How krey-yon/cliven works
Project type and purpose
• This is a command-line tool (Cliven) implementing a Retrieval-Augmented Generation (RAG) pipeline for chatting with PDF documents using fully local services.
• It orchestrates PDF extraction, text chunking, embedding, vector storage (ChromaDB), and LLM inference (Ollama + Gemma models).
File structure
• Top-level:
• main/ – core CLI application modules
• utils/ – helper modules for chunking, parsing, embedding, and vector DB operations
• docker-compose.yml – orchestration of ChromaDB and Ollama services
• setup.py, requirements.txt – packaging and dependencies
• Key patterns:
• Layered design: CLI layer (main/), processing layer (utils/), persistent/external services (Docker).
• RAG pipeline flow from PDF → chunks → embeddings → vector store → query engine → LLM.
README for additional insights
• Describes a five-stage architecture: PDF Parser, Text Chunker, Embedder, Vector DB, Chat Engine.
• Enumerates external services: ChromaDB and Ollama (with lightweight/high-performance models).
• Highlights configuration via environment variables and Docker Compose.