AI Assistants

To accelerate your research, JupyterHub ships with built-in, on-premise Generative AI assistants: Jupyter AI, Claude Code, and Posit Assistant.

Jupyter AI (Chat & Cell Magic in JupyterLab)Copy Anchor Link

Best for: Generating plotting code, explaining errors, writing documentation, and quick code snippets.

1. Spawn a server using the DataScience NB image with the AI badge.
2. Look for the Jupyternaut chat icon in the left sidebar to open an interactive chat panel.

Try It Out: Practical ExamplesCopy Anchor Link

• Sidebar Chat: Click the chat icon, highlight a block of messy analysis code in your notebook, and ask:

  “Refactor this pandas code to be faster and add docstrings explaining the statistical transformations.”

• Learn from Local Files: In the chat panel, type /learn followed by a path to your local documentation or .py scripts so the assistant can answer questions about your specific codebase.
• In-Cell %%ai Magic: You can prompt the LLM directly inside a notebook code cell (refer to the Jupyter AI docs for full syntax):

  Copy Text

  # Load the AI extension in a cell, then generate a publication-ready plot:
  %load_ext jupyter_ai_magics

  Copy Text

  %%ai
  Write Python code using seaborn and matplotlib to create a Raincloud plot 
  comparing gene expression levels across 3 treatment groups from a DataFrame `df`.
  Include ANOVA p-value annotations.

Claude Code (Agentic AI for Complex Workflows)Copy Anchor Link

Best for: Multi-file refactoring, building end-to-end data pipelines, debugging across scripts, and exploring unfamiliar datasets.

Unlike a simple chat window, Claude Code is an agentic assistant that can read your notebooks, inspect data schemas, and navigate your project directory.

• Human-in-the-Loop Safety: Claude Code cannot execute notebook cells or modify critical files without your explicit approval. It will always ask for your permission before making changes.

Launching Claude Code in JupyterLabCopy Anchor Link

1. Start a server with the DataScience NB image with the AI badge.
2. Open your project folder (working inside a dedicated project folder helps the agent understand your file structure).
3. Open a terminal or the Claude interface to begin interacting with your codebase. (See the e-INFRA CZ Blog for more visual walkthroughs).

Launching Claude Code in RStudioCopy Anchor Link

1. Go to hub.cloud.e-infra.cz. Select the RStudio image with the AI badge (version 4.6.1). Configure your required server resources and click Start.

2. In RStudio, create a project context: go to File → New Project → New Directory → New Project, name your folder inside your Home directory, and click Create Project.
3. Open the Terminal tab inside RStudio, type claude, and press Enter:

   Copy Text

   claude

4. Follow the on-screen prompts to approve permissions.

Try It Out: Example Prompts for Claude CodeCopy Anchor Link

Once claude is running in your project terminal, try typing natural-language tasks:

• Data Cleaning Pipeline:

  “Inspect the CSV files in ./raw_data/. Some have missing timestamps and inconsistent column names. Write a Python script clean_data.py to standardize them and save the output as a Parquet file.”

• Debugging & Optimization:

  “Read simulation.py. It runs out of memory when N > 10000. Identify the memory bottleneck and rewrite it using NumPy vectorization or chunking.”

• R Analysis:

  “Look at my analysis.R script and the dataset clinical_trial.csv. Add a section that fits a mixed-effects model using lme4 and outputs a summary table.”

Posit Assistant in RStudioCopy Anchor Link

Best for: Native R coding assistance inside the RStudio IDE.

1. Spawn a server with the RStudio image with the AI badge (version 4.6.1).
2. Once RStudio opens, click Posit Assistant in the top-right corner.
3. Click Install Posit Assistant, and when prompted, select Trust Workspace.
4. Select an On-Premise Model:
   • Click the ... (three dots) icon in the bottom-left corner of the Assistant panel.
   • Under the OpenAI Compatible category, select mini or GPT-OSS-120B (recommended).

VS Code with LLM IntegrationCopy Anchor Link

To use AI assistants inside a browser-based VS Code environment, spawn a server with a VS Code image and follow our dedicated LLM Integration Guide.