Getting Started

Relevant source files

The following files were used as context for generating this wiki page:

• README.md
• Cargo.toml
• examples/index_example.rs
• examples/query_search.rs
• examples/authentication.rs
• src/authentication.rs
• src/proto/milvus.proto.milvus.rs

The Milvus Rust SDK is the official library for interacting with the Milvus vector database using the Rust programming language. It facilitates connection management, collection schema definition, data mutation (insert, delete, upsert), and high-performance vector similarity searches or scalar queries. Sources: README.md:1-5, Cargo.toml:1-5

This guide provides the necessary steps to integrate the SDK into a Rust project, establish a connection to a Milvus cluster, and perform fundamental operations such as collection creation and indexing. Sources: README.md:7-10, examples/index_example.rs:17-25

Installation and Setup

To use the Milvus SDK in your Rust application, add it as a dependency in your Cargo.toml file or use the cargo command line tool. Sources: README.md:10-12

cargo add milvus-sdk-rust

The SDK relies on several core dependencies to handle asynchronous networking and data serialization, specifically tonic for gRPC communication, tokio for the runtime, and prost for Protocol Buffers. Sources: Cargo.toml:12-28

Development Prerequisites

For developers contributing to or testing the SDK, the following tools are required:

• Cargo: Rust package manager and build tool.
• Protocol Compiler: Required for compiling .proto files via build.rs.
• Docker: Used to spin up local Milvus instances for integration testing.
  Sources: README.md:31-35, build.rs:5-18

Establishing a Connection

Connections are managed through the Client or ClientBuilder structures. The client communicates with the Milvus proxy using gRPC. Sources: src/authentication.rs:53-55, examples/authentication.rs:7-12

Client Lifecycle Flow

The following diagram illustrates the typical lifecycle of a Milvus client from initialization to collection management.

Authentication Configuration

The SDK supports basic authentication. Credentials can be provided during client initialization. Sources: examples/authentication.rs:7-10, src/authentication.rs:18-35

Option	Type	Description
URL	String	Endpoint of the Milvus server (e.g., http://localhost:19530)
username	String	Username for basic auth (default: "root")
password	String	Password for basic auth (default: "Milvus")
Sources: examples/authentication.rs:7-12, README.md:18

Schema and Collection Management

Before data can be inserted, a CollectionSchema must be defined. This involves using the CollectionSchemaBuilder to specify fields, data types, and primary keys. Sources: examples/index_example.rs:25-35, src/proto/milvus.proto.milvus.rs:76-92

Field Schemas

Milvus supports various data types, including scalars (Int64, VarChar) and vectors (FloatVector). Sources: examples/query_search.rs:104-110, src/proto/milvus.proto.milvus.rs:434-445

let mut schema = CollectionSchemaBuilder::new(COLLECTION_NAME, "Description");
schema
    .add_field(FieldSchema::new_primary_int64("id", "primary key", true))
    .add_field(FieldSchema::new_float_vector("embeddings", "vector field", 128))
    .add_field(FieldSchema::new_varchar("title", "scalar field", 64))
    .enable_dynamic_field();
let schema = schema.build()?;

Sources: examples/index_example.rs:25-35

Data Operations

Once a collection is created, you can insert data and build indexes. Data is organized into FieldColumn objects before being sent via the insert method. Sources: examples/index_example.rs:51-60, examples/query_search.rs:114-125

Indexing and Loading

Indexing is required for efficient vector similarity searches. After creating an index, the collection must be loaded into memory. Sources: examples/index_example.rs:70-85, src/proto/milvus.proto.milvus.rs:175-188

Search and Query

• Search: Performed on vector fields using query vectors and specific search parameters (e.g., nprobe, radius). Sources: examples/query_search.rs:54-68, src/proto/milvus.proto.milvus.rs:475-495
• Query: Used for scalar filtering using boolean expressions (e.g., id > 100). Sources: examples/query_search.rs:31-45, examples/index_example.rs:90-95

Summary

Starting with the Milvus Rust SDK involves adding the dependency to Cargo.toml, initializing a Client with optional authentication, and defining a structured CollectionSchema. Users must follow the sequence of Inserting data, Creating an Index, and Loading the collection before vector searches can be executed. The library utilizes gRPC for all server communications, ensuring type-safe and performant interactions with the Milvus ecosystem. Sources: README.md:7-30, examples/index_example.rs:17-110, Cargo.toml:12-25