Skip to main content

Welcome to the ContentsVision Handbook

ContentsVision turns a field specialist's walkthrough into a structured contents inventory, ready to push into an Adjust Square claim. Upload a walkthrough video, a batch of photos, or a voice memo, and minutes later you have a reviewed list of items with rooms, conditions, quantities, ages, brands, and supporting photos.

This handbook is the shared reference for everyone who builds, supports, sells, and runs ContentsVision. It is written so a brand new engineer, a support agent on their first call, a project manager scoping a feature, and an operator doing a deploy can each find what they need without reading the code.

New here? Read these three pages first
  1. System overview for the big picture in plain language.
  2. Core concepts so the vocabulary clicks.
  3. The user journey to see it end to end.

Pick your track

What ContentsVision does, in one diagram

A specialist captures a property loss. ContentsVision extracts the items with AI, a human reviews and approves them, and the finished inventory is submitted to the Adjust Square claims platform.

How this handbook is organized

SectionWhat is insideBest for
Get orientedThe plain-language overview, the vocabulary, the user journeyEveryone
ArchitectureSystem design, the AI pipeline, data model, API, frontendDevelopers
SubsystemsAuth and SSO, Adjust Square, billing, automation, observability, analyticsDevelopers, PMs
OperationsLocal dev, deployment, configuration, testing, troubleshootingOperations, developers
ReferenceGlossary and engineering conventionsEveryone

The product in five sentences

  1. A user creates a claim, adds rooms to it, and uploads media (video, photos, or audio) into each room.
  2. Each upload becomes a job that runs through an AI pipeline which extracts inventory items.
  3. The user reviews the items (manually or with Guided Review), edits them, and approves the good ones.
  4. Auto-approval rules and custom AI prompts let a team tune the output to its needs.
  5. The user submits the claim to Adjust Square, where the items and their photos land on the real claim.
A note on accuracy

This handbook is generated from the codebase as it stands today and links to real files (for example backend/app/services/pipeline.py). Where the running code and an older README disagree, the handbook follows the code. The most notable example: transcription is handled by Deepgram Nova-2, not Whisper, even though some older docs still mention Whisper. See the AI pipeline.