CSV Studio — DuckDB Edition

A compact, browser-based CSV workbench powered by DuckDB + Streamlit.
Upload a file or paste a CSV URL, filter + chart it, edit rows, and export—fast.

Live demo · Source

---

Features

• Load data
  • CSV / TSV / XLSX upload or fetch a public CSV by URL
  • “Load sample” button for instant play
• Filter & search
  • Global text search across all string columns
  • Category pickers + multi-select values
  • Metric range slider (int/float/boolean-safe, constant/NaN-safe)
  • Optional date column for time series
• Charts & KPIs
  • Sum & median KPIs
  • Time series (date + metric) and By category (metric grouped by dimension)
• Table / Edit (CRUD)
  • Fast, paginated editor with sticky headers
  • Add / update / delete rows (stable internal _id)
  • Quick filter input above the table
• Export
  • Page CSV, filtered CSV, or full dataset (Parquet)
• Ergonomics
  • Noir+Gold premium dark theme (w/ Light toggle)
  • Share this view: saves filters to URL params for easy linking
  • Clear data & Clear filters one-click actions

---

Quickstart

# 1) Create & activate a venv (Windows PowerShell shown)
python -m venv .venv
.\.venv\Scripts\Activate.ps1

# 2) Install dependencies
pip install -r requirements.txt  # streamlit, duckdb, pandas, plotly, openpyxl

# 3) Run the app
python -m streamlit run app.py

---

How to use

Load data from the sidebar: upload CSV/TSV/XLSX, paste a CSV URL, or press Load sample.

Use Global search and Category / Metric filters to shape your view.

Switch between Charts and Table / Edit tabs.

Table tab has a Quick filter box (any column).

Use Prev / Next to paginate large data.

Click Apply changes to persist edits back to the in-memory dataset.

Export page CSV, filtered CSV, or the full dataset as Parquet.

Press 🔗 Share this view to push filters into the URL and share the exact state.

---

Why DuckDB?

DuckDB is an in-process analytical database—perfect for ad-hoc querying without maintaining a server. Query performance stays snappy in the browser app via Python bindings and in-memory tables.

---

Good test CSV URLs

Chicago community areas (small):
https://raw.githubusercontent.com/datadesk/census-data-1/master/data/education.csv

California cities (medium):
https://raw.githubusercontent.com/plotly/datasets/master/2014_usa_states.csv

NYC 311 sample (larger):
https://raw.githubusercontent.com/vaibhavk97/NYC-311-Data-Analysis/master/311_Service_Requests_from_2010_to_Present.csv

Any public raw CSV link works (GitHub “Raw” URLs are great). Very large files depend on your machine/host memory.

---

License

MIT © 2025 — Feel free to fork, adapt, and ship.