HushQuote: quotes that stay yours

Description and use

HushQuote enables rapid, accurate quote generation from spoken job descriptions, entirely in the browser, with no servers or accounts.

Problem

Tradespeople struggle to quickly generate accurate, editable quotes from spoken job descriptions, often leading to errors or delays.

Outcome

HushQuote enables rapid, accurate quote generation entirely in the browser, eliminating server costs, preserving privacy, and reducing quoting time for tradespeople.

Workflow

• 1. User speaks or types a job description into the command bar.
• 2. The Gemma 3 270M model runs in the browser (WebGPU or WASM) and outputs JSON via a schema‑enforced response format.
• 3. The app validates the JSON with Zod; if validation fails, the model retries once with the error attached.
• 4. Validated line items are added to the sheet with provisional styling.
• 5. The user confirms or edits each line; totals are computed in JavaScript from confirmed items.
• 6. The final quote can be exported as a PDF.
• 7. All data remains on the local device; no server or account is required.

Evidence

The whole app is one living document, not a wizard: The Margin, The Sheet, The Command Bar

Source

A 270-million-parameter model is small enough to ship as a file instead of a service, but it is also small enough to be wrong in specific, repeatable ways

Source

If the first response fails validation, the model gets one more try with the exact error attached. If that fails too, the app does not guess - it drops back to manual mode

Source

Model invented a fully priced fake quote for a job that explicitly had no numbers in it

Source