https://github.com/gin66/metaagentsystem
Experiment to generate an AI bootstrapped MetaAgentSystem (initial with grok4/gemini)
https://github.com/gin66/metaagentsystem
Last synced: 11 months ago
JSON representation
Experiment to generate an AI bootstrapped MetaAgentSystem (initial with grok4/gemini)
- Host: GitHub
- URL: https://github.com/gin66/metaagentsystem
- Owner: gin66
- License: mit
- Created: 2025-07-20T16:44:57.000Z (11 months ago)
- Default Branch: main
- Last Pushed: 2025-07-20T17:54:41.000Z (11 months ago)
- Last Synced: 2025-07-20T18:25:48.813Z (11 months ago)
- Language: Swift
- Size: 99.6 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Meta Agentic AI System
## Overview
This project involves developing a Meta Agentic AI system using Swift and Ollama. The system is designed to be self-improving, optimizing agent and LLM performance through adaptive process design and rigorous performance evaluation.
The project is structured to ensure that all development and operations are contained within the `projects/MetaAgentSystem` directory. This includes all source code, tests, documentation, and operational scripts.
## Vision
The vision is to create a dynamic, self-improving meta agentic AI system that optimizes agent and LLM performance, ensuring continuous progress and breakthrough results through adaptive process design and rigorous performance evaluation.
## Stakeholder Requirements
The system will manage agent-LLM pairings, optimize processes, and ensure continuous progress via performance evaluation, with agents running in isolated Swift containers. All operations and file modifications are restricted to the `projects/MetaAgentSystem` directory.
## Agile Plan
The project will follow an agile implementation plan with fast sprints. Each sprint will deliver a testable deliverable. All work, including code, tests, and documentation, is managed within the `projects/MetaAgentSystem` directory.
## How it Works
The system uses a `bootstrap.swift` script to automate the development process. This script operates exclusively within the `projects/MetaAgentSystem` directory and performs the following steps:
1. **Verifies a clean git state.** Before starting, it ensures that there are no uncommitted changes.
2. **Reads the current sprint plan.** It reads the `NextSteps.json` file to determine the tasks for the current sprint.
3. **Generates code.** An AI agent generates Swift code and unit tests based on the sprint plan.
4. **Builds and tests.** The script compiles the code and runs the unit tests.
5. **Commits the changes.** If the build and tests are successful, the script commits the changes with a descriptive message.
6. **Plans the next sprint.** An AI agent updates `NextSteps.json` with the plan for the next sprint.
This entire process is confined to the `projects/MetaAgentSystem` directory, ensuring that no other part of the file system is affected.