Zero‑Retention Architecture Proposal

Core Goal

Build a privacy‑first document assistant where:

Documents are processed but not stored on servers
All persistent data lives in the user's browser
The backend is stateless
Teams can build features independently

To accomplish this we use:

Frontend Storage → EntityDB (IndexedDB underneath)
Backend → Stateless APIs

What is EntityDB?

EntityDB is a client‑side entity graph database built on IndexedDB.

Instead of storing one giant JSON document, we store entities and relationships.

Think of it like this:

SQL	EntityDB
Table	Entity Type
Row	Entity
Foreign Key	Relationship
Join	Graph traversal

Documentation:

https://github.com/babycommando/entity-db

High Level Architecture

User Browser
│
├── EntityDB (IndexedDB)
│
│   Entities:
│   - Document
│   - OCRBlock
│   - FieldCandidate
│   - Chunk
│   - Embedding
│   - Summary
│   - ChatSession
│   - ChatMessage
│   - RiskFlag
│   - Language
│
│
└── API Requests
     │
     ▼

Stateless Backend
- OCR
- LLM summarization
- Embedding generation
- Translation
- Risk classification