LNMP SDK for Rust

The official Rust SDK for the LLM Native Minimal Protocol (LNMP).

This SDK provides a unified entry point to the entire LNMP ecosystem, leveraging the lnmp meta crate to give you access to all protocol features, from basic encoding to advanced spatial streaming and neural embeddings.

Installation

Add the following to your Cargo.toml:

[dependencies]
lnmp = "0.5.12"

Note

Most users should depend directly on the lnmp crate from crates.io. This SDK repository serves as a documentation hub and example collection.

Quick Start

use lnmp::prelude::*;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create a simple text token
    let token = Token::text("Hello LNMP");
    
    // Encode it
    let encoded = token.to_lnmp()?;
    println!("Encoded: {:?}", encoded);
    
    Ok(())
}

Modules Overview

The SDK exposes several powerful modules via the lnmp crate:

1. Core (lnmp::core)

The foundation of the protocol. Handles basic types, serialization traits, and the unified Token enum.

Key Features:

• Token-based data representation
• LNMP encoding/decoding
• Type-safe serialization

2. Embedding (lnmp::embedding)

Optimized vector operations for AI applications.

Key Features:

• Multiple quantization types (F32, F16, I8, Binary)
• Delta Mode: Efficiently transmit only changes
• Dimension validation

use lnmp::embedding::Vector;

let vec = Vector::from_f32(vec![0.1, 0.2, 0.3]);

3. Spatial (lnmp::spatial)

Protocol extensions for 3D environments, robotics, and AR/VR.

Key Features:

• Pose tracking (position + orientation)
• SpatialStreamer for efficient updates
• Predictive delta compression

4. Net (lnmp::net)

Networking primitives for LNMP.

Key Features:

• Abstract transport layer
• QUIC support
• Message routing

5. Envelope (lnmp::envelope)

Metadata wrapping for routing, priority, and distributed tracing.

Key Features:

• Priority levels
• Sender/recipient routing
• Custom metadata fields

Examples

The SDK includes comprehensive examples demonstrating various LNMP features. Each example is a standalone program you can run directly.

Running Examples

# Navigate to SDK directory
cd sdk/rust

# Run any example with:
cargo run --example <example_name>

Available Examples

1. Basic Encoding (basic_encoding.rs)

Learn the fundamentals of LNMP encoding and decoding.

cargo run --example basic_encoding

Covers:

• Creating text, number, float, and boolean tokens
• Encoding to LNMP format
• Building compound structures (records)

2. Embedding Operations (embedding_ops.rs)

Work with vector embeddings and delta compression.

cargo run --example embedding_ops

Covers:

• Creating F32 vectors
• Computing delta between vectors
• Applying delta patches
• Different quantization types (F16, I8)

3. Spatial Protocol (spatial_demo.rs)

3D spatial tracking and streaming for robotics/AR/VR.

cargo run --example spatial_demo

Covers:

• Creating spatial poses (position + orientation)
• Tracking multiple objects
• Streaming spatial updates
• LNMP encoding for spatial data

4. Envelope Wrapping (envelope_demo.rs)

Add metadata, routing, and priority to messages.

cargo run --example envelope_demo

Covers:

• Wrapping messages with envelopes
• Setting priority levels
• Adding routing information (sender/recipient)
• Custom metadata fields

5. Complete Workflow (complete_workflow.rs)

Real-world scenario: IoT sensor data pipeline.

cargo run --example complete_workflow

Covers:

• Collecting sensor readings
• Generating embeddings for semantic search
• Enriching data with metadata
• Routing through envelopes
• Complete encode-transmit pipeline

6. All Modules Overview (all_modules.rs)

Comprehensive demonstration of ALL 11 LNMP modules.

cargo run --example all_modules

Covers:

• Complete module inventory
• Core, Codec, Embedding, Envelope
• LLB, Net, Quant, Sanitize
• SFE, Spatial, Transport
• Real examples from each module

Example Output

All examples include detailed console output showing:

• Step-by-step operations
• Data transformations
• Encoded LNMP format
• Success confirmations

Advanced Usage

Custom Protocol Extensions

You can extend LNMP with custom types:

use lnmp::prelude::*;

// Define custom types using Token::record
let custom_message = Token::record(vec![
    ("type".to_string(), Token::text("custom_sensor")),
    ("data".to_string(), Token::int(42)),
]);

Performance Tips

• Use delta mode for frequently updated embeddings
• Enable quantization (F16/I8) for large vectors
• Batch spatial updates when possible
• Set appropriate envelope priorities

Resources

• Main Documentation: See project README.md
• API Reference: Run cargo doc --open in the workspace root
• Specification: Check spec/ directory
• More Examples: Browse examples/ in the main repository

Contributing

Contributions are welcome! Please see the CONTRIBUTING.md file for guidelines.

License

See LICENSE for details.