Rusted Recall

Inspiration

Modern AI-generated media can become outdated or legally incorrect the moment a product label, marketing claim, or approved source changes. Updating every generated asset manually is expensive, slow, and difficult to verify.

I built Rusted Recall to solve this problem by bringing software-style change management to AI media. Instead of searching manually for outdated content, organizations can declare a source-of-truth change and automatically identify, review, repair, and verify every affected asset.

What it does

Rusted Recall is an AI-powered recall management platform for generated media.

The platform allows organizations to:

Register approved Source-of-Truth items.
Version approved claims and media.
Create a Recall whenever an approved source changes.
Automatically analyze which assets are affected.
Build an explainable dependency graph.
Present affected assets for human review.
Repair assets using deterministic or generative workflows.
Produce before/after comparisons.
Preserve immutable technical evidence.
Store repair artifacts and reports on Backblaze B2.

The result is an end-to-end workflow that transforms source changes into verified media updates with complete traceability.

How i built it

The application is built as a production-style SaaS platform.

Main components include:

React frontend
FastAPI backend
PostgreSQL database
Backblaze B2 for immutable evidence and generated artifacts
Genblaze media workflow integration
Interactive impact graph
Review queue
Audit logging
Technical evidence generation
JSON, CSV, HTML and PDF reporting

Challenges i ran into

The biggest challenge was making the system explainable instead of acting like a black box.

Every detected dependency must be traceable, every repair must be verifiable, and every operation must leave an audit trail. We focused on deterministic evidence, immutable storage, and reproducible workflows rather than only generating media.

Accomplishments that i am proud of

End-to-end recall workflow.
Source-of-Truth version management.
Automatic impact analysis.
Interactive dependency visualization.
Human review pipeline.
Automated repair workflow.
Before/After gallery.
Immutable evidence stored on Backblaze B2.
Complete audit history.

https://youtu.be/TrYPHfgxLWs

https://rusted-recall.onrender.com/

What i learned

Building trustworthy AI systems requires more than generation quality. Organizations need governance, traceability, explainability, and evidence that every generated asset can be audited and repaired after approved information changes.