AI PDF Knowledge Assistant
Built a local RAG-based document intelligence platform with Spring Boot 4, Spring AI, Ollama, Qdrant, and Java 21, implementing chunked 768-dimensional embeddings and vector similarity search for citation-aware semantic search over private PDF documents. Developed streaming AI chat with SSE and Reactor Flux and LLM tool calling for calculator, filesystem, and weather tasks, containerizing Qdrant with Docker Compose and adding path traversal protection.
Tech Stack
Spring Boot 4Spring AIOllamaQdrant Vector DBJava 21RedisDocker Compose
Key Features
- Local RAG-based Document Intelligence: Chunked 768-dimensional embeddings and vector similarity search for citation-aware semantic search over private PDFs.
- Streaming AI Chat: Real-time response streaming using Server-Sent Events (SSE) and Reactor Flux reactive streams.
- LLM Tool Calling: Autonomous agent execution for calculator, filesystem, and weather tasks.
- Containerized Infrastructure: Qdrant vector database containerized with Docker Compose with path traversal protection.
Architecture & System Design