RexiO Agent is a self-improving, persistent AI agent framework. It uses a hybrid Python and TypeScript architecture, allowing it to perform backend execution, run cron routines, learn skills dynamically, and stream updates to chat clients or a web-based dashboard.
- Core ReAct Loop: Built-in planning, acting, and reflection cycles.
- SQLite Persistence: Keeps tracking of chat sessions, system states, schedules, and semantic memory.
- Built-in Tools: Integrated file management, web search, and a timed Python sandbox executor.
- Self-Improving Learning Loop: Ability to compile new dynamic tools from successful runs and execute them at runtime.
- Interactive UI: A real CLI client and an upcoming FastAPI-based React + Tailwind CSS dashboard.
- Python 3.11+
uv(recommended for faster package dependency handling)
Clone the repository, configure the environment, and install dependencies.
# Copy and configure environment variables
cp .env.example .env
# Install dependencies
uv pip install -e .
# Or standard pip install:
pip install -r requirements.txtLaunch the interactive session:
python cli.pyPlease refer to the architecture plan in .agents/skills/ (or local config directories) for detailed UML and technical diagrams.
MIT