rajz117/ios-ml-artifacts

iOS ML Artifacts

Pre-built iOS ML artifacts for on-device inference. Includes ONNX Runtime xcframework, sqlite-vec static libraries, and embedding models optimized for iOS.

Contents

1. ONNX Runtime iOS XCFramework

• Version: 1.22.0
• Size: ~49MB
• Architectures:
  • ios-arm64 - Physical iOS devices
  • ios-arm64-simulator - iOS Simulator (Apple Silicon)
• Features: CoreML execution provider for Apple Neural Engine acceleration

2. sqlite-vec Static Libraries

• Version: 0.1.6
• Size: ~340KB per architecture
• Architectures:
  • ios-arm64 - Physical iOS devices
  • ios-arm64-simulator - iOS Simulator (Apple Silicon)
• Purpose: Vector similarity search extension for SQLite

3. BGE-Small Text Embedding Model

• Model: bge-small-en-v1.5-int8.onnx
• Size: 32MB (INT8 quantized)
• Dimensions: 384
• Use Case: Text embeddings for semantic search
• Includes: tokenizer.json for text preprocessing

Quick Start

Download from GitHub Releases

# Download all artifacts
gh release download v1.0.0 --repo vannmex/vannmex-ios-ml-artifacts

# Or download specific files
gh release download v1.0.0 --pattern "onnxruntime-*.zip" --repo vannmex/vannmex-ios-ml-artifacts
gh release download v1.0.0 --pattern "sqlite-vec-*.zip" --repo vannmex/vannmex-ios-ml-artifacts
gh release download v1.0.0 --pattern "bge-small-*.zip" --repo vannmex/vannmex-ios-ml-artifacts

Download from Hugging Face

# Install HF CLI
brew install huggingface-cli

# Download all artifacts
hf download vannmex/ios-ml-artifacts --local-dir ./artifacts

# Or download specific files
hf download vannmex/ios-ml-artifacts onnxruntime/onnxruntime.xcframework --local-dir ./onnxruntime

Integration

ONNX Runtime in Xcode

1. Unzip onnxruntime-ios-1.22.0.xcframework.zip
2. Drag onnxruntime.xcframework into your Xcode project
3. In Build Phases > Link Binary With Libraries, ensure it's listed
4. Add to Framework Search Paths if needed

sqlite-vec in Rust/Tauri

// Load the extension
unsafe {
    conn.load_extension_enable()?;
    conn.load_extension("libsqlite_vec0", None)?;
    conn.load_extension_disable()?;
}

BGE-Small Embedding Model

use ort::{Session, SessionBuilder};

let session = SessionBuilder::new()?
    .with_execution_providers([CoreMLExecutionProvider::default().build()])?
    .with_model_from_file("bge-small-en-v1.5-int8.onnx")?;

Checksums

Verify file integrity with SHA-256:

shasum -a 256 *.zip

License

• ONNX Runtime: MIT License
• sqlite-vec: MIT License
• BGE-Small: MIT License (BAAI)

Related Projects

• ONNX Runtime
• sqlite-vec
• BGE Models