lzszq/dsh-scholar
152Last commit Aug 15, 2026
dsh-scholar DSH plugin
DSH Scholar is a research workbench for computational studies. It organizes research materials, project conversations, code and data, experiment runs, evidence ledger, and TeX manuscripts within a single recoverable project. It can start from a new problem or resume an ongoing study.
How to install the dsh-scholar DSH plugin
dsh plugin --profile web add /absolute/path/to/dsh-scholarCopying does not run this command. Review the repository and version before installing the dsh-scholar DSH plugin.
dsh-scholar DSH plugin data source
dsh-scholar DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-scholar DSH plugin can do
- Governable research workflow with human gates at key stages (Scope, Idea, Experiment Contract, Evidence, Claim, Release).
- Controlled experiment execution via Docker or remote runners with full logging, status, and artifacts.
- Traceable evidence linking claims to controlled runs, artifacts, and reviewed evidence.
- Integrated workspace with Chat, Workspace, Terminal, Manuscript, Trajectory, and Settings sharing the same project context.
- Recoverable and auditable state via Research Kernel, preserving authority state, next actions, approval traces, and artifact references.
- Plugin configuration for default governance mode, unattended mode, standalone address, and shortcut key.
Where the dsh-scholar DSH plugin fits
- Start a new research project from scratch, filling in a research brief via Grill Me and defining scope.
- Resume an existing project, recovering its stage, sessions, files, and tasks.
- Upload papers, code, data, or logs to integrate into an ongoing research pipeline.
- Run controlled experiments with immutable code snapshots and fixed execution environments.
- Write and compile TeX manuscripts, review evidence, and release findings through a release gate.
Who the dsh-scholar DSH plugin is for
- Researchers in machine learning, data science, bioinformatics, and other computational fields.
- Research teams that require traceable, auditable, and reproducible workflows for their studies.
dsh-scholar DSH plugin limitations
- DSH Scholar assists research but does not replace researchers in scientific judgment, approval, authorship, or publication responsibility.
- Formal experiments must bind immutable code/data snapshots and fixed execution environments, executed by a controlled Runner.
- Product focuses on computational research (ML, data science, bioinformatics) and is not suitable for clinical decisions, human trials, wet lab, or other high-risk research.
- The `@dsh-scholar/*` packages are not yet published; installation requires building from source and adding as a local plugin.
- Requires Linux, Node.js 24, pnpm 11.20.0, and Docker Engine for formal experiments and TeX compilation.
dsh-scholar DSH plugin: from the repository README
Quoted from the lzszq/dsh-scholar README, the upstream source of the dsh-scholar DSH plugin. Copyright remains with the original authors.
> **加急开发适配中！** **简体中文** | [English](README.en.md) DSH Scholar 是面向纯计算研究的 AI 科研工作台。它把研究资料、项目对话、代码与数据、实验运行、证据账本和 TeX 手稿放在同一个可恢复项目中，既可以从新问题开始，也可以接入已经进行到一半的研究。  ## 核心能力 - **可治理的研究流程**：从 Scope、Idea、Experiment Contract 到 Evidence、Claim 和 Release，关键节点由人类 Gate 把关。 - **受控实验**：Runner 在本机 Docker 或受控远端机器上执行冻结的实验计划，记录日志、状态与产物。 - **可追溯证据**：论文主张可追溯到受控 Run、Artifact 和经评审的 Evidence。 - **一体化工作台**：Chat、Workspace、Terminal、Manuscript、Trajectory 和 Settings 共用同一项目上下文。 - **可恢复与可审计**：Research Kernel 保存权威状态、NextAction、审批轨迹和产物引用。 ## 使用边界 - DSH Scholar 辅助研究，不代替研究者承担科学判断、审批、署名和发布责任。 - 默认使用 `gate-only`；Agent 不能批准 Human Gate、伪造 accepted Evidence 或绕过实验合同。 - 正式实验必须绑定不可变代码/数据快照和固定执行环境，并由受控 Runner 真实执行。 - Chat、普通 stdout 和 Interactive Terminal 输出不会自动成为正式 Evidence。 - 产品聚焦机器学习、数据科学、生物信息学等纯计算研究，不适用于临床决策、人体试验、湿实验或其他高风险研究。 ## 快速开始 完整的环境、端口、变量和验收矩阵见 [开发、测试与部署运行规范](docs/test-instance-plan.md)。本地体验需要： - Linux； - Node.js 24； - pnpm 11.20.0； - Docker Engine（正式实验、TeX 编译和 clean-room 复现必需）。 ### 1. 安装与构建 ```bash pnpm install --frozen-lockfile pnpm run build ``` ### 2. 启动独立工作台 ```bash bash scripts/start-standalone-ui.sh ``` 默认页面为 <http://127.0.0.1:18610>，Research
Read the full READMERepository license: MIT
dsh-scholar DSH plugin questions
How do I install DSH Scholar?
Since the `@dsh-scholar/*` packages are not yet published, you need to clone the repository, run `pnpm install --frozen-lockfile` and `pnpm run build`, then add it as a local plugin to DSH using `pnpm dsh plugin --profile web add /absolute/path/to/dsh-scholar`. For the standalone UI, follow the steps in the README to start it without DSH.
Can I use DSH Scholar without DSH?
Yes, you can run the standalone workbench using `bash scripts/start-standalone-ui.sh`. However, this will not include Agent tools, slash commands, Skills, configuration cards, or the 'dsh Scholar' tab. For the full integrated experience, DSH is required.
How do I configure DSH Scholar after installation?
Open DSH Settings → Plugin Configuration → dsh Scholar. You can set the default governance mode, enable unattended mode, change the standalone address, and customize the shortcut key. Changes take effect after restarting DSH.
What are the system requirements?
You need Linux, Node.js 24, pnpm 11.20.0, and Docker Engine (for formal experiments and TeX compilation). The standalone UI requires a token file located at `~/.dsh-scholar-standalone/research-ui-standalone/standalone-token`.
How do I start a research project?
After launching the workbench, you can start a new project by clicking 'Init' and filling in the research brief via Grill Me, resume an existing project, or upload existing papers/code/data. The typical workflow proceeds through scope, idea, experiment contract, run, evidence, and release gates.