SoloDawn  by huanchong-99

Automated AI software development with hallucination repair

Created 5 months ago
297 stars

Top 89.2% on SourcePulse

GitHubView on GitHub
Project Summary

<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> SoloDawn automates complex, production-grade software development from requirements to merged code, acting as a full development team. It targets AI hallucination detection and repair via a multi-stage quality system, enabling non-programmers and solo developers to build sophisticated applications.

How It Works

It uses a two-layer Agent architecture where a primary orchestrator dynamically spawns and manages child AI CLIs (e.g., Claude Code, Codex) for development tasks. Workflows are Git-driven, executing tasks in parallel on isolated branches. This dynamic, non-preconfigured approach grants the primary Agent full decision authority, offering flexibility and avoiding rigid workflow constraints.

Quick Start & Requirements

Clone the repository (https://github.com/huanchong-99/SoloDawn.git) and run pnpm run dev. Prerequisites include Rust nightly-2025-12-04, Node.js >= 18, pnpm 10.13.1, Git, a C/C++ toolchain, protoc 31.1, LLVM/libclang, cmake, nasm, perl, and sqlx-cli 0.8.6. A 32-character SOLODAWN_ENCRYPTION_KEY environment variable is required. Initial backend compilation takes several minutes.

Highlighted Details

  • Automated Development: Orchestrated Workspace mode handles the entire development lifecycle from requirements to merge.
  • Quality System: Three-layer gates (Terminal, Branch, Repo) with 31 built-in rules, self-healing loops for automatic error correction, and a 90-point acceptance review bar.
  • Native Skill Inheritance: Integrates and utilizes existing AI CLIs, plugins, and MCP servers seamlessly without re-configuration.
  • Multi-AI-CLI Collaboration: Supports concurrent execution of different AI CLIs within a single workflow, assigning distinct roles.
  • Continuation Rounds: Enables iterative development by allowing subsequent conversation rounds to build upon previous deliveries using context capsules.

Maintenance & Community

Initiated in January 2026, with version 1.0 released in July 2026. No specific community links or contributor details are provided in the README.

Licensing & Compatibility

SoloDawn is licensed under Apache-2.0. CC-Switch components are MIT licensed, and Quality Gate models are LGPL-3.0. Apache-2.0 is permissive for commercial use and closed-source integration.

Limitations & Caveats

Continuation rounds, while implemented, are new and not yet fully battle-tested at scale. Multi-AI-CLI collaboration and the Feishu connector require re-testing and may not be fully stable. Users might need to leverage AI CLIs to resolve environment-specific setup issues.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
15
Issues (30d)
0
Star History
124 stars in the last 30 days

Explore Similar Projects

Starred by John Mullan John Mullan(MTS at xAI; Cofounder of Hotshot AI), Thomas Wolf Thomas Wolf(Cofounder of Hugging Face), and
24 more.

n8n by n8n-io

0.4%
196k
AI agent and workflow automation platform
Created 7 years ago
Updated 22 hours ago
Feedback? Help us improve.