neuroflow-ml 0.2.0

A local-first, node-based ML orchestrator with a visual pipeline builder

NeuroFlow

A local-first, node-based ML orchestrator with a visual pipeline builder.

Features

• Visual Pipeline Builder - Drag-and-drop nodes to create ML workflows
• Node-Based Architecture - Connect data loaders, preprocessors, models, and evaluators
• Custom Python Nodes - Inject custom Python code directly into your workflows
• Code View Toggle - Switch between visual editor and code representation (Overleaf-style)
• Built-in Visualizations - View metrics, charts, and model performance
• Local-First - All data stays on your machine
• Export/Import - Save and share workflows as JSON
• One Command - Install and run with a single command

Quick Start

Installation

# Install from PyPI
pip install neuroflow-ml

# Or install from source
git clone https://github.com/aayushgokhale/neuroflow.git
cd neuroflow/NeuroFlow
pip install -e .

Running NeuroFlow

# Start NeuroFlow (opens browser automatically)
neuroflow

# Or specify a custom port
neuroflow --port 3000

# Don't open browser automatically
neuroflow --no-browser

Visit http://localhost:8000 to access the visual pipeline builder.

Usage

Building a Pipeline

1. Drag nodes from the left sidebar onto the canvas
2. Connect nodes by dragging from output ports to input ports
3. Configure nodes by clicking on them and editing parameters
4. Run the workflow by clicking the "Run Workflow" button
5. View results in the Results Panel with metrics and charts

Custom Python Nodes

NeuroFlow supports custom Python nodes for maximum flexibility:

1. Click "Add Custom Node" in the left sidebar
2. Write your Python code with access to input_data, pd, np, and sklearn
3. Set the output variable with your results
4. Connect to other nodes and run

View Toggle (Overleaf-Style)

Switch between visual and code views using the toggle in the top navigation:

• Visual: Drag-and-drop node canvas
• Code: VS Code-style syntax-highlighted view of all nodes

Node Types

Category	Nodes
Data	CSV Loader, SQL Query, API Fetch
Preprocessing	Normalize, Encode, Split Data, Handle Missing
Models	Random Forest, Neural Net, XGBoost, Gradient Boosting
Evaluate	Accuracy Chart, Loss Curve, Confusion Matrix
Custom	Custom Python Node

Example Workflow

CSV Loader -> Split Data -> Random Forest -> Evaluator

Development

Prerequisites

• Python 3.9+
• Node.js 18+
• pnpm (recommended)

Setup

# Clone the repository
git clone https://github.com/aayushgokhale/neuroflow.git
cd neuroflow

# Install backend dependencies
cd NeuroFlow
pip install -e ".[dev]"

# Install frontend dependencies
cd ../Frontend
pnpm install

# Run development servers
# Terminal 1: Backend
cd NeuroFlow && uvicorn app.main:app --reload

# Terminal 2: Frontend
cd Frontend && pnpm dev

Building for Distribution

cd NeuroFlow
./build.sh
pip install -e .

Project Structure

neuroflow/
|-- Frontend/                 # Next.js frontend
|   |-- app/                  # App router pages
|   |-- components/           # React components
|   |   |-- neuroflow/        # NeuroFlow-specific components
|   |-- public/               # Static assets
|-- NeuroFlow/                # Python backend package
|   |-- app/                  # FastAPI application
|   |   |-- cli.py            # CLI entry point
|   |   |-- main.py           # FastAPI app
|   |   |-- engine.py         # Workflow execution engine
|   |   |-- schemas.py        # Pydantic models
|   |   |-- executors/        # Node executors
|   |-- data/                 # Demo datasets
|   |-- static/               # Built frontend (after build)
|   |-- pyproject.toml        # Package configuration

API Reference

REST Endpoints

Method	Endpoint	Description
GET	/health	Health check
POST	/api/workflows/execute/sync	Execute a workflow synchronously
POST	/api/workflows/execute	Execute a workflow asynchronously
GET	/api/workflows/{id}/status	Get execution status
POST	/api/data/upload	Upload data files
GET	/api/models/{id}/download	Download trained model

WebSocket

Connect to /ws/workflow/{run_id} for real-time execution progress updates.

Changelog

v0.2.0 (2026-03-04)

New Features

• Added Overleaf-style view toggle between visual editor and code view
• Added custom Python node support with inline code execution
• Added custom node drawer for creating nodes with configurable inputs/outputs
• Added "Load Custom Nodes Demo" workflow showcasing custom node capabilities
• Added edit/save functionality in code view

Improvements

• Improved JSON serialization for pandas/numpy objects in workflow results
• Enhanced storage manager with proper type conversion
• Updated node icons - custom nodes now use Code2 icon
• Better error handling in workflow execution

Bug Fixes

• Fixed async executor issues with custom Python nodes
• Fixed pandas Series serialization errors in API responses
• Fixed connection line rendering on canvas zoom

v0.1.0 (2026-02-15)

Initial Release

• Visual drag-and-drop pipeline builder
• Node-based ML workflow architecture
• Built-in node types: Data, Preprocessing, Models, Evaluation
• Local-first execution with FastAPI backend
• Demo workflow with auto-mpg dataset
• Export/import workflows as JSON
• Real-time execution progress via WebSocket

License

MIT License - see LICENSE for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Contact

Created by Aayush Gokhale