🧠 NeuroFlow: The AI-Architected Training Platform

NeuroFlow is a state-of-the-art node-based platform designed to simplify the creation, management, and deployment of AI training pipelines. By combining the power of AutoTrain Advanced, a robust React-Flow frontend, and Gemini-powered AI Architecture, NeuroFlow enables users to build production-ready AI workflows in minutes.

---

🚀 Vision

The goal of NeuroFlow is to democratize fine-tuning and AI orchestration. Whether you are a researcher or an engineer, you can design complex pipelines—from data ingestion to model deployment—using a visual canvas aided by an AI co-pilot.

---

✨ Key Features

🤖 AI Build (NeuroBot Architect)

• Intelligent Design: Describe your goal in natural language (e.g., "Build a LLM fine-tuning pipeline for medical data"), and NeuroBot will automatically architect the entire node graph for you.
• Dual-Mode Assistant: Supports both general architectural advice and direct graph generation.
• Powered by Gemini: Leverages high-performance LLMs to ensure logically sound connections.

🎨 Visual Workflow Editor

• Reactive Canvas: Drag-and-drop nodes to build sequences of datasets, models, trainers, and logic routers.
• Node Library:
  • Data: Hub and local dataset support.
  • Logic: Routers, Python REPL, and conditional switches.
  • Interface: One-click deployment to Streamlit and Gradio.
  • Deployment: Direct push-to-hub integration.

⚙️ Production-Ready Engine

• AutoTrain Integration: Deeply integrated with autotrain-advanced for high-performance training on local or cloud backends.
• Dynamic UI Generation: Automatically generates and launches interactive web interfaces (Streamlit/Gradio) tailored to your pipeline's inputs and outputs.
• Local Persistence: Graphs are automatically saved to your browser's local storage.
• Pipeline Export: Download your architectures as JSON for sharing or version control.

---

🏗️ Technical Architecture

Frontend: React + Vite

• Core: TypeScript-based SPA.
• Graph Engine: React Flow for high-performance node orchestration.
• Styling: Premium UI design with dark mode, glassmorphism, and smooth animations.

Backend: Python (AutoTrain)

• API Layer: FastAPI/Uvicorn for low-latency communication.
• Execution Engine: GraphExecutor handles topological sorting and sequential/parallel execution of nodes.
• Subprocess Orchestration: Manages orphan UI processes (Streamlit/Gradio) with automated port management and logging.

AI Core: Gemini API

• Integration: Official @google/genai SDK for structured JSON generation of graph states.

---

🛠️ Setup and Installation

Prerequisites

• Node.js (v18 or higher)
• Python (3.10 or higher)
• Gemini API Key (from Google AI Studio)

1. Clone the Repository

git clone https://github.com/Esmail-ibraheem/NeuroFlow.git
cd NeuroFlow

2. Backend Setup

# Install dependencies
pip install -e .
pip install streamlit gradio uvicorn fastapi

3. Frontend Setup

cd neuroflow
npm install

4. Configuration

Create a .env file in the neuroflow directory:

VITE_API_KEY=your_gemini_api_key_here

---

🚀 Running the Application

Start the Backend (API)

From the root directory:

cd src
uvicorn autotrain.app.app:app --reload

Start the Frontend

From the neuroflow directory:

npm run dev

The app will be available at http://localhost:5173.

---

🧪 Testing and Verification

Manual Verification

1. AI Build: Click the "AI Build" button and type "Fine-tune Llama 3 on a medical dataset".
2. Save/Load: Create a graph, refresh the page, and verify the graph persists.
3. UI Launch: Add a "Streamlit UI" node, connect it to a model output, and click "Train". A new browser tab will open with the generated interface.

Automated Diagnostics

A diagnostic script is provided to verify your Gemini model access:

cd neuroflow
node list_models.js

---

📁 Repository Structure

.
├── src/                # Backend Source (Python)
│   └── autotrain/      # AutoTrain Core & Graph Engine
├── neuroflow/          # Frontend Source (React/Vite)
│   ├── src/            # UI Components & Services
│   └── .env            # API Keys
├── configs/            # Training Configurations
└── README.md           # This file!

---

📜 License

Developed as part of the NeuroTron project. MIT.