Corex Codes

.gitignore

@@ -0,0 +1,40 @@
+# Ignore virtual environment
+venv/
+ragenv/
+ENV/
+env/
+.venv/
+
+# Python compiled files
+__pycache__/
+*.py[cod]
+*.so
+
+# Environment variables
+.env
+
+# VS Code settings
+.vscode/
+*.code-workspace
+
+# OS-specific
+.DS_Store
+Thumbs.db
+
+# Logs and databases (optional)
+*.log
+*.sqlite3
+
+# Jupyter/IPython
+.ipynb_checkpoints/
+
+# Cache
+*.cache
+*.pkl
+*.db
+
+# Node modules (if ever added)
+node_modules/
+
+Kubernetes/secret.yml
+

Dockerfile

@@ -0,0 +1,37 @@
+# Use an official Python runtime as a parent image
+FROM python:3.10-slim
+
+# Set the working directory in the container
+WORKDIR /app
+
+# Create a non-root user and set cache directory permissions
+RUN useradd --create-home --shell /bin/bash app && \
+    mkdir -p /home/app/.cache && \
+    chown -R app:app /home/app/.cache && \
+    chown -R app:app /app
+
+# Set environment variables for Hugging Face cache
+ENV HF_HOME=/home/app/.cache/huggingface
+ENV TRANSFORMERS_CACHE=/home/app/.cache/huggingface/transformers
+ENV HF_DATASETS_CACHE=/home/app/.cache/huggingface/datasets
+
+# Copy the requirements file into the container
+COPY requirements.txt .
+
+# Install dependencies
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the rest of the application code into the container
+COPY . .
+
+# Change ownership of all files to app user
+RUN chown -R app:app /app
+
+# Switch to non-root user
+USER app
+
+# Expose the port your app runs on
+EXPOSE 8000
+
+# Command to run the application; Hugging Face Spaces sets PORT env
+CMD sh -c "uvicorn main:app --host 0.0.0.0 --port ${PORT:-8000}"

Kubernetes/deployment.yml

@@ -0,0 +1,30 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: rag-app
+  namespace: rag
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: rag-app
+  template:
+    metadata:
+      labels:
+        app: rag-app
+    spec:
+      containers:
+      - name: rag-container
+        image: yadavkapil23/rag-app:latest
+        ports:
+        - containerPort: 8000
+        # --- NEW CODE: INJECT HUGGINGFACE TOKEN FROM A SECRET ---
+        env:
+        - name: HUGGINGFACE_API_TOKEN
+          valueFrom:
+            secretKeyRef:
+              # You must create a secret named 'huggingface-secret' beforehand
+              name: huggingface-secret
+              # Assuming the key inside the secret is also named HUGGINGFACE_API_TOKEN
+              key: HUGGINGFACE_API_TOKEN
+        # --------------------------------------------------------

Kubernetes/namespace.yml

@@ -0,0 +1,4 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: rag

Kubernetes/service.yml

@@ -0,0 +1,13 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: rag-service
+  namespace: rag
+spec:
+  type: NodePort
+  selector:
+    app: rag-app
+  ports:
+  - port: 8000
+    targetPort: 8000
+    nodePort: 30036  # optional fixed port, else Kubernetes assigns a random one

