TutorBot — AI tutor website

TutorBot is an AI tutor that helps learners interactively study course materials with LLM assistance. It is more than a generic chat wrapper: TutorBot implements a full retrieval-augmented generation (RAG) pipeline against the user's course documents so answers are grounded in those materials.

Tech stack and rationale

• Java + Spring Boot — mature ecosystem, strong typing, and production-ready frameworks. Project uses idiomatic Spring: controllers, service layer, and configuration via TutorBotApplication.
• Spring Data JPA (Hibernate) — transactional, relational mapping for users, roles, and metadata where ACID guarantees are important.
• Spring Security — centralized authentication and authorization with a custom UserDetailsService and clear separation between credential concerns and business logic.
• PostgreSQL + pgvector — user/auth data lives in Postgres; embeddings are stored with pgvector when keeping semantic vectors colocated with relational data is desirable (simpler infra, transactional consistency).
• MongoDB — flexible document storage for extracted text and chunked documents; well-suited for variable-length text segments and more complex Json data structures (Quizzes).
• langchain4j RAG pipeline — the AI flow follows: ingest → chunk → embed → index → retrieve (metadata filtering) → generate. The orchestration is decoupled from storage so vector backends can be swapped.
• Docker & Docker Compose — container-first packaging and multi-stage builds via backend/compose.yaml for reproducible deployments and environment-driven configuration.

Concise architecture notes

• Code lives under backend/src/main/java/ch/frupp/lecturevault/ (packages: auth, ai, course, user).
• The ai package separates ingestion (PDF parsing, chunking), embedding persistence, and query-time retrieval + prompt orchestration.
• Persistence strategy mixes Postgres (structured user/auth data, embeddings via pgvector) and MongoDB (document chunks and provenance) to balance transactional integrity with schema flexibility.

Files of interest

• backend/src/main/java/ch/frupp/lecturevault/ — main backend code (controllers, services, repositories)
• backend/compose.yaml — container orchestration example (local dev environment, no production deployment yet)
• backend/src/test/java — unit and integration tests demonstrating testability and mocking of AI providers