BNN Code

Terminal-native AI coding agent powered by Binarized Neural Networks

BNN Code is a terminal-based AI coding assistant that helps you understand, refactor, and test your codebase using Binarized Neural Networks (BNNs). It runs entirely in your terminal with zero IDE lock-in.

Features

• Code Understanding — Index your codebase and ask questions in natural language
• Multi-language Support — Python, JavaScript, TypeScript, Rust, Go, Java, C++, Ruby, Swift, Kotlin
• AST-aware Indexing — Smart chunking using tree-sitter grammars
• Terminal-native UI — Built with Ratatui + Crossterm, no IDE required
• BNN Inference — On-device AI with ONNX Runtime (SSE4.2 compatible)
• SQLite-backed — Fast local search and retrieval
• Editor Integrations — VS Code extension + Neovim plugin included

Installation

From GitHub Releases (recommended)

Download the latest binary for your platform from the Releases page:

# Linux (glibc)
tar -xzf bnn-code-linux-amd64.tar.gz
sudo mv bnn-code /usr/local/bin/

# macOS (Intel)
tar -xzf bnn-code-macos-amd64.tar.gz
sudo mv bnn-code /usr/local/bin/

# macOS (Apple Silicon)
tar -xzf bnn-code-macos-arm64.tar.gz
sudo mv bnn-code /usr/local/bin/

# Windows
unzip bnn-code-windows-amd64.exe.zip
# Move bnn-code.exe to a directory in your PATH

From crates.io

cargo install bnn-code

Deepin / Debian / Ubuntu

Download the .deb package from the Releases page:

sudo dpkg -i bnn-code_0.1.1_amd64.deb

Build from Source

git clone https://github.com/Yotsawarit/bnn-code.git
cd bnn-code
cargo build --release
./target/release/bnn-code --help

Quick Start

# Index your codebase
bnn-code index --path /path/to/your/project

# Search for code patterns
bnn-code search "function that handles authentication"

# Start interactive TUI
bnn-code

# Get help
bnn-code --help

Editor Integrations

BNN Code comes with first-class editor support.

VS Code

1. Install the extension from VS Code Marketplace (coming soon) or install manually:

bash scripts/install-vscode.sh

2. Open command palette (Ctrl+Shift+P) and run:
   • BNN: Index Workspace
   • BNN: Search Code
   • BNN: Query with AI

Neovim

bash scripts/install-neovim.sh

Then use the following keymaps:

• <Leader>bn — Open BNN Code
• <Leader>bi — Index workspace
• <Leader>bs — Search code
• See :help bnn-code for full documentation

Download Models

# Download CodeBERTa-small model (recommended for most users)
bash scripts/download_model.sh

# Or use the Python script for more options
python3 scripts/download_model.py --model codeberta-small

Commands

Command	Description
bnn-code index	Index a codebase
bnn-code search	Search indexed code
bnn-code query	Query with AI inference
bnn-code	Launch interactive TUI
bnn-code --help	Show help

Architecture

┌─────────────────────────────────────┐
│           CLI (clap)                │
├─────────────────────────────────────┤
│  ┌─────────┐  ┌──────────┐         │
│  │ Indexer │  │Retrieval │         │
│  │ (AST)   │  │ (SQLite) │         │
│  └────┬────┘  └────┬─────┘         │
│       │            │               │
│  ┌────▼────────────▼─────┐         │
│  │   Inference Engine    │         │
│  │  (ONNX Runtime + BNN) │         │
│  └───────────────────────┘         │
│  ┌───────────────────────┐         │
│  │   Terminal UI (TUI)   │         │
│  │  (Ratatui + Crossterm)│         │
│  └───────────────────────┘         │
├─────────────────────────────────────┤
│  VS Code Extension │ Neovim Plugin │
└─────────────────────────────────────┘

Supported Platforms

Platform	Binary
Linux x86_64 (glibc)	bnn-code-linux-amd64.tar.gz
Linux x86_64 (static)	bnn-code-linux-amd64-static.tar.gz
macOS x86_64	bnn-code-macos-amd64.tar.gz
macOS ARM64	bnn-code-macos-arm64.tar.gz
Windows x86_64	bnn-code-windows-amd64.exe.zip
Deepin/Debian/Ubuntu	bnn-code_0.1.1_amd64.deb

Development

# Check
cargo check

# Run tests
cargo test

# Build release
cargo build --release

# Build static binary
RUSTFLAGS="-C target-feature=+crt-static" cargo build --release

Project Structure

src/
├── cli/         # CLI argument parsing (clap)
├── indexer/     # Code indexing with tree-sitter AST
│   ├── chunker.rs
│   ├── database.rs
│   └── parser.rs
├── inference/   # ONNX Runtime + BNN inference
│   └── model.rs
├── retrieval/   # SQLite search and retrieval
│   └── search.rs
├── ui/          # Terminal UI (Ratatui)
│   ├── mod.rs
│   ├── streaming.rs
│   └── terminal.rs
├── utils/       # Shared utilities
│   ├── cache.rs
│   └── config.rs
└── main.rs

License

MIT License — see LICENSE.

Copyright (c) 2026 Mr. Yotsawarit Pudpong