Procfile

@@ -0,0 +1 @@
+web: uvicorn main:app --host=0.0.0.0 --port=8000

README.md

@@ -0,0 +1,129 @@
+---
+title: RAG Project
+emoji: 🧠
+colorFrom: blue
+colorTo: purple
+sdk: docker
+app_port: 8000
+python_version: 3.10
+---
+
+# 🚀 RAG System with LangChain and FastAPI 🌐
+
+Welcome to this repository! This project demonstrates how to build a powerful RAG system using **LangChain** and **FastAPI** for generating contextually relevant and accurate responses by integrating external data into the generative process.
+
+## 📋 Project Overview
+
+The RAG system combines retrieval and generation to provide smarter AI-driven responses. Using **LangChain** for document handling and embeddings, and **FastAPI** for deploying a fast, scalable API, this project includes:
+
+- 🗂️ **Document Loading**: Load data from various sources (text, PDFs, etc.).
+- ✂️ **Text Splitting**: Break large documents into manageable chunks.
+- 🧠 **Embeddings**: Generate vector embeddings for efficient search and retrieval.
+- 🔍 **Vector Stores**: Store embeddings in a vector store for fast similarity searches.
+- 🔧 **Retrieval**: Retrieve the most relevant document chunks based on user queries.
+- 💬 **Generative Response**: Use retrieved data with language models (LLMs) to generate accurate, context-aware answers.
+- 🌐 **FastAPI**: Deploy the RAG system as a scalable API for easy interaction.
+
+## ⚙️ Setup and Installation
+
+### Prerequisites
+
+Make sure you have the following installed:
+- 🐍 Python 3.10+
+- 🐳 Docker (optional, for deployment)
+- 🛠️ PostgreSQL or FAISS (for vector storage)
+
+### Installation Steps
+
+1. **Clone the repository**:
+   ```bash
+   git clone https://github.com/yadavkapil23/RAG_Project.git
+   ```
+
+2. **Set up a virtual environment**:
+   ```bash
+   python -m venv venv
+   source venv/bin/activate   # For Linux/Mac
+   venv\Scripts\activate      # For Windows
+   ```
+
+3. **Install dependencies**:
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+4. **Run the FastAPI server**:
+   ```bash
+   uvicorn main:app --reload
+   ```
+
+   Now, your FastAPI app will be running at `http://127.0.0.1:8000` 🎉!
+
+### Set up Ollama 🦙
+
+This project uses Ollama to run local large language models.
+
+1.  **Install Ollama:** Follow the instructions on the [Ollama website](https://ollama.ai/) to download and install Ollama.
+
+2.  **Pull a model:** Pull a model to use with the application. This project uses `llama3`.
+    ```bash
+    ollama pull llama3
+    ```
+
+
+## 🛠️ Features
+
+- **Retrieval-Augmented Generation**: Combines the best of both worlds—retrieving relevant data and generating insightful responses.
+- **Scalable API**: FastAPI makes it easy to deploy and scale the RAG system.
+- **Document Handling**: Supports multiple document types for loading and processing.
+- **Vector Embeddings**: Efficient search with FAISS or other vector stores.
+
+## 🛡️ Security
+
+- 🔐 **OAuth2 and API Key** authentication support for secure API access.
+- 🔒 **TLS/SSL** for encrypting data in transit.
+- 🛡️ **Data encryption** for sensitive document storage.
+
+## 🚀 Deployment
+
+### Hugging Face Spaces (Docker) Deployment
+This project is configured for a Hugging Face Space using the Docker runtime.
+
+1. Push this repository to GitHub (or connect local).
+2. Create a new Space on Hugging Face → Choose "Docker" SDK.
+3. Point it to this repo. Spaces will build using the `Dockerfile` and run `uvicorn` binding to the provided `PORT`.
+4. Ensure the file `data/sample.pdf` exists (or replace it) to allow FAISS index creation on startup.
+
+Notes:
+- Models `Qwen/Qwen2-0.5B-Instruct` and `all-MiniLM-L6-v2` will be downloaded on first run; initial cold start may take several minutes.
+- Dependencies are CPU-friendly; no GPU is required.
+- If you see OOM, consider reducing `max_new_tokens` in `vector_rag.py` or swapping to an even smaller instruct model.
+
+### Docker Deployment (Local)
+If you want to deploy your RAG system using Docker, simply build the Docker image and run the container:
+
+```bash
+docker build -t rag-system .
+docker run -p 8000:8000 rag-system
+```
+
+### Cloud Deployment
+Deploy your RAG system to the cloud using platforms like **AWS**, **Azure**, or **Google Cloud** with minimal setup.
+
+## 🧠 Future Enhancements
+
+- 🔄 **Real-time Data Integration**: Add real-time data sources for dynamic responses.
+- 🤖 **Advanced Retrieval Techniques**: Implement deep learning-based retrievers for better query understanding.
+- 📊 **Monitoring Tools**: Add monitoring with tools like Prometheus or Grafana for performance insights.
+
+## 🤝 Contributing
+
+Want to contribute? Feel free to fork this repository, submit a pull request, or open an issue. We welcome all contributions! 🛠️
+
+## 📄 License
+
+This project is licensed under the MIT License.
+
+---
+
+🎉 **Thank you for checking out the RAG System with LangChain and FastAPI!** If you have any questions or suggestions, feel free to reach out or open an issue. Let's build something amazing!

