All builds

In-browser AI

classify

Write a playbook of categories and rules, drop in a spreadsheet, and get every row decided. Unsure rows queue for review, and files never leave the tab.

Replay of classify on a support-ticket spreadsheet: the file is dropped in, each of seven rows gets a category, a decision and who decided it, one unsure row at 47% confidence asks a person, and the same file comes back with the new columns.

Write a playbook of categories, decisions and rules, then drop in Excel, CSV, JSON or logs. It fills in each row, queues the unsure ones for you, and hands back the same file with new columns. Everything runs in the tab.

What it does

  1. Rules always win. Then a word-based model, then embedding similarity, then an optional second opinion from a larger model on unsure rows.

  2. A review queue with a reason for every row, and your corrections become training examples.

  3. Excel comes back as the same workbook, with other sheets, formatting and formulas kept.

How it is built

  • Measured on held-out lines: categories 93% of 29, ticket decisions 87% of 45.
  • Mixing weights were tuned on separate lines before the test set was run.
  • Models live in the browser's private file system and projects in IndexedDB.

Honest limits

  • The test sets are tiny and hand-labelled.
  • The larger model is 847 MB and needs WebGPU.