Coding Studio
Coding Studio at /codelab is a browser IDE with project files, Monaco editors, Jupyter-like notebooks, and Wasm kernels for Python (Pyodide), SQL (DuckDB), JavaScript/TypeScript, and R (WebR). Ordinary execution stays in your tab — nothing uploads to UtilBloom servers. Use it when transforms need real code beside Data Lab’s click-based profiling. Browser-max packages mean Pyodide/micropip and WebR Wasm binaries, not full desktop pip/CRAN or remote warehouses. Expect download latency and memory limits; heavy training belongs on GPU infrastructure, while AI Research Lab covers lightweight token experiments elsewhere on the site.
What you can do
- Python IDE + Browser-max packages
Multi-file Pyodide workspace with micropip for Wasm and pure-Python wheels, starter packs for data science and plotting, and clear errors when a native wheel is unavailable.
- SQL Studio on DuckDB-WASM
Query local CSV, Parquet, and JSON from the project tree, browse schemas, and keep an OPFS-backed DuckDB database when the browser allows.
- Notebook mode
Jupyter-like cells with markdown, Python, SQL, JS/TS, and R — native .ubnb.json plus .ipynb import/export.
- JavaScript and TypeScript
Sandboxed runtimes with Monaco editing for small algorithms and JSON transforms without installing Node.
- R Studio lite (WebR)
Edit and run R with console output, canvas plots, and packages from the WebR binary repository — not full RStudio Server.
- Local project files
OPFS file tree with drag-and-drop import so CSVs stay on your device and sync into kernels under /data.
Worked examples
CSV summary with Python
pandas read_csv on data/input.csv, groupby region
Import a modest CSV into the project tree, compute mean revenue by region, and hand off to Data Lab or Charts when you want click charts.
DuckDB over a local file
SELECT * FROM read_csv_auto('input.csv') LIMIT 100Drop a CSV into data/, run analytical SQL in SQL Studio, and export results or open the same file in Data Lab.
Notebook + micropip
await micropip.install('networkx') then build a tiny graphUse the Packages panel or a micropip cell; if the wheel is not Wasm-compatible, Coding Studio explains the Browser-max limit.
How it compares
Google Colab, JupyterHub, and Replit offer persistent notebooks, GPUs, and collaboration UtilBloom does not host. VS Code locally remains superior for large monorepos and debuggers. Coding Studio wins when you need a zero-install IDE beside UtilBloom PDF and data tools on the same origin. For spreadsheet-first exploration without code, use /datalab. For ML token math without training clusters, see /ailab. Tool pages under /coding deep-link into studio presets.
Questions
Does Coding Studio send my source to UtilBloom?
Execution is designed to stay in-browser via Pyodide, DuckDB-WASM, WebR, and the JS sandbox. Kernels may download Wasm assets and packages from public CDNs.
Why is the first run slow?
Each kernel downloads a WASM runtime on first use. Subsequent runs in the same tab are faster; a hard refresh may re-download.
Can I pip install arbitrary packages?
Browser-max only: Pyodide-built packages and pure-Python / wasm32 wheels via micropip. Native GPU stacks and many C-extension packages need a local venv or Colab.
Is SQL a real Postgres server?
No. SQL Studio is DuckDB in the browser over local files. Warehouse-scale or remote Postgres/MySQL stays on infrastructure you control.
Is R the same as RStudio?
No. WebR runs R in Wasm with prebuilt packages from repo.r-wasm.org. It cannot compile arbitrary CRAN source packages in the tab.
Can Coding Studio replace CI tests?
Use it for interactive debugging only. Commit scripts to git and run pytest or jest in CI for reproducible gates.