data/my_document.txt

@@ -0,0 +1,33 @@
+Knowledge Base
+
+
+Quantum computing uses qubits that can represent both 0 and 1 simultaneously, offering immense parallelism for computation.
+A transformer model uses self-attention to weigh the importance of each word in a sentence for tasks like translation or summarization.
+Python 3.12 introduced new error messages, better performance, and support for isolated subinterpreters.
+
+The French Revolution (1789–1799) radically transformed French society, ending monarchy and spreading ideas of liberty and equality.
+Mahatma Gandhi led the Indian independence movement through nonviolent civil disobedience, notably during the Salt March.
+
+A Random Forest is an ensemble of decision trees used for classification or regression. It reduces overfitting and improves accuracy.
+LangChain is a framework for developing LLM-powered apps with components like chains, tools, memory, and agents.
+
+Meditation helps in reducing stress, enhancing concentration, and improving emotional regulation. Regular practice can reduce anxiety.
+Intermittent fasting involves alternating periods of eating and fasting. It can help in weight loss and metabolic health.
+
+GDP (Gross Domestic Product) measures a country's economic output. A growing GDP usually indicates a healthy economy.
+Inflation refers to the general rise in prices over time, reducing purchasing power. Central banks use interest rates to control inflation.
+
+Photosynthesis is the process where green plants use sunlight, CO₂, and water to produce oxygen and glucose.
+Black holes are regions in space where gravity is so strong that nothing—not even light—can escape.
+
+A binary search tree is a node-based data structure where left children are smaller and right children are larger than the parent node.
+Recursion is a function calling itself until a base condition is met. It’s used in tree traversal, backtracking, and divide-and-conquer.
+
+Japan is an island country in East Asia known for its technology, cherry blossoms, and cultural traditions like tea ceremony and sumo.
+The Eiffel Tower was constructed in 1889 in Paris and is one of the most visited monuments in the world.
+
+Q: What is a black hole?
+A: A black hole is a region in space where gravity is so strong that nothing, not even light, can escape its pull.
+
+Q: How do neural networks work?
+A: Neural networks consist of layers of nodes that process inputs through weighted connections and activation functions to detect patterns.

endpoints.py

@@ -0,0 +1,35 @@
+from fastapi import APIRouter, HTTPException
+from pydantic import BaseModel
+from typing import List, Literal
+
+router = APIRouter()
+
+from rag import get_smart_rag_response
+
+# Pydantic models for request/response validation
+class Message(BaseModel):
+    role: Literal["user", "assistant"]
+    content: str
+
+class QueryRequest(BaseModel):
+    query: str
+    conversation_history: List[Message] = []
+
+class QueryResponse(BaseModel):
+    query: str
+    response: str
+    source: str
+
+@router.post("/query/")
+async def query_rag_system(request: QueryRequest):
+    try:
+        # Convert Pydantic models to dicts for processing
+        history = [msg.dict() for msg in request.conversation_history]
+        response, source = await get_smart_rag_response(request.query, history)
+        return QueryResponse(
+            query=request.query,
+            response=response,
+            source=source
+        )
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))

main.py

@@ -0,0 +1,24 @@
+from fastapi import FastAPI
+from fastapi.staticfiles import StaticFiles
+from fastapi.templating import Jinja2Templates
+from fastapi.requests import Request
+from endpoints import router
+import uvicorn
+
+app = FastAPI()
+
+# Serve static files (CSS, JS)
+app.mount("/static", StaticFiles(directory="static"), name="static")
+
+# Serve HTML templates
+templates = Jinja2Templates(directory="templates")
+
+@app.get("/")
+def home(request: Request):
+    return templates.TemplateResponse("index.html", {"request": request})
+
+# Include your API endpoints
+app.include_router(router)
+
+if __name__ == "__main__":
+    uvicorn.run(app, host="127.0.0.1", port=8000)

