Technical Documentation

Detailed system diagrams covering session lifecycle, LLM routing, RAG pipeline, authentication, and database schema.

Session Lifecycle

End-to-end flow from starting an interview to storing the distilled knowledge.

LLM Provider Routing

Priority-based resolution of which AI provider to use for each session.

RAG Pipeline

How documents and past transcripts are ingested, embedded, and retrieved during interviews.

Authentication Flow

How user identity flows from Keycloak through oauth2-proxy to the application.

Database Schema

Core PostgreSQL tables and their relationships.

S3 Storage Layout

STACKIT S3 Bucket
├── transcripts/
│   └── {email}/
│       └── {session_id}.json          # Distilled transcript + summary
├── recordings/
│   └── {room_name}_{timestamp}.mp4    # Full session recording (egress)
├── media/
│   └── {session_id}/
│       └── snapshots/                 # Screen/camera captures
│           └── {timestamp}_{source}_{type}.jpg
└── questionnaires/
    └── {session_id}.json              # Post-session survey responses