Updated main.py

README.md

@@ -1,42 +1,47 @@
-# 📚 Chat + RAG PDF Assistant (Streamlit)
+---
+title: RagBot
+emoji: 📚
+colorFrom: indigo
+colorTo: blue
+sdk: streamlit
+sdk_version: "1.32.0"
+app_file: app.py
+pinned: false
+---
 
-This is a Streamlit-based chatbot that allows you to upload a PDF and ask questions about its content, while also answering general knowledge questions using an open-source language model.
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
-## 🚀 Features
+---
 
-- Upload a PDF and extract its content
-- Chatbot interface with persistent chat history
-- Document-specific Q&A using Retrieval-Augmented Generation (RAG)
-- General knowledge support (open-domain QA)
-- Anti-hallucination strategies
-- Runs fully locally using Hugging Face models
+# 📚 RagBot: Chatbot + Document QA
 
-## 📦 Tech Stack
+This app lets you chat with any uploaded PDF and ask questions — either from the document or general ones — powered by **local open-source models**. Combines RAG with chatbot UX.
 
-- `Streamlit` - For the frontend and interactive chat UI
-- `PyMuPDF` - To extract text from PDF files
-- `sentence-transformers` - For vector embeddings
-- `FAISS` - For fast similarity search
-- `transformers` - Lightweight Hugging Face inference with distilBERT
-- `Python` - Core application logic
+## 🚀 Features
 
-## 🛠 How it Works
+- Upload PDFs and chat with them
+- General knowledge + document-specific answers
+- Accurate retrieval using FAISS and sentence-transformers
+- Clean, sticky UI with full chat history
+- Runs 100% free using lightweight Hugging Face models
 
-1. Upload your PDF
-2. PDF is chunked and embedded
-3. FAISS builds an index of these chunks
-4. Ask any question:
-   - For general questions: uses general knowledge
-   - For doc-specific: retrieves relevant context and answers accordingly
+## 🧠 Powered By
 
-## 🔒 Anti-Hallucination
+- 🧠 Embeddings: `sentence-transformers/paraphrase-MiniLM-L6-v2`
+- 📖 QA Model: `distilbert-base-cased-distilled-squad`
+- 🧰 Backend: Streamlit + FAISS + Transformers
 
-- Clear system prompt restricts out-of-scope answers
-- Uses `do_sample=False` and `temperature=0.0` for consistency
-- Returns "I’m sorry, I don’t know..." when context is missing
+## 💡 How It Works
+
+1. Upload your PDF
+2. App chunks & embeds the text
+3. Uses FAISS to retrieve context
+4. Hugging Face model generates answers:
+   - Uses the doc context if available
+   - Falls back to general QA
+   - Refuses to answer if uncertain
 
-## 📥 Installation
+## 📦 Dependencies
 
 ```bash
 pip install streamlit faiss-cpu sentence-transformers transformers PyMuPDF
-streamlit run app.py