rag.py

@@ -0,0 +1,107 @@
+from vector_rag import query_vector_store, llm # <--- FIX: Import llm here!
+import wikipedia
+from typing import List, Dict
+# REMOVED: All duplicate model/pipeline/tokenizer imports and initialization code
+
+# The 'llm' instance is now imported from vector_rag.py and is ready to use.
+wikipedia.set_lang("en")
+
+def format_conversation_context(history: List[Dict], max_messages: int = 10) -> str:
+    """
+    Formats conversation history into a context string for the LLM.
+    Keeps only the most recent messages to prevent token overflow.
+    
+    Args:
+        history: List of message dicts with 'role' and 'content' keys
+        max_messages: Maximum number of messages to include (default: 10)
+    
+    Returns:
+        Formatted conversation history string
+    """
+    if not history:
+        return ""
+    
+    # Keep only the last N messages
+    recent_history = history[-max_messages:]
+    
+    formatted_lines = []
+    for msg in recent_history:
+        role = "User" if msg["role"] == "user" else "Assistant"
+        formatted_lines.append(f"{role}: {msg['content']}")
+    
+    return "\n".join(formatted_lines)
+
+async def get_smart_rag_response(query: str, conversation_history: List[Dict] = None) -> tuple[str, str]:
+    """
+    Get a smart RAG response with conversation context.
+    
+    Args:
+        query: The user's current question
+        conversation_history: List of previous messages (optional)
+    
+    Returns:
+        Tuple of (response, source)
+    """
+    print(" Received Query:", query)
+    
+    if conversation_history is None:
+        conversation_history = []
+    
+    # Format conversation history for context
+    context_str = format_conversation_context(conversation_history)
+
+    # First: Try Wikipedia
+    try:
+        summary = wikipedia.summary(query, sentences=5)
+        print("Wikipedia summary found.")
+        
+        # Build prompt with conversation context
+        prompt = f"""You are a helpful assistant engaged in a conversation.
+"""
+        if context_str:
+            prompt += f"""
+Previous conversation:
+{context_str}
+
+"""
+        prompt += f"""Use the following Wikipedia information to answer the current question as clearly as possible.
+
+Wikipedia Context:
+{summary}
+
+Current question: {query}
+Answer:"""
+        result = llm.invoke(prompt) 
+        answer = result.replace(prompt, "").strip()
+        return answer, "Wikipedia"
+    except wikipedia.exceptions.PageError:
+        print("Wikipedia page not found.")
+    except wikipedia.exceptions.DisambiguationError as e:
+        return f"The query is ambiguous. Did you mean: {', '.join(e.options[:5])}", "Wikipedia"
+
+    # Second: Fallback to LLM with conversation context
+    try:
+        print("Fallback: LLM with conversation context")
+        
+        fallback_prompt = "You are a knowledgeable assistant engaged in a conversation.\n\n"
+        if context_str:
+            fallback_prompt += f"Previous conversation:\n{context_str}\n\n"
+        fallback_prompt += f"Current question: {query}\nAnswer:"
+        
+        llm_answer = llm.invoke(fallback_prompt) 
+        answer = llm_answer.replace(fallback_prompt, "").strip()
+        if answer and "not sure" not in answer.lower():
+            return answer.strip(), "LLM"
+    except Exception as e:
+        print("Error during LLM fallback:", e)
+
+    # Finally: Fallback to Local Documents
+    try:
+        print("Fallback: Local vector search")
+        vector_answer = query_vector_store(query, conversation_history)
+        if vector_answer:
+            return vector_answer, "Local Document"
+    except Exception as e:
+        print("Error during local vector search:", e)
+
+    return "Sorry, I couldn't find any information to answer your question.", "System"

requirements.txt

@@ -0,0 +1,14 @@
+fastapi
+uvicorn
+langchain
+langchain-community 
+python-dotenv
+langchain-huggingface
+faiss-cpu
+jinja2
+wikipedia
+pypdf
+sentence-transformers 
+torch                 
+transformers          
+accelerate

