Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Catches and repairs AI hallucinations before delivery: three quality gates, 31 built-in rules, a self-healing loop and a 90-point acceptance review. Fully automated from requirement to merged code. A primary Agent dynamically spawns and orchestrates child Agents on the AI CLIs on your machine (Claude Code, Codex — 8 supported).
| Date | Stars |
|---|---|
| 2026-07-31 | 304 |
| 2026-08-02 | 307 |
| 2026-08-06 | 307 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
15.0
growth rate 0.00%/day
<p align="center"> <img src="installer/assets/solodawn.png" alt="SoloDawn" width="120" /> </p> <h1 align="center">SoloDawn</h1> <p align="center"> <strong>Give it one sentence. It builds the entire project. You just wait for delivery.</strong> </p> <p align="center"> <a href="README.zh-CN.md">简体中文</a> · <a href="https://linux.do/">Community</a> </p> --- ## What Is SoloDawn? SoloDawn is an open-source web app that runs on your own machine (Rust backend + React frontend): an upper-layer orchestrator Agent (the primary Agent) commands the **AI CLIs actually installed on your computer** (Claude Code, Codex — 8 in total) to carry out fully automated development inside a Git repository — requirement clarification → technical spec generation → task decomposition → parallel development on isolated branches → three-layer quality gates → scored acceptance review → automatic merge. SoloDawn's ultimate design goal is to **complete complex, production-grade products through a simple conversation on a social platform** — not toy demos, but real, complex, production-ready software. > **One-liner:** Whether you're a programmer or not, just describe what you need. Everything else is automatic. > **About AI hallucination, honestly:** SoloDawn cannot stop a model from hallucinating at the moment it emits tokens — nobody can. Through architecture and workflow design it catches and repairs hallucination artifacts before delivery: 31 built-in quality rules + three quality gates + a self-healing loop + a 90-point acceptance bar, so hallucinations, security holes, and integration conflicts each get intercepted on a different dimension. See [Quality System in Depth](#quality-system-in-depth). --- ## Community QQ Group <p align="center"> <img src="docs/assets/qq-group-qr.jpg" alt="SoloDawn QQ group QR code" width="240" /> </p> Scan the QR code (or search group ID **1053609555**) to join the SoloDawn QQ group — questions, feedback, and usage ideas are all welcome. --- ## Two Core Highlights ### Highlight 1: Fully Automated AI Development The Orchestrated Workspace uses a **two-layer Agent architecture**: a primary Agent plus child Agents — not three layers, and no child Agent is preconfigured. Every child Agent is dynamically spawned and terminated by the primary Agent: whenever the work needs more hands (additional tasks, code review, defect fixing, integration repair), the primary Agent creates new child Agents — exactly as many as required, closed when done. This design is fully implemented. Why abandon fixed workflow definitions? Because rigid workflows introduce hard constraints: they're overly cumbersome for simple tasks yet lack capacity for complex ones, and the conditional-judgment logic inside a workflow inevitably misjudges. So the two-layer architecture grants full authority to the primary Agent — it alone decides what to spawn, how many, and when to close them. AI-assisted coding today has a fundamental contradiction: programmers still have to build workflows, configure skills, wire up MCP servers, and write plans themselves — the human remains the driver; non-programmers can't even turn a requirement into a technical spec. SoloDawn takes the driving over from the human: | Who you are | What you do | What the system does | |---|---|---| | **Programmer** | Throw in a precise task goal | Your input is used directly as the technical spec and executed — no follow-up questions. No commands to issue, no "Continue" to click; just wait for acceptance | | **Non-programmer** | Describe what you want in plain language | Plain-language follow-up questions close the gaps → a technical spec + acceptance scoring rubric are generated in the background → fully automated development begins | Delivery isn't "it ran, so it's done": every commit passes quality gates, and every completed task is scored item-by-item against a **rubric generated from your requirements** — **below 90 points it is automatically sent back
Excerpt of 57,310 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:7c77c25951eb7f2f, llm:Repository description: 'Catches and repairs AI hallucinations... three quality gates, 31 built-in rules, a self-healing loop and a 90-point acceptance review. Fully automated from requirement to merged code. A primary Agent dynamically spawns and orchestrates child Agents on the AI CLIs on your machine (Claude Code, Codex — 8 supported).' Topics: gitcortex. Language: Rust.
matched fp:7c77c25951eb7f2f, llm:Repository description: 'Catches and repairs AI hallucinations... three quality gates, 31 built-in rules, a self-healing loop and a 90-point acceptance review. Fully automated from requirement to merged code. A primary Agent dynamically spawns and orchestrates child Agents on the AI CLIs on your machine (Claude Code, Codex — 8 supported).' Topics: gitcortex. Language: Rust.
matched fp:7c77c25951eb7f2f, llm:Repository description: 'Catches and repairs AI hallucinations... three quality gates, 31 built-in rules, a self-healing loop and a 90-point acceptance review. Fully automated from requirement to merged code. A primary Agent dynamically spawns and orchestrates child Agents on the AI CLIs on your machine (Claude Code, Codex — 8 supported).' Topics: gitcortex. Language: Rust.
matched fp:7c77c25951eb7f2f, llm:Repository description: 'Catches and repairs AI hallucinations... three quality gates, 31 built-in rules, a self-healing loop and a 90-point acceptance review. Fully automated from requirement to merged code. A primary Agent dynamically spawns and orchestrates child Agents on the AI CLIs on your machine (Claude Code, Codex — 8 supported).' Topics: gitcortex. Language: Rust.