CSVSnap

Lightweight CSV dashboard. Python stdlib only — no npm, no build, no dependencies.

Install

chmod +x csnap.py
ln -sf "$PWD/csnap.py" ~/.local/bin/csnap     # any dir on your PATH

Use

csnap .                       # serve CSVs in the current directory
csnap dir/dir/dir             # or any other directory
csnap data --port 8080

Prints the URL; open it. Default host is 127.0.0.1, default port 8000.

Tabs

Tables — lists every *.csv in the directory as a card (name, column count, file size). The search box filters by table name or column name; when a column matches, the card says which. Click a card to load it. Rows are fetched 100 at a time from the server, so the list view never reads row data.

Relations — draws each table as a box and connects columns using relation.json. Drag a table header to move it. Relations pointing at an unknown table or column are skipped and reported at the top.

relation.json

Lives next to the CSVs. Table name = filename without .csv, column names come from the header row.

{
  "relations": [
    { "from": "orders.customer_id", "to": "customers.id", "type": "many-to-one" }
  ]
}

type is a free-text label drawn on the edge; omit it if you don't want one.

API

Endpoint	Returns
GET /api/tables	[{name, bytes, columns}] — header row only
GET /api/table?name=X&offset=0&limit=100	{columns, rows, total}
GET /api/relations	contents of relation.json

Tests

python3 test_csnap.py

Notes

• Binds to 127.0.0.1 by default. --host 0.0.0.0 exposes the served directory to your network.
• Table names are resolved inside the served directory only; ../ is rejected.
• Row counts require a full file scan. Fine up to ~100 MB; add an index if you go bigger.
• data/ holds sample CSVs — csnap data to see it working.