static/script.js

@@ -0,0 +1,371 @@
+document.addEventListener('DOMContentLoaded', () => {
+  const queryInput = document.getElementById('queryInput');
+  const askButton = document.getElementById('askButton');
+  const chatMessages = document.getElementById('chatMessages');
+
+  // 💬 Conversation state management
+  let conversationHistory = [];
+
+  function addUserMessage(content) {
+    conversationHistory.push({
+      role: 'user',
+      content: content,
+      timestamp: Date.now()
+    });
+  }
+
+  function addAssistantMessage(content, source) {
+    conversationHistory.push({
+      role: 'assistant',
+      content: content,
+      timestamp: Date.now(),
+      source: source
+    });
+  }
+
+  function clearConversation() {
+    conversationHistory = [];
+  }
+
+  function getHistory() {
+    return conversationHistory;
+  }
+
+  // 💬 Message rendering functions
+  function renderUserMessage(content, timestamp = null) {
+    const ts = timestamp || Date.now();
+    return `
+      <div class="message user" data-timestamp="${ts}">
+        <div class="message-avatar">
+          <i class="fas fa-user"></i>
+        </div>
+        <div class="message-content">
+          <div class="message-bubble">
+          <div class="message-text">${escapeHtml(content)}</div>
+          </div>
+        </div>
+      </div>
+    `;
+  }
+
+  function renderAssistantMessage(content, source) {
+    const sourceBadge = source ? `<span class="source-badge">${source}</span>` : '';
+    return `
+      <div class="message assistant">
+        <div class="message-avatar">
+          <i class="fas fa-robot"></i>
+        </div>
+        <div class="message-content">
+          ${sourceBadge}
+          <div class="message-bubble">
+          <div class="message-text">${formatAnswer(content)}</div>
+          </div>
+        </div>
+      </div>
+    `;
+  }
+
+  function renderLoadingMessage() {
+    return `
+      <div class="message assistant">
+        <div class="message-avatar">
+          <i class="fas fa-robot"></i>
+        </div>
+        <div class="message-content">
+          <div class="message-bubble">
+            <div class="loading-message">
+              <span>Thinking...</span>
+          <div class="typing-indicator">
+            <span></span>
+            <span></span>
+            <span></span>
+              </div>
+            </div>
+          </div>
+        </div>
+      </div>
+    `;
+  }
+
+  function renderWelcomeMessage() {
+    return `
+      <div class="welcome-message">
+        <h2>Welcome to Corex!</h2>
+        <p>Ask me anything and I'll help you with accurate, document-backed answers.</p>
+      </div>
+    `;
+  }
+
+  function escapeHtml(text) {
+    const div = document.createElement('div');
+    div.textContent = text;
+    return div.innerHTML;
+  }
+
+  function displayAllMessages() {
+    if (conversationHistory.length === 0) {
+      chatMessages.innerHTML = renderWelcomeMessage();
+      return;
+    }
+
+    let html = '';
+    conversationHistory.forEach(msg => {
+      if (msg.role === 'user') {
+        html += renderUserMessage(msg.content);
+      } else {
+        html += renderAssistantMessage(msg.content, msg.source);
+      }
+    });
+    chatMessages.innerHTML = html;
+    scrollToBottom();
+  }
+
+  function scrollToBottom() {
+    chatMessages.scrollTop = chatMessages.scrollHeight;
+  }
+
+function formatAnswer(text) {
+  if (typeof text !== "string") {
+    text = String(text ?? "No response received.");
+  }
+  return text
+    .split('\n')
+    .filter(line => line.trim())
+    .map(line => `<p>${line}</p>`)
+    .join('');
+}
+
+  // 🔍 Query handler
+  async function handleQuery() {
+    const query = queryInput.value.trim();
+    if (!query) return;
+
+    // Add user message to conversation
+    addUserMessage(query);
+    displayAllMessages();
+
+    // Clear input
+    queryInput.value = '';
+
+    // Show loading message
+    const loadingMessage = renderLoadingMessage();
+    chatMessages.innerHTML += loadingMessage;
+    scrollToBottom();
+
+    try {
+      // Send conversation history to backend
+      const response = await fetch('/query/', {
+        method: 'POST',
+        headers: {
+          'Content-Type': 'application/json',
+        },
+        body: JSON.stringify({
+          query: query,
+          conversation_history: getHistory()
+        })
+      });
+
+      if (!response.ok) throw new Error(`Server returned ${response.status}`);
+      const data = await response.json();
+
+      // Remove loading message and add assistant response
+      chatMessages.innerHTML = chatMessages.innerHTML.replace(loadingMessage, '');
+      addAssistantMessage(data.response, data.source);
+      displayAllMessages();
+
+    } catch (err) {
+      // Remove loading message and add error message
+      chatMessages.innerHTML = chatMessages.innerHTML.replace(loadingMessage, '');
+      addAssistantMessage(`Failed to get response: ${err.message}`, 'Error');
+      displayAllMessages();
+    }
+}
+
+  // 🔗 Event listeners
+  askButton.addEventListener('click', handleQuery);
+  queryInput.addEventListener('keypress', e => {
+    if (e.key === 'Enter') handleQuery();
+  });
+
+  // Auto-resize input
+  queryInput.addEventListener('input', () => {
+    queryInput.style.height = 'auto';
+    queryInput.style.height = queryInput.scrollHeight + 'px';
+  });
+
+  // Dropdown menu functionality
+  const optionsBtn = document.getElementById('optionsBtn');
+  const optionsMenu = document.getElementById('optionsMenu');
+  const downloadTxtBtn = document.getElementById('downloadTxt');
