Table to JSON + Diff

Turn HTML tables on pages you own or are allowed to read into clean JSON records, and get what changed since last time (added / removed / changed rows) instead of re-reading the whole table.

Typical uses: your own price list or inventory page, a partner's public rate table you are permitted to monitor, a public timetable, a league table, a government open-data page, a status board. Feed the output to a spreadsheet, a Slack / Discord notification, or an LLM.

What you get

One dataset item per table:

{
  "url": "https://example.com/prices",
  "table_index": 0,
  "caption": "Price list",
  "headers": ["Item", "Price", "Stock"],
  "row_count": 42,
  "rows": [{"Item": "A-100", "Price": "1,200", "Stock": "in stock"}, ...],
  "diff": {
    "first_run": false,
    "previous_run_at": "2026-11-03T09:00:00+00:00",
    "counts": {"added": 1, "removed": 0, "changed": 3, "previous_rows": 41, "current_rows": 42},
    "added": [...], "removed": [...],
    "changed": [{"key": "A-100", "fields": {"Price": {"before": "1,100", "after": "1,200"}}, "after": {...}}]
  },
  "run_at": "2026-11-04T09:00:00+00:00"
}

• Header row becomes the keys (duplicate headers get _2, _3; empty headers become col_n).
• colspan is flattened; thead / tbody / th-less tables are handled.
• Give a Key column (e.g. ID) to get field-level changes. Without it, rows are matched by full content (added / removed only).
• Snapshots live in a named key-value store in your account (snapshotStore), so you can run this on a schedule and only act on diff.counts.

Input

Field	Meaning
urls	Pages to read (required)
tableSelector	CSS selector, default table
headerRow	First row is the header (default on)
keyColumn	Column that identifies a row, enables changed detection
diffMode	Compare with previous run (default on)
includeRowsInDiff	Off = counts only
snapshotStore	Name of the key-value store that keeps snapshots
maxTablesPerPage	Safety cap (default 20)

Pricing (pay per event)

You are charged only when a table is actually delivered: one table-extracted event per table, plus one diff-computed event per table that had a previous snapshot. Failed requests and pages with no matching table are free.

Fair use

Plain HTTP fetch, one request per URL, no login, no anti-bot evasion, no JavaScript rendering. Use it on pages you own or have permission to read, and respect the site's terms and robots rules. For JavaScript-rendered tables, pre-render the page with a browser Actor and feed this Actor the HTML URL.

Scheduling

Create a Task with your input, add a Schedule (e.g. every morning), and connect a webhook or an integration (Slack, Google Sheets, Make, Zapier) that fires when diff.counts.added + changed + removed > 0.