Auction Sheet Decipher: Full Technical Guide

The Auction Sheet Decipher sends the auction sheet image to Claude (Anthropic’s frontier AI model) along with a structured prompt that instructs it to extract and translate all inspection data. This guide explains the technical process and how to get the most from it.

The decipher pipeline

1. The auction sheet URL (from R2 storage or allowed auction house domains) is fetched server-side
2. The image is base64-encoded and sent to Claude with a structured extraction prompt
3. Claude returns a JSON object with all detected grades, damage codes, and translated comments
4. The result is displayed in the VDP auction sheet section
5. One quota credit is consumed only if the Anthropic API call succeeds

Interpreting the grades

Overall grade reflects the inspector’s composite assessment. Grade 5 is near-new. Grade 4 is good condition. Grade 3.5 is above average. R and RA indicate repaired vehicles where prior accident or structural damage has been professionally addressed.

Exterior grade uses letters A through D. Grade A is clean. Grade B has minor marks. Grade C has visible wear. Grade D has significant interior damage.

Damage indicators are translated damage codes with their panel locations. Each code is expanded to plain language: “A” becomes “Surface scratch”, “U” becomes “Dent”, “X” becomes “Requires panel replacement”.

Language selection

The decipher translates into your preferred language as set in the Triple-Picker. If you switch language and re-decipher, a new credit is consumed. We recommend setting your preferred language before running deciphers.

When the AI gets it wrong

Claude is highly accurate on clear, well-scanned auction sheets but may misread damaged, faded, or poorly photographed documents. If a result seems wrong: check the original image quality, look for handwritten additions that may be outside the standard form structure, and note that regional auction house variants (HAA, CAA) use slightly different form layouts that the AI handles with lower confidence.

Quota rollback on API failure

If the Anthropic API returns an error after the quota has been incremented, WATTSHIP automatically rolls back the credit. You will not lose a decipher credit due to a server error on our side.