Milvus Node.js SDK

The Milvus Node.js SDK API reference is organized into several functional areas to help you find specific method signatures and configurations. Use the following categories to navigate the API:

• Client: Access MilvusClient for core SDK entry points and HttpClient for REST-specific configurations and HTTP-only operations.
• Operations: Grouped by resource type, including:
  • Collection: Manage lifecycle (create, load, release, rename, etc.).
  • Data: Perform CRUD and search (insert, upsert, query, search, delete, bulk import).
  • Index: Manage vector and scalar indexes.
  • Partition: Manage collection partitions.
  • Database: Manage database instances.
  • User & Role: Manage access control (users, passwords, roles, privileges).
  • Resource Group: Manage capacity and node transfers.
  • System: Health checks, metrics, and analyzer management.
  • BulkWriter: Client-side JSON/Parquet bulk file preparation.
• Types: Reference common data types, error codes, and response shapes.

Client telemetry tracks the following information:

• Operation Metrics: Request, success, and error counts, plus average, p99, and maximum latency for supported data operations. High-level operations include their retries in a single logical operation. Note that iterator setup and batches are excluded.
• Heartbeat Metadata: SDK version, client identity, hostname, user, and database context.
• Error Details: Operation, collection, request ID, and error text.

Important Notes:

• Sampling: The samplingRate reduces the number of recorded operations rather than extrapolating counts.
• Latency: Metrics arrive in windows and do not provide an instantaneous view of requests.
• Server Control: The server can request metrics, inspect configuration, or push changes to your enabled state, heartbeatIntervalMs, or samplingRate. If the server pushes a 'disabled' state, the heartbeat control channel remains active.

Filter expressions are used in search(), query(), and delete() methods. They support comparison operators, logical operators, IN clauses, string matching with LIKE, array operations, and JSON field access. You can also use parameterized template expressions to prevent injection and improve readability.

// Comparison
age > 18
price <= 99.9
status == "active"
name != "test"

// Logical operators
age > 18 AND status == "active"
price < 10 OR price > 100
NOT (status == "deleted")

// IN operator
id IN [1, 2, 3]
category IN ["books", "music"]

// String matching
title LIKE "hello%"         # Starts with "hello"
title LIKE "%world"         # Ends with "world"

// Array operations
ARRAY_CONTAINS(tags, "ai")
ARRAY_LENGTH(tags) > 3

// JSON field access
metadata["key"] == "value"
metadata["nested"]["field"] > 10

// Template expressions (parameterized)
await client.query({
  collection_name: 'test',
  filter: 'age > {min_age}',
  filter_params: { min_age: 18 },
});

For highly structured or nested data, Milvus supports Array and Struct types.

Array

An Array is a collection of scalar values. Requires element_type and max_capacity.

Struct

A Struct is a nested structure. Requires element_type (which is an object defining the child fields).

StructArray (Milvus 3.0+)

A StructArray is an Array where each element follows a nested Struct schema. This allows storing arrays of complex objects.

Using StructArray in Queries:

• Use bracketed paths for child fields: features[label].
• Use element_filter for predicates matching values within the same array element: element_filter(features, $[score] >= 90).
• To add a StructArray to an existing collection, set nullable: true to maintain compatibility with existing rows.

// StructArray definition
{
  name: 'features',
  data_type: DataType.Array,
  element_type: DataType.Struct,
  max_capacity: 16,
  nullable: true,
  fields: [
    { name: 'label', data_type: DataType.VarChar, max_length: 64 },
    { name: 'score', data_type: DataType.Int32 },
    { name: 'active', data_type: DataType.Bool },
    { name: 'embedding', data_type: DataType.FloatVector, dim: 128 },
  ],
}

// Querying a StructArray
const rows = await client.query({
  collection_name: 'documents',
  filter: 'element_filter(features, $[score] >= 90 AND $[active] == true)',
  output_fields: ['id', 'features[label]', 'features[score]'],
});