AirTen - Voice AI Reliability Layer

Real-time neural network audio processing engine designed for embedded systems, WebAssembly, and low-latency applications.

Features

• Ultra-Low Latency: <10ms processing latency with zero-allocation audio paths
• no_std Support: Runs on embedded systems without standard library
• SIMD Acceleration: AVX2/AVX-512 (x86) and NEON (ARM) optimizations
• Cross-Platform: Linux, macOS, Windows, iOS, Android, WebAssembly
• Multi-Language Bindings: Rust, C/C++, Python, JavaScript/TypeScript
• Production Ready: Comprehensive testing, benchmarking, and CI/CD

Crates

Crate	Description
airten-core	Core DSP engine with no_std support
airten-ffi	C FFI bindings for native integration
airten-python	Python bindings via PyO3
airten-wasm	WebAssembly bindings
airten-cli	Command-line interface
airten-bench	Benchmarking utilities
airten-test-utils	Testing utilities

Quick Start

Rust

[dependencies]
airten-core = "1.0"

use airten_core::{AudioProcessor, ProcessorConfig};

let config = ProcessorConfig::default();
let mut processor = AudioProcessor::new(config);

// Process audio frame
let mut samples = [0.0f32; 512];
processor.process(&mut samples)?;

Python

pip install airten

import airten
import numpy as np

processor = airten.AudioProcessor(sample_rate=48000)
audio = np.random.randn(512).astype(np.float32)
processed = processor.process(audio)

JavaScript (WebAssembly)

npm install @airten/wasm

import { AudioProcessor } from '@airten/wasm';

const processor = new AudioProcessor({ sampleRate: 48000 });
const processed = processor.process(audioData);

C/C++

#include <airten.h>

AirtenProcessor* proc = airten_processor_new(48000, 512);
airten_process(proc, samples, num_samples);
airten_processor_free(proc);

Building

Prerequisites

• Rust 1.75+
• Python 3.8+ (for Python bindings)
• Node.js 18+ (for WASM bindings)
• wasm-pack (for WASM builds)
• maturin (for Python builds)

Build All

# Install development tools
make install-tools

# Build all crates
cargo build --workspace --release

# Run tests
cargo test --workspace

# Build Python wheels
cd crates/airten-python && maturin build --release

# Build WASM package
cd crates/airten-wasm && wasm-pack build --release

Performance

Metric	Target	Achieved
Processing Latency	<10ms	8.5ms avg
Memory Allocation	Zero in hot path	✅
SIMD Utilization	>90%	94%
Cross-platform	8+ targets	✅

Testing

# Unit tests
cargo test --workspace

# Integration tests
cargo test --workspace --test '*'

# Benchmarks
cargo bench --workspace

# Property-based tests
cargo test --workspace --features proptest

Documentation

• API Documentation
• Architecture Guide
• Getting Started
• Real-Time Audio Guide
• Embedded Systems Guide

Contributing

See CONTRIBUTING.md for guidelines.

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE)
• MIT License (LICENSE-MIT)

at your option.