Psynapse

A visual node-based workflow editor for creating and executing computational graphs with Python. Psynapse is meant to be the single no-code workflow editor that lets you harness the entire power of the python ecosystem using simple nodepacks and intuitive drag-and-drop interface.

https://github.com/user-attachments/assets/309ea22a-1fdc-4685-baf5-cd728441c329

https://github.com/user-attachments/assets/660d8857-9c8e-4287-a45e-83637e9bdd12

Features

• Visual Node Editor: Drag-and-drop interface built with ReactFlow
• Python Backend: FastAPI server with real-time execution streaming
• Extensible Nodepacks: Add custom Python functions as nodes
• Type-Safe: Full TypeScript frontend and type-hinted Python backend
• Progress Tracking: Real-time progress updates for long-running operations
• Multiple Node Types: Functions, variables, lists, and view nodes

Quick Start

Using Docker Compose (Recommended)

docker compose -f docker/docker-compose.yml up --build

For LLM nodepacks, use:

OPTIONAL_DEPS=llm docker compose -f docker/docker-compose.yml up --build

For using nodepacks with a Pytorch GPU backend, use:

OPTIONAL_DEPS=llm,z-image docker compose -f docker/docker-compose-torch-gpu.yml up --build

Access the editor at http://localhost:5173

Local Development

Backend:

uv sync
psynapse-backend run --reload

Frontend:

cd frontend
npm install
npm run dev

Creating Custom Nodepacks

Add a new directory in nodepacks/ with an ops.py file:

def my_function(text: str, count: int = 1) -> str:
    """Repeats text multiple times."""
    return text * count

Restart the backend to auto-register your nodes.