+  const downloadPdfBtn = document.getElementById('downloadPdf');
+  const clearChatBtn = document.getElementById('clearChat');
+
+  // Toggle dropdown menu
+  optionsBtn.addEventListener('click', (e) => {
+    e.stopPropagation();
+    optionsMenu.classList.toggle('show');
+  });
+
+  // Close dropdown when clicking outside
+  document.addEventListener('click', (e) => {
+    if (!optionsBtn.contains(e.target) && !optionsMenu.contains(e.target)) {
+      optionsMenu.classList.remove('show');
+    }
+  });
+
+  // Download as TXT
+  downloadTxtBtn.addEventListener('click', () => {
+    downloadChatAsTxt();
+    optionsMenu.classList.remove('show');
+  });
+
+  // Download as PDF
+  downloadPdfBtn.addEventListener('click', () => {
+    downloadChatAsPdf();
+    optionsMenu.classList.remove('show');
+  });
+
+  // Clear chat
+  clearChatBtn.addEventListener('click', () => {
+    clearConversation();
+    displayAllMessages();
+    optionsMenu.classList.remove('show');
+  });
+
+  // Download functions
+  function downloadChatAsTxt() {
+    if (conversationHistory.length === 0) {
+      alert('No conversation to download');
+      return;
+    }
+
+    let content = 'Corex Chat History\n';
+    content += '='.repeat(50) + '\n\n';
+    
+    conversationHistory.forEach((msg, index) => {
+      const timestamp = new Date(msg.timestamp).toLocaleString();
+      const role = msg.role === 'user' ? 'You' : 'Corex';
+      const source = msg.source ? ` (${msg.source})` : '';
+      
+      content += `[${timestamp}] ${role}${source}:\n`;
+      content += msg.content + '\n\n';
+    });
+
+    const blob = new Blob([content], { type: 'text/plain' });
+    const url = URL.createObjectURL(blob);
+    const a = document.createElement('a');
+    a.href = url;
+    a.download = `corex-chat-${new Date().toISOString().split('T')[0]}.txt`;
+    document.body.appendChild(a);
+    a.click();
+    document.body.removeChild(a);
+    URL.revokeObjectURL(url);
+  }
+
+  function downloadChatAsPdf() {
+    if (conversationHistory.length === 0) {
+      alert('No conversation to download');
+      return;
+    }
+
+    try {
+      const { jsPDF } = window.jspdf;
+      const doc = new jsPDF();
+      
+      // Set up the document
+      let yPosition = 20;
+      const pageHeight = doc.internal.pageSize.height;
+      const pageWidth = doc.internal.pageSize.width;
+      const margin = 20;
+      const maxWidth = pageWidth - (margin * 2);
+      
+      // Helper function to add text with word wrapping
+      function addTextWithWrap(text, x, y, maxWidth, fontSize = 10) {
+        doc.setFontSize(fontSize);
+        const lines = doc.splitTextToSize(text, maxWidth);
+        doc.text(lines, x, y);
+        return y + (lines.length * (fontSize * 0.4));
+      }
+      
+      // Helper function to check if we need a new page
+      function checkNewPage(requiredSpace) {
+        if (yPosition + requiredSpace > pageHeight - 20) {
+          doc.addPage();
+          yPosition = 20;
+          return true;
+        }
+        return false;
+      }
+      
+      // Title
+      doc.setFontSize(16);
+      doc.setFont(undefined, 'bold');
+      doc.text('Corex Chat History', pageWidth / 2, yPosition, { align: 'center' });
+      yPosition += 10;
+      
+      // Date
+      doc.setFontSize(10);
+      doc.setFont(undefined, 'normal');
+      doc.text(`Generated on: ${new Date().toLocaleString()}`, pageWidth / 2, yPosition, { align: 'center' });
+      yPosition += 15;
+      
+      // Add a line
+      doc.line(margin, yPosition, pageWidth - margin, yPosition);
+      yPosition += 10;
+      
+      // Process each message
+      conversationHistory.forEach((msg, index) => {
+        const timestamp = new Date(msg.timestamp).toLocaleString();
+        const role = msg.role === 'user' ? 'You' : 'Corex';
+        const source = msg.source ? ` (${msg.source})` : '';
+        
+        // Check if we need a new page for this message
+        const messageText = `[${timestamp}] ${role}${source}:\n${msg.content}`;
+        const estimatedHeight = (messageText.split('\n').length * 4) + 10;
+        
+        if (checkNewPage(estimatedHeight)) {
+          // Add a continuation marker
+          doc.setFontSize(8);
+          doc.text('...continued from previous page...', margin, yPosition);
+          yPosition += 5;
+        }
+        
+        // Message header
+        doc.setFontSize(10);
+        doc.setFont(undefined, 'bold');
+        yPosition = addTextWithWrap(`[${timestamp}] ${role}${source}:`, margin, yPosition, maxWidth, 10);
+        
+        // Message content
+        doc.setFont(undefined, 'normal');
+        yPosition = addTextWithWrap(msg.content, margin + 5, yPosition, maxWidth - 5, 9);
+        
+        // Add some space between messages
+        yPosition += 8;
+        
+        // Add a subtle line between messages (except for the last one)
+        if (index < conversationHistory.length - 1) {
+          doc.setDrawColor(200, 200, 200);
+          doc.line(margin, yPosition, pageWidth - margin, yPosition);
+          yPosition += 5;
+        }
+      });
+      
+      // Save the PDF
+      const fileName = `corex-chat-${new Date().toISOString().split('T')[0]}.pdf`;
+      doc.save(fileName);
+      
+    } catch (error) {
+      console.error('Error generating PDF:', error);
+      alert('Error generating PDF. Please try downloading as TXT instead.');
+    }
+  }
+
+  // Scroll to bottom when new messages arrive
+  const observer = new MutationObserver(() => {
+    scrollToBottom();
+  });
+  observer.observe(chatMessages, { childList: true, subtree: true });
+
+  // Initialize
+  displayAllMessages();
+});

