Structuring Incoming Emails for Valuable Extraction

---

SIEVE is a web application that structures and analyzes incoming emails with LLM support and enriches results with internal knowledge and product data. Emails can be analyzed manually or imported from IMAP inboxes, and all completed runs are available in the History view with source-aware filtering.

This repository contains:

• apps/frontend (Next.js UI)
• apps/backend (NestJS API + database orchestration)
• apps/ai-backend (FastAPI + LangGraph analysis flows)

Getting Started

Run the full stack with Docker Compose:

git clone git@github.com:SE-UUlm/sieve.git
cd sieve

cp .env.example .env
# Generate a key and add it as SETTINGS_ENCRYPTION_KEY in .env
openssl rand -base64 32

docker compose up

If you build images locally with docker compose up --build, generate clients first:

cd apps/backend
pnpm install
pnpm run generate:clients

After startup, open:

• Frontend: http://localhost:3000
• Backend API: http://localhost:5175/api
• Backend docs: http://localhost:5175/docs
• AI-Backend docs: http://localhost:8000/docs

For detailed setup, configuration, architecture, and contribution workflow, use the wiki:

• Home
• Getting Started
• Configuration
• Architecture
• Contributing