CoolFace
Apppublic

MadeiraG/aeon-next

sourceHugging Faceupdated 1mo agoView on Hugging Face
0likes
App README

AEON NEXT — Home AI Agent System

A local-first multi-agent AI project for a Mac mini / home server. It is designed to run with free local models first, then optionally use Hugging Face or paid providers when available.

What it is

  • —FastAPI backend
  • —React dashboard
  • —SQLite persistence
  • —4 agent roles: Trader, Analyst, Creator, Auditor
  • —local model support via Ollama
  • —optional cloud model support via Hugging Face inference
  • —Docker-based deployment for Hugging Face Spaces

Local-first setup

  1. 1.Copy .env.example to .env
  2. 2.Prefer free local AI:
  3. 3.MODEL_PROVIDER=auto
  4. 4.OLLAMA_BASE_URL=http://localhost:11434
  5. 5.OLLAMA_MODEL=llama3.1
  6. 6.Start the app with Docker or local uvicorn

Free AI options

Ollama (recommended)

bash
MODEL_PROVIDER=ollama
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3.1

Hugging Face

bash
MODEL_PROVIDER=huggingface
HF_TOKEN=your_token_here
HF_MODEL=google/gemma-2-2b-it
HF_BASE_URL=https://api-inference.huggingface.co

Optional Anthropic fallback

bash
ANTHROPIC_API_KEY=your_key_here

Run locally

bash
cp .env.example .env
pip install -r requirements.txt
python -m uvicorn app.backend.main:app --reload --host 0.0.0.0 --port 8000

Frontend in another terminal:

bash
cd app/frontend
npm install
npm run dev

Open:

  • —http://localhost:5173
  • —http://localhost:8000/docs

Docker setup

bash
docker compose up -d

This includes an Ollama service for a home/server environment.

Hugging Face deployment

This repo is already connected to a Hugging Face Space remote and is configured for a Docker-based Space.

Use space secrets/variables for:

  • —MODEL_PROVIDER
  • —OLLAMA_BASE_URL
  • —OLLAMA_MODEL
  • —HF_TOKEN
  • —HF_MODEL
  • —ANTHROPIC_API_KEY
  • —NEWSAPI_KEY

Architecture

  • —backend: app/backend/
  • —frontend: app/frontend/
  • —database: SQLite in /data/aeon.db
  • —providers: provider abstraction in app/backend/llm_provider.py

Notes

This project is designed to work at home on a Mac mini with free local AI where possible, and to degrade gracefully if no paid endpoint is configured.