Skip to content

Releases: CUHK-AIM-Group/NeuroDiscovery

NeuroDiscovery v1.0.0

Choose a tag to compare

@chengwang96 chengwang96 released this 16 Sep 12:25

NeuroDiscovery v1.0.0

Highlights

  • Refreshed desktop research workspace and project-centered workflows
  • Improved AutoResearch continuity and expert-study workflows
  • Added three configurable hypothesis novelty modes: strict, novelty_first, and weighted
  • Expanded NeuroOracle graph exploration and evidence-grounded hypothesis generation
  • Added academic literature integration through MCP
  • Added CTP processing and clinical research outputs
  • Expanded statistical ML, survival analysis, causal treatment, temporal modeling, imaging genetics, connectome discovery, and brain-age workflows
  • Expanded NeuroBench to 500 research and engineering tasks
  • Improved runtime setup, usage monitoring, and model-provider support, including Ollama Cloud

macOS Apple Silicon

Download either:

  • NeuroDiscovery-1.0.0-arm64.dmg
  • NeuroDiscovery-1.0.0-arm64-mac.zip

The current macOS build is unsigned and not notarized. If macOS blocks the first launch, open System Settings → Privacy & Security and select Open Anyway.

NeuroClaw v0.2.1

Choose a tag to compare

@chengwang96 chengwang96 released this 14 Jul 00:14

NeuroClaw Desktop v0.2.1

This release adds scoped autoresearch guidance and improves desktop workflows and reliability.

Highlights

  • Added /help guidance for data processing, model development, idea generation, and end-to-end autoresearch.
  • Added / command autocomplete with English and Chinese instructions.
  • Added parallel execution across multiple chats.
  • Added project creation using existing or new folders.
  • Added drag-and-drop attachments with support for all file formats.
  • Added model discovery and alphabetical model sorting.
  • Improved chat restoration, error recovery, workspace handling, and Windows shell detection.
  • Collapsed tool execution details by default.
  • Removed the desktop requirement for project-level environment configuration.

Downloads

For Windows users:

  • NeuroClaw Setup 0.2.1.exe is recommended for normal installation.
  • NeuroClaw 0.2.1.exe is the portable version. It may take longer to start because it extracts the app before opening.
  • NeuroClaw-0.2.1-win.zip contains the unpacked application.

Notes

The bundled runtime has been rebuilt. Release files were checked for personal configuration, hardcoded credentials, and local development paths.

NeuroClaw v0.2

Choose a tag to compare

@chengwang96 chengwang96 released this 20 Jun 16:02

NeuroClaw Desktop v0.2

This release introduces the first cross-platform NeuroClaw desktop release for Windows and macOS.

Highlights

  • Added Windows and macOS desktop clients.
  • Added Electron-based desktop shell for launching and using NeuroClaw locally.
  • Added desktop settings for runtime mode, Python path, FSL path, proxy, language, and text size.
  • Added local runtime startup and health detection.
  • Added built-in Python runtime support for easier first-time setup.
  • Added project-aware chat workspace with separate Projects and Chats sections.
  • Added Skills page with categorized skill descriptions.
  • Added NeuroOracle entry point with graph file detection, download, and update flow.
  • Added native right-click context menu.
  • Added single-instance desktop behavior.
  • Added English and Chinese desktop UI support.

Downloads

For Windows users:

  • NeuroClaw Setup 0.2.0.exe is recommended for normal installation.
  • NeuroClaw 0.2.0.exe is the portable version. It may take longer to start because it extracts the app before opening.

For macOS users:

  • Use the macOS build artifact uploaded with this release.

Notes

Linux is still supported through the repository, command-line workflow, and web interface. A dedicated Linux desktop package is not included in this release.

This is an early desktop release. Some advanced workflows still depend on local runtime configuration, optional neuroscience tools such as FSL, and configured model endpoints.