Kaggle Studio

Enables developers to edit Jupyter notebooks locally in VS Code while submitting them to Kaggle's cloud infrastructure for execution, with dynamic GPU/TPU/CPU selection via kaggle.yml configuration. The extension reads .ipynb files from the local filesystem, serializes them via the Kaggle API client, and pushes them to Kaggle's kernel execution service, which handles environment setup, dependency resolution, and compute allocation. Results are automatically downloaded to a .kaggle-outputs/ directory for local inspection.

Manages Kaggle API token authentication through two configurable methods: file-based credentials stored at ~/.kaggle/kaggle.json (recommended for persistent, shared environments) or in-memory credentials via VS Code's built-in credential storage (for ephemeral or single-user setups). The extension validates tokens by calling Kaggle's API status endpoint and provides Sign In/Sign Out commands to manage credentials without manual file editing. Expired tokens trigger 401 Unauthorized errors, requiring manual regeneration from kaggle.com/settings/account.

Scaffolds new Kaggle projects by generating kaggle.yml and kernel-metadata.json configuration files that define project identity, compute requirements, dataset dependencies, and internet access policies. The 'Init Project' command creates these files in the workspace root with sensible defaults; the 'Link Notebook' command associates an existing Kaggle notebook with the local project by populating kernel_slug. The extension reads and validates these YAML/JSON files on startup to configure subsequent operations (execution, dataset attachment, submission).

Provides a searchable sidebar tree view of Kaggle datasets filtered by name, owner, and competition context. Users can browse dataset metadata (size, file count, description) without downloading, attach datasets to projects by adding them to the kaggle.yml datasets array, and download entire datasets to the local workspace via the 'Download Dataset' command. The extension uses Kaggle's dataset API to list available datasets and the dataset download API to fetch files, with progress indication in the VS Code status bar.

Displays a sidebar tree view of Kaggle competitions filtered by status (entered, featured, all) and searchable by name. Users can submit predictions to competitions directly from VS Code via the 'Submit to Competition' command, which uploads a CSV file and returns a submission ID and leaderboard score. The extension tracks submission history in a 'Runs' tree view, showing execution timestamps, compute resources used, and output file locations.

Maintains a 'Runs' tree view that displays all notebook executions triggered from VS Code, including execution timestamp, compute resource used (GPU/TPU/CPU), execution status (running, completed, failed), and output file location in .kaggle-outputs/. Users can click on a run to view its outputs or logs. The extension queries Kaggle's kernel execution API to populate this view and polls for status updates until execution completes.

Adds a rocket button (🚀) to the VS Code notebook editor toolbar that triggers immediate execution of the current notebook on Kaggle infrastructure. Clicking the button is equivalent to running the 'Kaggle: Run Current Notebook' command, which reads the .ipynb file, validates the kaggle.yml configuration, and submits the notebook to Kaggle's kernel execution API. The extension displays execution progress in the status bar and automatically downloads outputs to .kaggle-outputs/ when complete.

The 'Push & Run' command combines notebook upload and execution into a single operation: it reads the local .ipynb file, pushes it to Kaggle via the API, triggers execution with the compute resources specified in kaggle.yml, monitors execution status via polling, and automatically downloads all output files (including the executed notebook with cell outputs) to the .kaggle-outputs/ directory when complete. This eliminates the need for separate push and run commands.

The 'Download Outputs' command retrieves all output files generated by a notebook execution from Kaggle's storage and saves them to the .kaggle-outputs/ directory in the local workspace. The extension automatically creates this directory if it does not exist and organizes files by execution ID or timestamp (exact organization is undocumented). Users can also manually download outputs from the 'Runs' tree view by clicking on a specific run.

Provides three searchable sidebar tree views that display Kaggle resources without leaving VS Code: 'My Notebooks' (lists user's notebooks filtered by language, type, competition), 'Datasets' (searchable dataset browser), and 'Competitions' (lists entered and featured competitions). Each tree view supports incremental search by name, owner, or competition context. Clicking on a resource in the tree view opens its details or triggers an action (e.g., clicking a notebook opens it, clicking a dataset attaches it).