Abstract

This project is a simple Retrieval-Augmented Generation (RAG) system that can load documents, index them, and answer user questions based on their content. It supports multiple document formats, generates embeddings, stores them in a vector database, and retrieves the most relevant context for each query. The assistant then uses an LLM to produce accurate, context-aware answers.

Introduction

This is my first core project in the certification, where I applied key concepts from Module 1 including:

• Document ingestion and embedding
• Vector database retrieval (Chroma)
• LLM-based response generation

What this project consists of

We will implement a complete RAG system that can:

• Load and chunk documents from the data/ directory
• Create embeddings and store them in a vector database
• Search for relevant context based on user queries
• Generate responses using retrieved context and an LLM

Tools & Frameworks

• Python
• LangChain – for chaining prompt → retriever → LLM
• ChromaDB – vector store for semantic search
• OpenAI Embeddings – document encoding

How It Works

This project is subdivided in 7 main steps:

1. Prepare Your Documents - Adding our own documents to the data directory
2. Document Loading - Loading documents from files into the system
3. Text Chunking - Spliting documents into smaller, searchable chunks
4. Document Ingestion - Processing and storing documents in the vector database
5. Similarity Search - Finding relevant documents based on queries
6. RAG Prompt Template - Designing effective prompts for the LLM
7. RAG Query Pipeline - Completing query-response pipeline using retrieved context

Results and conclusion

By completing this project, we are able now:

• ✅ Understand RAG architecture and data flow
• ✅ Implement text chunking strategies
• ✅ Work with vector databases and embeddings
• ✅ Build LLM-powered applications with LangChain
• ✅ Handle multiple API providers
• ✅ Create production-ready AI applications

Our implementation is complete since:

1. ✅ We can load your own documents
2. ✅ The system chunks and embeds documents
3. ✅ Search returns relevant results
4. ✅ The RAG system generates contextual answers
5. ✅ We can ask questions and get meaningful responses