRes = await milvusClient.loadCollectionSync({
      collection_name: collectionName,
    });
    console.log("--- Load collection (" + collectionName + ") ---", loadCollectionRes);


    const result = await milvusClient.search({
      collection_name: collectionName,
      vectors: [vectorsData[0]["float_vector"]],
      search_params: {
        anns_field: "float_vector",
        topk: "4",
        metric_type: "L2",
        params: JSON.stringify({ nprobe: 1024 }),
        round_decimal: 4,
      },
      output_fields: ["count"],
      vector_type: DataType.FloatVector,
    });

    console.log("--- Search collection (" + collectionName + ") ---", result);

• Releases the collection:

    const releaseRes = await milvusClient.releaseCollection({
      collection_name: collectionName,
    });
    console.log("--- Release Collection ---", releaseRes);

• Drops the collection:

    const dropRes = await milvusClient.dropCollection({
      collection_name: collectionName,
    });
    console.log("--- Drop Collection ---", dropRes);

5. Compile the file

npx tsc MilvusHello.ts

6. Run the example

node MilvusHello.js

Congratulations! You have successfully booted Milvus Standalone and created your first collection.