Summary

Dual-Graph is a context engine that drastically cuts AI coding assistant costs (30-45%) and enhances quality by building a semantic codebase graph. It pre-loads relevant files into prompts for assistants like Claude Code, Codex CLI, and GitHub Copilot, enabling AI to reason more effectively and reducing token expenditure. This solution targets developers and power users seeking optimized AI-assisted coding workflows.

How It Works

The system scans projects to build a semantic graph of files, symbols, and imports. Upon user queries, it identifies and packs relevant files into the AI's context, minimizing exploration tokens. Context memory compounds across sessions, prioritizing recent activity for further cost reduction. All processing is local, ensuring code privacy.

Quick Start & Requirements

Install via shell scripts (macOS/Linux: curl ... | bash; Windows: irm ... | iex) or Scoop (scoop bucket add dual-graph ...). Prerequisites: Python 3.10+, Node.js 18+, and a compatible AI assistant. The installer aids with dependency management. Community Discord: discord.gg/ptyr7KJz.

Highlighted Details

• Achieves 30-45% cost reduction, with benchmarks showing lower average prompt costs ($0.27 vs $0.46), fewer turns (10.3 vs 16.8), faster responses (134s vs 186s), and improved quality (87.1/100 vs 82.7/100).
• Supports TypeScript, JavaScript, Python, Go, Swift, Rust, Java, Kotlin, C#, Ruby, and PHP.
• Guarantees local processing and data privacy; no code leaves the machine.
• Features compounding context memory for progressively cheaper and more efficient interactions.

Maintenance & Community

An active community thrives on Discord (discord.gg/ptyr7KJz), facilitating feedback and bug reporting. Launcher scripts are open source (Apache 2.0), welcoming contributions. The core graph intelligence is proprietary. The tool auto-updates on each run.

Licensing & Compatibility

Launcher scripts and tooling are Apache License 2.0. The core graperoot graph engine (PyPI dependency) is proprietary. This distinction is critical for commercial integration, as the core intelligence is not open source.

Limitations & Caveats

The core graperoot graph engine is proprietary, not open source. The tool requires specific, user-installed AI coding assistants. Setup necessitates Python 3.10+ and Node.js 18+.