Skip to content
IsaacHuoPublic

About

A local-first open research workspace connecting research tools, AI agents, literature, data and compute. Foundation stage.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

14 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

OpenLab

OpenLab flask and wave mark

The Open Research Operating System

Repository checks License: Apache-2.0 Latest release

简体中文 · Product specification · Architecture · Contributing

OpenLab is a local-first research workspace being designed to connect open research tools, AI agents, literature, data, and compute. Research projects and their evidence belong to the researcher and should remain usable when a model or tool changes.

Current status

This source is the v0.1.0-alpha.1 research preview. It runs locally with Node.js 24 and a browser; no installer is provided. The historical v0.0.1 release remains foundation-only. See the changelog for the preview boundary and PLAN for delivery progress.

The published source prerelease includes the reviewed notes and GitHub source archives.

The preview combines projects, a supervised Harness, DeepSeek-tested model configuration, Crossref metadata retrieval, plan confirmation, source-linked Markdown reports and portable research archives. See development and the implementation plan for tested scope and remaining gates.

Capability Status
Contribution, issue, review, and release processes Available in this repository
Automated documentation, workflow, and secret checks Available in GitHub Actions and local tooling
Local service with browser UI, Windows first Implemented in development source; desktop/mobile Chrome checked
Projects, pinned inert skill installation, model configuration, and metadata research Implemented; two live DeepSeek/Crossref tasks checked
OpenLab Harness: default single-agent literature runtime Implemented with supervised native Windows worker and replaceable interfaces
Supervisor-Skills execution through the platform runtime adapter Planned; not an endorsed or verified integration yet
Conversation, execution state, and live artifacts workspace React/assistant-ui with AG-UI; plan, input, denial and replay checked
Source tracking, execution history, and artifact export Immutable revisions, material copies and verified archive round trips
Markdown/LaTeX document plugins Recoverable drafts, immutable manual revisions, comparison, BusyTeX and PDF.js; portable v2 archives with v1 imports

Run the development preview

Use Node.js 24 on Windows:

npm ci
npm run build
npm start

Open the launch link printed in the terminal. Add and test your model provider in settings, create a project, and submit a question. Review the plan before research starts. Model planning and execution use your provider account. Windows protects credentials under the current user; saved projects remain readable without provider access. Crossref retrieval needs a network connection. Public-page reading asks for scoped access and can be unavailable; metadata alone does not verify a paper's conclusions. DeepSeek currently uses non-thinking mode.

For direct metadata search, select the metadata tool without a model credential. Document editing and TeX compilation also require no model. Before first TeX use, optionally run npm run plugins:prepare: it downloads about 203 MB of pinned components using curl, separately from the source archive. The workspace can prepare them on demand; failed downloads retain ranges for an explicit retry. See compiler setup and limits. On PowerShell, use npm.cmd if execution policy blocks npm.ps1.

Native Windows preview with a real DeepSeek/Crossref metadata research record

This screenshot shows a real metadata-only task, not a full-text or scientific claim-verification result. Verification details distinguish live probes from synthetic browser fixtures.

First research release

The first application milestone targets a literature research loop:

  1. Create a project and describe a research question.
  2. Install a versioned skill and configure a model and retrieval provider.
  3. Review the research plan and requested access.
  4. Follow execution, respond to questions, or stop the task.
  5. Inspect sources, distinguish verified evidence from inference, and export the research record.

The initial application is intended to start locally and open in a browser. Users bring their own model credentials. Existing projects and artifacts should remain readable offline; model execution and online retrieval require their configured services.

Architecture direction

Planned OpenLab architecture

OpenLab is an open research platform, not a wrapper around a single agent framework. It will own projects, conversations, execution records, sources, artifacts and provenance. Agent runtimes, models, scientific tools and user interfaces must be independently replaceable; research records remain owned by OpenLab and its users.

The planned workspace combines project navigation, conversation with execution state, and live artifacts. assistant-ui provides the conversation foundation; the Platform gateway uses AG-UI. shadcn/ui and AI Elements supply supporting views, while CopilotKit is an interaction reference without an SDK dependency. The first runtime is the minimal self-developed OpenLab Harness behind RuntimeAdapter; OpenCode is a later optional adapter. See architecture and ADR 0002 for the current design and its implementation gates.

Documentation

Document Purpose
Development execution plan Current phase, ordered tasks, acceptance exits and retained implementation history
Original vision, Chinese Founding product vision, preserved as supplied
Product specification Users, first milestone, behavior, and acceptance criteria
Architecture and contracts Components, research objects, and runtime interface
Research workspace Three-region layout, interaction states, live artifacts and UI acceptance
OpenLab Harness Minimal first runtime, replaceable capabilities and execution gates
Technical decisions Accepted baseline and implementation gates
Workspace and Harness decision Current UI/protocol and default-runtime selection; supersedes OpenCode first
Logging and provenance Diagnostics, research records, retention, and export
Product mark Vector assets, exact symmetry, and reproduction
Repository operations GitHub settings, labels, checks, and release procedure
Verification record Baseline verification and remaining limitations
Changelog Delivered changes by version

Contribute

For repository tooling, install Node.js 24 LTS, clone the repository, and run:

npm ci
npm run check
npm run check:actions
npm run check:secrets

The last two commands download checksum-verified tool releases into an ignored local cache. See CONTRIBUTING for Windows instructions and review expectations. Start with an issue or discussion. English and Chinese contributions are welcome.

License and attribution

OpenLab-authored code and documentation use Apache-2.0, unless a file explicitly states otherwise. Third-party skills, tools, models, and research material retain their own licenses; installing them does not relicense them under Apache-2.0.

Supervisor-Skills is the planned first integration reference. Its repository and individual files have differing license declarations that must be resolved before distributing an integration. Its content is not bundled here. See third-party notices and contributors.

About

A local-first open research workspace connecting research tools, AI agents, literature, data and compute. Foundation stage.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages