AI Agentic Data Analysis (Jupyter + SSH Service) ================================================= Data collected in ARIM-mdx Nextcloud can be analyzed through ARIM-mdx Jupyter. It also supports long-running autonomous execution by AI agents. ARIM-mdx Jupyter provides: - `Jupyter Lab `_ : Interactive data analysis, e.g., visualization - `VSCode `_ : Software development - `MateriApps LIVE! `_ (Remote Desktop): All-in-one software package for materials sciecne Jupyter Initial Setup ----------------------- Jupyter-Nextcloud integration is required as initial setup. Please refer to :ref:`jupyter-activation` for details. Logging into Jupyter --------------------- | Log in to ARIM-mdx Jupyter service (https://jupyter.arim.mdx.jp). .. image:: _static/login-6.png :alt: login-6 :align: center :width: 400px | Starting Jupyter Server ------------------------ | Select a server and click Start to launch. Below are the server types for different purposes. .. image:: _static/jupyter-spec.png :alt: jupyter-spec :align: center :width: 500px | .. list-table:: :header-rows: 1 :widths: 40 60 * - Type - Description * - Jupyter / VSCode / Remote Desktop - Environment including Remote Desktop (`MateriApps LIVE! `_) * - Jupyter / VSCode Only - Environment with only JupyterLab and VSCode .. list-table:: :header-rows: 1 :widths: 20 40 40 * - Type - Specifications - Description * - Default - 16 vCPUs, 24GB memory - Standard computing environment * - High Spec - 32 vCPUs, 48GB memory - High-performance PC equivalent * - Max Spec - 64 vCPUs, 96GB memory - High-performance server equivalent .. note:: Starting Jupyter may take up to several minutes. Starting Jupyter VSCode ------------------------ | Select **VSCode** from the Launcher to start. .. image:: _static/vscode.png :alt: vscode :align: center :width: 500px | | VSCode will launch. .. image:: _static/vscode-run.png :alt: vscode-run :align: center :width: 500px | Starting MateriApps LIVE! on Jupyter Remote Desktop ---------------------------------------------------- | Select **Remote Desktop** from the Launcher to start. .. image:: _static/vscode.png :alt: vscode :align: center :width: 500px | | MateriApps LIVE! on Jupyter Remote Desktop will launch. .. image:: _static/remote-desktop.png :alt: remote-desktop :align: center :width: 500px | SSH Access ----------- | ARIM-mdx Jupyter also accepts direct SSH connections. .. note:: The web interface at `https://jupyter.arim.mdx.jp `_ remains the primary way to use ARIM-mdx Jupyter. SSH is an optional additional path. Prerequisites for SSH ---------------------- 1. Complete the Jupyter initial setup (see :ref:`jupyter-activation`). 2. Start a Jupyter server from the web UI — the SSH endpoint is tied to your running Jupyter server. 3. Generate an SSH keypair on your local machine if you do not already have one: .. code-block:: bash ssh-keygen -t ed25519 4. Append the **public key** (``~/.ssh/id_ed25519.pub``) to ``~/.ssh/authorized_keys`` on the Jupyter server: .. code-block:: bash echo "" >> ~/.ssh/authorized_keys SSH Connection Information --------------------------- .. list-table:: :header-rows: 1 :widths: 25 75 * - Item - Value * - Host - ```` * - Port - ```` * - Username - ```` * - Authentication - SSH public key (password authentication is disabled) Connecting with VSCode Remote-SSH ---------------------------------- 1. Install the `Remote - SSH `_ extension in VSCode. 2. Add the following entry to your local ``~/.ssh/config`` (replace the ```` values): .. code-block:: text Host arim-jupyter HostName Port User IdentityFile ~/.ssh/id_ed25519 3. Open the Command Palette in VSCode and run **Remote-SSH: Connect to Host...**, then select ``arim-jupyter``. .. TODO: add screenshot _static/ssh-vscode.png once the flow is demonstrable Connecting from Terminal ------------------------- Use a standard ``ssh`` command: .. code-block:: bash ssh @ -p .. TODO: add screenshot _static/ssh-terminal.png once the flow is demonstrable File Transfer (SFTP / scp) --------------------------- Upload a local file to your Jupyter home directory: .. code-block:: bash scp -P ./local-file.csv @:~/ Download a file from your Jupyter home directory: .. code-block:: bash scp -P @:~/output.csv . Interactive SFTP session: .. code-block:: bash sftp -P @ Files transferred here are visible from inside the Jupyter server at the same path.