Neuro-Workflow

A second-generation brain model builder — organizing multi-scale computational neuroscience as a graph of reusable, schema-defined components, designed to be understood and operated by both humans and AI agents.

---

Why Neuro-Workflow?

Brain modeling today is fragmented. NEST, TVB, NEURON, and analysis tools each have separate APIs, data formats, and execution models.

Neuro-Workflow is a second-generation model builder. Unlike first-generation tools designed solely for human users (i.e. SNNbuilder), it organizes multi-scale brain modeling as a graph of reusable components — each a well-defined Python class with a schema describing its role, inputs, outputs, and parameters. This architecture was built from the ground up to be understood and operated by both humans and AI agents.

The key innovation is not the addition of LLMs — it is the AI-ready infrastructure. Because every node carries structured metadata, AI agents can support the modeling process through few-shot learning and protocols such as MCP (Model Context Protocol) without deep domain fine-tuning. Even small or locally deployed models can perform well, keeping computational overhead and token costs low.

This architecture enables:

• Simulator interoperability — NEST, TVB, NEURON, and custom solvers run as interchangeable nodes through a unified interface
• Human + AI collaboration — users and agents compose nodes into models, generate executable Python scripts and notebooks, and run simulations
• AI-assisted parametrization — agents retrieve parameter values from open data sources and suggest configurations grounded in the literature
• Reproducibility by design — workflows are serializable graphs; the same pipeline runs on a laptop or a supercomputer
• Extensibility — any Python function becomes a node; new simulators integrate without changing the core

"By providing well-documented, schema-defined nodes, Neuro-Workflow establishes a foundation for systematically organizing computational neuroscience functions, algorithms, and tools — enabling AI-augmented scientific discovery in which humans and agents jointly build, test, and extend brain models."

---

Support and Development

This project is supported by the Brain/MINDS 2.0 initiative and is being developed by the Neural Computation Unit at the Okinawa Institute of Science and Technology (OIST) in collaboration with partners.

---

Preview

Get a first impression of Neuro-Workflow in action:

🎥 Video demonstrations:

Basal Ganglia Model of the Macaque on Neuro-Workflow using NEST
Credits: Carlos Enrique Gutierrez

Marmoset Full-Brain Model on Neuro-Workflow using TVB
Credits: Carlos Enrique Gutierrez and Henrik Skibbe

First View of Neuro-Workflow
Credits: Carlos Enrique Gutierrez

---

Current Status

Neuro-Workflow Python API

Neuro-Workflow provides a comprehensive Python API for building and executing computational neuroscience workflows using a node-based system. The core functionality is organized as follows:

Node System

• Node Storage: All available nodes are stored in src/neuroworkflow/nodes/
• Organization: Nodes are organized in customizable categories for easy navigation
• Extensibility: New custom nodes can be created and integrated into the system

Creating Custom Nodes

For developers interested in extending Neuro-Workflow with custom functionality:

• 📋 Node Schema: See NODE_SCHEMA.md for detailed node structure specifications
• 📝 Template: Use CustomNodeTemplate.py as a starting point for new nodes
• 📖 Tutorial: Follow CUSTOM_NODE_TUTORIAL.md for step-by-step node creation guide

Python API Examples

The following examples demonstrate how to use the Neuro-Workflow Python API to create and execute workflows:

Examples folder:

• sonata_simulation.py - Basic simulation example
• neuron_optimization.py - Parameter optimization example (in development)
• epilepsy_rs.py - Epileptic resting state simulation using The Virtual Brain (TVB)

Notebooks folder:

• 01_Basic_Simulation.ipynb - Interactive basic simulation tutorial
• epilepsy_rs.ipynb - Interactive epileptic resting state example with TVB
• SNNbuilder_example1.ipynb - Spiking Neural Network building with SNNbuilder custom nodes

Neuro-Workflow Web Application

For users who prefer a graphical interface, Neuro-Workflow includes a comprehensive web application that provides visual workflow building capabilities.

Installation

To set up the web application, follow the detailed instructions in gui/README.md.

Important Setup Notes

Node Synchronization:

• The web app requires nodes to be copied from src/neuroworkflow/nodes/ to gui/workflow_backend/django-project/codes/nodes/
• This copy is regularly performed by administrators
• For developers: If you create new custom nodes, ensure they are copied to the web app directory to make them available in the GUI

Core API Synchronization:

• The Python API base code from src/neuroworkflow/core/ is also copied to the web application
• Web app location: gui/workflow_backend/django-project/codes/neuroworkflow/core/
• This ensures the web app stays synchronized with the latest API updates

---

Conference Presentations

This work has been presented at several conferences and workshops, receiving valuable feedback that has contributed to its ongoing development:

2026

• Unified Theory Workshop (April 23, 2026)

  • "NeuroWorkflow: Agent-Assisted Brain Modeling"
  • 📄 Poster

2025

• INCF/EBrains Summit

  • "NeuroWorkflow: A Node-Based Framework for Scalable Computational Neuroscience with AI-Ready Infrastructure"
  • 📄 Abstract
  • 📄 Poster

• RIKEN CBS Hackathon (September 28, 2025)

  • "Building BrainModeling Workflows: A proof-of-concept framework"
  • 📄 Hackathon Material

• CNS 2025 (Computational Neuroscience Society)

  • "A Graph-Based, In-Memory Workflow Library for Brain/MINDS 2.0 – The Japan Digital Brain Project"
  • 📄 Poster

• NEST Conference 2025 (June 17, 2025)

  • "A Graph-Based, In-Memory Workflow Library for Brain/MINDS 2.0"
  • 📄 Presentation Slides

• Unified Theory Workshop (May 30, 2025)

  • "NeuroWorkflow: A python-based Graph Framework for Modular Brain Modeling Workflows"
  • 📄 Poster

• Winter Workshop

  • "Towards a Generic and Open Software for Building Digital Brains"
  • 📄 Poster

---

Publications

Neuro-Workflow is currently under preparation for publication. If you use it in your research, please check back for the citation or contact us.

Related Publications

• Gutierrez et al. (2022). A Spiking Neural Network Builder for Systematic Data-to-Model Workflow. Frontiers in Neuroinformatics. https://doi.org/10.3389/fninf.2022.855765

• Gutierrez et al. (2025). Topological basal ganglia model with dopamine-modulated spike-timing-dependent plasticity reproduces reinforcement learning, discriminatory learning, and neuropsychiatric disorders. bioRxiv. https://doi.org/10.1101/2025.11.10.687760

---

License

This project is licensed under the GNU Affero General Public License v3.0 or later (AGPL-3.0-or-later) - see the LICENSE file for details.