Discover and explore top open-source AI tools and projects—updated daily.
ai4s-researchAI workbench for reproducible scientific research
New!
Top 55.0% on SourcePulse
Summary
Open Science provides an open-source, local-first AI research workbench, acting as a reproducible alternative to proprietary tools like Claude Science. It targets scientists and researchers, enabling rigorous, auditable workflows by integrating literature, code, figures, and reports.
How It Works
This workbench uses a plan-approve-execute-artifacts-review cycle, emphasizing traceable artifacts linked to their origin (code, data, conversation). Its local-first design keeps data and compute on the user's machine. Model-agnostic support is via OpenRouter, OpenAI-compatible APIs, or local models, built on Tauri and an agent skills framework.
Quick Start & Requirements
Installers are available for macOS (.dmg) and Windows (.exe/.msi), with Linux builds also provided. Requires macOS 13+ or Windows 10/11 (x64). Building from source needs Node.js ≥ 20, pnpm 9, and Rust. macOS users may need xattr -cr "/Applications/Open Science.app" to bypass Gatekeeper.
Highlighted Details
provenance.jsonl) for recoverable artifact versions.Maintenance & Community
Currently v0.1, in active development as a macOS MVP. Future plans include R kernel, Windows installer, and HPC support. Discussed within the linux.do community. Contributions are welcomed.
Licensing & Compatibility
Licensed under MIT. Third-party components have their own licenses. Described as "beta research tooling"; outputs require verification and expert review before submission, advising caution for critical closed-source integration.
Limitations & Caveats
Beta tooling requires output verification. Unsigned builds may trigger OS warnings. The OpenCode agent runtime requires macOS 13+ or Windows 10/11 (x64). A dedicated Windows installer is pending.
19 hours ago
Inactive