DeepSeekSoftwareShipping codeAI score · 81/100AI-assisted edit · no obvious prompt-injection riskGitHub

Install dsh-notebook plugin in DSH for a persistent Jupyter workspace

Input: an existing DSH environment; Process: install plugin and configure kernel; Output: an Agent-controllable notebook workspace with read/edit/execute cells

ExperiencedStepsWorkflowTip
Who
Data scientists or developers with DeepSeek Harness (DSH) who need a persistent Jupyter kernel inside Agent sessions
You get
A Notebook tab in DSH Web with Python/R kernels, persistent variables across cells, and Agent tools to read/edit/execute cells
You need
DeepSeek

Constraints

Prerequisites: DSH (DeepSeek Harness) 0.1.x-rc or 0.2.0-rc installed and configured. Node.js/npm required for dsh plugin command. Python kernel needs ipykernel and jupyter_client in the target environment; R kernel needs IRkernel. Optional dsh-envsel plugin for conda environment selection. Cost: MIT licensed, no extra fees.

What they did

The author built dsh-notebook, a DSH plugin that embeds a persistent Jupyter kernel (ipykernel + jupyter_client sidecar) into the DeepSeek Harness web interface. It provides a Notebook tab, a set of Agent-callable tools (read/edit/execute cells, runtime introspection, version control), VS Code-aligned cell UI, and dual Python/R kernel support.

How it went

The author reports that the Agent can drive a live computational session: find relevant cells, read source and context, edit cells, execute them, check stdout/traceback, and iterate when needed. Because the kernel is persistent, the Agent operates on stateful runtime objects (e.g., Anndata, GPU models, loaded data) rather than merely assembling strings. (author_report, not tested by this site)

Worth knowing

DSH 0.2.0 support was added in v0.3.0; R kernel support in v0.4.0. nbformat is no longer a required dependency. Each DSH session owns an independent notebook, kernel, and working directory with no cross-session leakage. If dsh web is force-killed, the Python sidecar detects parent death and auto-shuts-down ipykernel to avoid orphan processes. The source does not provide DSH installation instructions, nor a sample notebook or verified reproduction steps.

Steps

  1. Install the plugin

    Run the install command in a terminal to add the dsh-notebook plugin to DSH's web profile.

    Setup guidance · Source text。Terminal command line

    dsh plugin --profile web add @beihaizb/[email protected]

  2. Restart DSH Web

    After installation, restart dsh web. A Notebook tab will appear at the top of DSH sessions.

  3. Select a kernel

    Open the Notebook tab and pick a kernel from the toolbar dropdown. If dsh-envsel is installed, the current session's selected conda environment is used by default; otherwise it falls back to the first discovered conda env with ipykernel. Both Python (ipykernel + jupyter_client) and R (IRkernel) are supported and switchable at any time.

    Setup guidance · Source text。Terminal command line (optional dependency)

    dsh plugin --profile web add @beihaizb/dsh-envsel

  4. Use Notebook and Agent tools

    Create, edit, and run cells in the Notebook. The DSH Agent can interact via tools: nb_get to read, nb_edit_cell to edit, nb_run_cell to execute, nb_context for runtime state, etc. Cell version history (nb_cell_history / nb_revert_cell), error context (nb_error_context), and batch execution (nb_run_all) are also available.

Prompt you can copy

Reference prompt · Editorial suggestion。Confirm the destination, input variables and output format before use.

This block is the prompt. Ads are never placed here.

No prompt needed — just follow the steps.

Copy the install command into your terminal; the Notebook tab appears after restarting DSH Web and selecting a kernel.

Source

安装 dsh-notebook，在 DSH 里跑 Jupyter 风格笔记本 · beihzb · GitHub · 2026-10-31T09:09:09Z

Comments & feedback

Tried this method? Share results, problems or improvements. Comments appear after moderation. Please avoid sensitive information.

Loading…

    Related