static/styles.css

@@ -0,0 +1,622 @@
+/* Corex Colorful Theme */
+:root {
+  --bg-primary: #0f0f23;
+  --bg-secondary: #1a1a2e;
+  --bg-tertiary: #16213e;
+  --text-primary: #ffffff;
+  --text-secondary: #e0e6ed;
+  --text-muted: #a0aec0;
+  --border-color: #2d3748;
+  --accent-color: #667eea;
+  --accent-hover: #5a67d8;
+  --accent-secondary: #f093fb;
+  --accent-tertiary: #4facfe;
+  --user-message-bg: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+  --ai-message-bg: linear-gradient(135deg, #4e78ae 0%, #000000 100%);
+  --input-bg: #2d3748;
+  --input-border: #4a5568;
+  --shadow: 0 10px 25px -5px rgba(0, 0, 0, 0.3);
+  --gradient-primary: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+  --gradient-secondary: linear-gradient(135deg, #f093fb 0%, #f5576c 100%);
+  --gradient-tertiary: linear-gradient(135deg, #4facfe 0%, #00f2fe 100%);
+}
+
+* {
+  margin: 0;
+  padding: 0;
+  box-sizing: border-box;
+}
+
+body {
+  font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
+  background-color: var(--bg-primary);
+  color: var(--text-primary);
+  height: 100vh;
+  overflow: hidden;
+}
+
+/* Chat Container */
+.chat-container {
+  display: flex;
+  flex-direction: column;
+  height: 100vh;
+  background-color: var(--bg-primary);
+}
+
+/* Header */
+.chat-header {
+  display: flex;
+  justify-content: space-between;
+  align-items: center;
+  padding: 1rem 1.5rem;
+  border-bottom: 1px solid var(--border-color);
+  background-color: var(--bg-secondary);
+}
+
+.header-left {
+  display: flex;
+  align-items: center;
+  gap: 0.5rem;
+}
+
+.header-left h1 {
+  font-size: 1.25rem;
+  font-weight: 600;
+  color: var(--text-primary);
+}
+
+.header-dropdown {
+  color: var(--text-muted);
+  cursor: pointer;
+  padding: 0.25rem;
+}
+
+.header-right {
+  display: flex;
+  gap: 0.5rem;
+}
+
+.header-btn {
+  background: none;
+  border: none;
+  color: var(--text-muted);
+  cursor: pointer;
+  padding: 0.5rem;
+  border-radius: 0.375rem;
+  transition: background-color 0.2s;
+}
+
+.header-btn:hover {
+  background-color: var(--bg-tertiary);
+  color: var(--text-primary);
+}
+
+/* Message Reactions */
+.reaction-btn {
+  background: none;
+  border: none;
+  color: var(--text-muted);
+  cursor: pointer;
+  padding: 0.25rem 0.5rem;
+  border-radius: 1rem;
+  transition: all 0.2s;
+  font-size: 0.875rem;
+  display: flex;
+  align-items: center;
+  gap: 0.25rem;
+}
+
+.reaction-btn:hover {
+  background-color: var(--bg-tertiary);
+  color: var(--text-primary);
+  transform: scale(1.05);
+}
+
+.reaction-btn.active {
+  background: var(--gradient-primary);
+  color: white;
+  box-shadow: 0 2px 8px rgba(102, 126, 234, 0.3);
+}
+
+.reaction-count {
+  font-size: 0.75rem;
+  font-weight: 500;
+}
+
+/* Search Bar */
+.search-bar {
+  position: absolute;
+  top: 100%;
+  left: 0;
+  right: 0;
+  background: var(--bg-secondary);
+  border: 1px solid var(--border-color);
+  border-radius: 0.5rem;
+  padding: 1rem;
+  box-shadow: var(--shadow);
+  z-index: 1000;
+  opacity: 0;
+  visibility: hidden;
+  transform: translateY(-0.5rem);
+  transition: all 0.3s ease;
+}
+
+.search-bar.show {
+  opacity: 1;
+  visibility: visible;
+  transform: translateY(0);
+}
+
+.search-input {
+  width: 100%;
+  background: var(--input-bg);
+  border: 1px solid var(--input-border);
+  border-radius: 0.5rem;
+  padding: 0.75rem 1rem;
+  color: var(--text-primary);
+  font-size: 0.875rem;
+  margin-bottom: 0.75rem;
+}
+
+.search-input:focus {
+  outline: none;
+  border-color: var(--accent-color);
+  box-shadow: 0 0 0 3px rgba(102, 126, 234, 0.1);
+}
+
+.search-results {
