Family Archive

A private agentic archive that reads multilingual source documents and proposes typed, sourced changes for people to approve.

A private archive for one family's history, built so a large and incomplete record stays understandable without ever becoming public. The archivist is an agent, not a chat window. It reads multilingual documents, calls typed tools to propose people, relationships and stories, and reconciles each proposal against the existing graph before anything changes. Compatible evidence enriches an existing record instead of creating a duplicate. Ambiguous names and contradictory dates fail closed into questions that carry their source, and every accepted operation enters an audit history that can be read and reversed.

My contribution

I designed and built it end to end: archive reconstruction, the product model, the graph schema, the privacy and access system, visual and interaction design, full stack implementation, the archivist itself, data migration, testing, Cloudflare infrastructure and ongoing operation. The family remains the authority for its own history and contributes the records, corrections, stories and photographs the product preserves.

AI-native workflow

Coding agents worked across source reconstruction, graph and data modeling, privacy UX, full stack and Cloudflare implementation, migration, review, testing and deployment. Product judgment, privacy decisions, source acceptance, verification and release stayed mine. The deployment is private by design and stops at a password gate. A generic version of the product is published separately as an open template with synthetic data.