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.
OpenAlpha_Evolve is an open-source Python framework inspired by the groundbreaking research on autonomous coding agents like DeepMind's AlphaEvolve.
| Date | Stars |
|---|---|
| 2026-07-31 | 1043 |
| 2026-08-06 | 1046 |
Today
+3 stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# OpenAlpha_Evolve: Contribute to Improve this Project
OpenAlpha_Evolve is an open-source Python framework inspired by the groundbreaking research on autonomous coding agents like DeepMind's AlphaEvolve. It's a **regeneration** of the core idea: an intelligent system that iteratively writes, tests, and improves code using Large Language Models (LLMs) via LiteLLM, guided by the principles of evolution.
Our mission is to provide an accessible, understandable, and extensible platform for researchers, developers, and enthusiasts to explore the fascinating intersection of AI, code generation, and automated problem-solving.
[](LICENSE.md)
## Table of Contents
- [✨ The Vision: AI-Driven Algorithmic Innovation](#-the-vision-ai-driven-algorithmic-innovation)
- [🧠 How It Works: The Evolutionary Cycle](#-how-it-works-the-evolutionary-cycle)
- [🚀 Key Features](#-key-features)
- [📂 Project Structure](#-project-structure)
- [🏁 Getting Started](#-getting-started)
- [💡 Defining Your Own Algorithmic Quests!](#-defining-your-own-algorithmic-quests)
- [🔮 The Horizon: Future Evolution](#-the-horizon-future-evolution)
- [🤝 Join the Evolution: Contributing](#-join-the-evolution-contributing)
- [📜 License](#-license)
- [🙏 Homage](#-homage)
---
## ✨ The Vision: AI-Driven Algorithmic Innovation
Imagine an agent that can:
* Understand a complex problem description.
* Generate initial algorithmic solutions.
* Rigorously test its own code.
* Learn from failures and successes.
* Evolve increasingly sophisticated and efficient algorithms over time.
OpenAlpha_Evolve is a step towards this vision. It's not just about generating code; it's about creating a system that *discovers* and *refines* solutions autonomously.
---
<img width="1253" alt="Screenshot 2025-05-19 at 12 17 58 AM" src="https://github.com/user-attachments/assets/43d7c5a8-f361-438c-ac38-39717f28ee1f" />
## 🧠 How It Works: The Evolutionary Cycle
OpenAlpha_Evolve employs a modular, agent-based architecture to orchestrate an evolutionary process:
1. **Task Definition**: You, the user, define the algorithmic "quest" – the problem to be solved, including examples of inputs and expected outputs.
2. **Prompt Engineering (`PromptDesignerAgent`)**: This agent crafts intelligent prompts for the LLM. It designs:
* *Initial Prompts*: To generate the first set of candidate solutions.
* *Mutation Prompts*: To introduce variations and improvements to existing solutions, often requesting changes in a "diff" format.
* *Bug-Fix Prompts*: To guide the LLM in correcting errors from previous attempts, also typically expecting a "diff".
3. **Code Generation (`CodeGeneratorAgent`)**: Powered by an LLM (currently configured for Gemini), this agent takes the prompts and generates Python code. If a "diff" is requested and received, it attempts to apply the changes to the parent code.
4. **Evaluation (`EvaluatorAgent`)**: The generated code is put to the test!
* *Syntax Check*: Is the code valid Python?
* *Execution*: The code is run in a temporary, isolated environment against the input/output examples defined in the task.
* *Fitness Scoring*: Programs are scored based on correctness (how many test cases pass), efficiency (runtime), and other potential metrics.
5. **Database (`DatabaseAgent`)**: All programs (code, fitness scores, generation, lineage) are stored, creating a record of the evolutionary history (currently in-memory).
6. **Selection (`SelectionControllerAgent`)**: The "survival of the fittest" principle in action. This agent selects:
* *Parents*: Promising programs from theExcerpt of 15,262 characters
Read on GitHub
41
14
6
4
4
2
1
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:ad242725423c5c9c, topic:llm-inference