Studio

Fit, inspect, and export.

SMILE Studio is an integrated desktop IDE for machine learning and data science. It combines an interactive multi-language notebook, an AI-powered agent panel, a file explorer, and a kernel explorer in a single window. Launch it from the release package:

cd bin
./setup    # first-time: installs native libraries
./smile    # launches SMILE Studio
    

SMILE Studio requires Java 25 and a graphical (non-headless) environment. At least 4 GB RAM; 8 GB recommended for large datasets. Open Studio from your project directory so the file tree, notebooks, and SMILE.md memory file all live next to your data.

Your First Session

1. Download the latest smile-<version>.zip from the releases page and unzip it.

2. Run first-time setup once (BLAS/LAPACK, Python venv — see Quick Start):

   Bash Copy

   
   /path/to/smile/bin/setup
       

3. Create a project folder and put datasets in input/:

   Bash Copy

   
   mkdir -p myproject/input
   # copy your CSV / Parquet / … files into myproject/input/
       

4. Launch Studio from the project directory (so the file tree and SMILE.md sit next to your data):

   Bash Copy

   
   cd myproject
   /path/to/smile/bin/smile
       

5. Configure an AI provider under File > Settings… (needed for agents, cell prompts, and Tab completion).

6. In Clair’s tab, initialize the project, then explore with natural language, AutoML, or a notebook:

   Agent Copy

   
   /init
   /automl
       

   Or open a notebook and work through the Tutorial examples cell by cell.

Application Layout

Using the Notebook

File > New creates an Untitled.java notebook with SMILE imports already filled in. File > Open… or double-click a file in the Project Explorer loads .java / .jsh, .scala / .sc, .py, or .ipynb. Other text files open in the Notepad instead of the notebook.

Each cell has a type combobox (Code, Markdown, or Raw), a Prompt field for AI code generation, and a toolbar to run, move, clear, or delete the cell. Execution is asynchronous; a second Run while a cell is still running shows a warning rather than overlapping jobs. After a successful snippet, Studio prints the last value (for example ⇒ DataFrame df = [200 rows × 5 columns]).

Save with File > Save. Enable File > Auto Save to write every 60 seconds (only notebooks that already have a path). If another process changes an open file, Studio asks whether to reload.

Project and Kernel Explorers

The left split has two tabs:

• Project — file tree of the working directory. Double-click a notebook/script to open it; double-click other text files to open them in Notepad.
• Kernel — live variables from the active notebook, grouped as DataFrames, Matrix, Models, and Services. Double-click a DataFrame or matrix to open a viewer; double-click a Model to save a .sml file; double-click a Service to start a REST inference server for that model.

Execution Kernels

Key Shortcuts

Magic Commands

Notebook cells can run kernel-specific magics without leaving the editor.

• Java
• Python

//!mvn org.apache.commons:commons-math3:3.6.1
    

%pip install pandas
%time df.describe()
%%timeit
model.fit(X, y)
    

AI Agents

The right-hand Agent Panel ships with seven built-in agents, each in its own tab. They appear in the order Frank, Steve, Clair, Ada, James, Guido, Chuck. Switch between them depending on the task:

All seven require an AI provider (OpenAI, Anthropic, Google Gemini, Azure OpenAI, or Vertex AI) and API key under File > Settings…. MCP (Model Context Protocol) servers can be wired in via .smile/mcp.json for extra tools.

Prompts

You can drive the agents from two places:

• Notebook cell Prompt field — type a natural-language description (e.g. “load iris.csv into a DataFrame”) and press Enter. Generated code streams into the cell editor. Press Tab in the editor for single-line completion.
• Agent Panel — send an Instructions turn (legend >) for a conversation, or a Shell turn (legend !) to run a subprocess. Agents keep project context in SMILE.md.

Slash Commands

In the Agent Panel, type / followed by a command and press Ctrl+Enter. A hint tooltip appears as you type. Session, memory, and CLI commands are shared; the longer skill list below is what Clair exposes for end-to-end ML work. Type /help in a given agent tab for that agent’s full set.

Session, memory, and tools

Train, predict, and serve

Clair skills