loupe

Turn any CSV or JSON into a beautiful, interactive HTML dashboard. One command. Zero dependencies.

Why

You have a CSV. You want to explore it — sort, filter, search, see some stats. Your options:

• Excel / Google Sheets — import, wait, click around, can't share without an account
• Pandas + Jupyter — pip install, open notebook, write df.head(), realize you need more code
• Observable / Datasette — serious tools for serious setups. You just want to look at data.

loupe takes any CSV or JSON and spits out a single .html file. Open it in any browser, offline, forever. Sort columns, search across everything, get stats on numeric columns automatically.

The output file has no external dependencies — no CDN calls, no server, no JS framework. It's just a well-crafted HTML file.

Install

pip install loupe-cli

Or just grab the file directly — it's a single Python script with zero dependencies beyond the standard library:

curl -O https://raw.githubusercontent.com/Jah-yee/loupe/main/loupe.py

Usage

# CSV file
python loupe.py sales.csv -o sales.html

# JSON file
python loupe.py users.json -o users.html

# Pipe from stdin
cat data.csv | python loupe.py - -o dashboard.html

# Force format (if extension doesn't match)
python loupe.py export.txt --format csv -o out.html

Then open the output file in your browser.

What you get

Table

• Sort by any column (click headers)
• Full-text search across all columns at once
• Pagination (50 rows per page)
• Inline mini-bar charts for numeric columns — see distribution at a glance

Stats cards

• Auto-detected for numeric columns: avg, min, max, distribution bar
• Up to 5 cards shown by default

Column type detection

• number — integers, floats, currency ($1,234), percentages (42%)
• date — ISO 8601, US, European, and more
• boolean — true/false, yes/no, 1/0 render as ✓ / ✗
• text — everything else

Dark / light mode

• Toggle in the top right corner

Output

• Single .html file — ~10–20kb for most datasets
• Works fully offline
• Embeds all data inline — no external files needed
• Shareable: send the html file directly

Examples

SaaS metrics CSV → dashboard:

python loupe.py saas_metrics.csv -o saas_metrics.html
# ✓ saas_metrics.html (16kb · 30 rows · 9 columns)

npm package stats JSON → dashboard:

python loupe.py npm_packages.json -o npm_packages.html
# ✓ npm_packages.html (13kb · 20 rows · 7 columns)

JSON formats

loupe handles common JSON shapes automatically:

// Plain array (most common)
[{"id": 1, "name": "Alice"}, ...]

// Wrapped in a common key
{"data": [...]}
{"rows": [...]}
{"results": [...]}
{"items": [...]}

How It Works

1. Parse CSV with csv.DictReader or JSON with json.loads — pure stdlib, no pandas
2. Auto-detect each column's type by scanning values
3. Compute stats (min/max/mean) for numeric columns
4. Inject data as a const DATA = [...] block inside a self-contained HTML template
5. The JS (< 100 lines, no framework) handles sort, filter, search, pagination at runtime

The template — CSS variables for theming, sticky headers, monospace numbers, inline bar charts — is all hand-crafted, no build step.

Supported input

Everything that outputs CSV or JSON — databases, APIs, spreadsheets, BI tools, logs, jq, psql -c "\copy ...", gh api, you name it.

# GitHub repo stats via gh CLI
gh api /repos/torvalds/linux/contributors | python loupe.py - -o contributors.html

# PostgreSQL query result
psql -c "COPY (SELECT * FROM orders) TO STDOUT WITH CSV HEADER" | python loupe.py - -o orders.html

# AWS Cost Explorer export
python loupe.py aws-cost-export.csv -o costs.html

Contributing

Issues and PRs welcome. Some ideas on the roadmap:

• Column hiding / reordering
• CSV export of filtered results
• Configurable page size (--page-size 100)
• Chart view for numeric columns (bar / scatter)
• Custom column type overrides (--types col:number,date_col:date)
• --title flag to set dashboard title

License

MIT

_{One command. One file. Your data, explorable.}