1. Overview
   This project implements a Retrieval-Augmented Generation (RAG) chatbot that allows users to ask natural language questions about the content of PDF documents.
   It runs entirely locally using:
   • Ollama (for embeddings + LLM responses)
   • ChromaDB (vector database for document retrieval)
   • Streamlit (UI to interact with the chatbot)
   The chatbot reads PDFs, converts them into vector embeddings, stores them in a vector database, and uses the embeddings to provide context-aware answers.

---

2. Architecture
   Components
3. PDF Loader
   o Reads PDF files from a folder (project_docs).
   o Extracts raw text using PyPDF.
4. Embedding Generator
   o Uses Ollama’s nomic-embed-text model to convert text into dense vector embeddings.
   o Each embedding represents the semantic meaning of a document.
5. Vector Database (ChromaDB)
   o Stores embeddings and corresponding document text.
   o Supports similarity search for retrieving relevant chunks when a question is asked.
   o Persistent mode ensures the database survives app restarts (chroma_db/ folder).
6. Query Processor
   o When a user asks a question:
    Converts the query into an embedding (via Ollama).
    Retrieves top n most relevant documents from ChromaDB.
    Passes these documents + question as a prompt to Ollama’s LLM (llama3.1:8b).
7. LLM Answer Generator
   o Ollama LLM uses the retrieved context to generate a natural-language answer.
   o Helps the model remain factual and grounded in the provided PDFs.
8. User Interface
   o Built with Streamlit.
   o Features:
    Displays status of loaded PDFs.
    Input field for user questions.
    Full chat history (Q&A between You and Bot).
    Clean, icon-free design for readability.

---

3. Workflow
   Step 1: Load Documents
   • PDFs placed in project_docs/.
   • Extracted text → embedded with Ollama → stored in ChromaDB.
   Step 2: User Asks Question
   • Input collected via Streamlit text box.
   • Converted to embedding using Ollama.
   • Similar documents retrieved from ChromaDB.
   Step 3: Generate Answer
   • Retrieved context + question passed to Ollama LLM.
   • LLM generates context-aware answer.
   • Answer displayed in UI alongside user’s query.
   Step 4: Maintain Conversation
   • All interactions stored in st.session_state.chat_history.
   • Chat history is displayed below new inputs.

---

4. Tech Stack
   Component Tool / Library
   LLM Ollama (llama3.1:8b)
   Embeddings Ollama (nomic-embed-text)
   Vector Database ChromaDB (PersistentClient)
   Document Loader PyPDF
   UI Streamlit
   Language Python 3.13