+  max-height: 200px;
+  overflow-y: auto;
+}
+
+.search-result {
+  padding: 0.5rem;
+  border-radius: 0.375rem;
+  cursor: pointer;
+  transition: background-color 0.2s;
+  border: 1px solid transparent;
+}
+
+.search-result:hover {
+  background-color: var(--bg-tertiary);
+  border-color: var(--accent-color);
+}
+
+.search-result-content {
+  font-size: 0.875rem;
+  color: var(--text-secondary);
+  margin-bottom: 0.25rem;
+}
+
+.search-result-meta {
+  font-size: 0.75rem;
+  color: var(--text-muted);
+}
+
+/* Streaming Animation */
+.streaming-text {
+  position: relative;
+}
+
+.streaming-cursor {
+  display: inline-block;
+  width: 2px;
+  height: 1em;
+  background: var(--accent-color);
+  animation: blink 1s infinite;
+  margin-left: 2px;
+}
+
+@keyframes blink {
+  0%, 50% { opacity: 1; }
+  51%, 100% { opacity: 0; }
+}
+
+/* Dropdown Menu */
+.dropdown-container {
+  position: relative;
+  display: inline-block;
+}
+
+.dropdown-menu {
+  position: absolute;
+  top: 100%;
+  right: 0;
+  background-color: var(--bg-secondary);
+  border: 1px solid var(--border-color);
+  border-radius: 0.5rem;
+  box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1), 0 2px 4px -1px rgba(0, 0, 0, 0.06);
+  min-width: 12rem;
+  z-index: 1000;
+  opacity: 0;
+  visibility: hidden;
+  transform: translateY(-0.5rem);
+  transition: all 0.2s ease;
+  margin-top: 0.5rem;
+}
+
+.dropdown-menu.show {
+  opacity: 1;
+  visibility: visible;
+  transform: translateY(0);
+}
+
+.dropdown-item {
+  display: flex;
+  align-items: center;
+  gap: 0.75rem;
+  width: 100%;
+  padding: 0.75rem 1rem;
+  background: none;
+  border: none;
+  color: var(--text-primary);
+  text-align: left;
+  cursor: pointer;
+  transition: background-color 0.2s;
+  font-size: 0.875rem;
+}
+
+.dropdown-item:hover {
+  background-color: var(--bg-tertiary);
+}
+
+.dropdown-item i {
+  width: 1rem;
+  text-align: center;
+  color: var(--text-muted);
+}
+
+.dropdown-divider {
+  height: 1px;
+  background-color: var(--border-color);
+  margin: 0.25rem 0;
+}
+
+/* Chat Messages */
+.chat-messages {
+  flex: 1;
+  overflow-y: auto;
+  padding: 1rem;
+  display: flex;
+  flex-direction: column;
+  gap: 1rem;
+}
+
+/* Welcome Message */
+.welcome-message {
+  text-align: center;
+  padding: 2rem;
+  color: var(--text-secondary);
+}
+
+.welcome-message h2 {
+  font-size: 1.5rem;
+  margin-bottom: 0.5rem;
+  color: var(--text-primary);
+}
+
+.welcome-message p {
+  font-size: 1rem;
+  color: var(--text-muted);
+}
+
+/* Message Bubbles */
+.message {
+  display: flex;
+  gap: 0.75rem;
+  margin-bottom: 1rem;
+  max-width: 100%;
+}
+
+.message.user {
+  justify-content: flex-end;
+}
+
+.message.assistant {
+  justify-content: flex-start;
+}
+
+.message-avatar {
+  width: 2rem;
+  height: 2rem;
+  border-radius: 50%;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  flex-shrink: 0;
+  margin-top: 0.25rem;
+}
+
+.message.user .message-avatar {
+  background: var(--gradient-primary);
+  color: white;
+  box-shadow: 0 4px 15px rgba(102, 126, 234, 0.4);
+}
+
+.message.assistant .message-avatar {
+  background: var(--gradient-secondary);
+  color: white;
+  box-shadow: 0 4px 15px rgba(240, 147, 251, 0.4);
+}
+
+.message-content {
+  max-width: 70%;
+  min-width: 0;
+}
+
+.message.user .message-content {
+  display: flex;
+  flex-direction: column;
+  align-items: flex-end;
+}
+
+.message.assistant .message-content {
+  display: flex;
+  flex-direction: column;
+  align-items: flex-start;
+}
+
+.message-bubble {
+  padding: 0.75rem 1rem;
+  border-radius: 1rem;
+  word-wrap: break-word;
+  line-height: 1.5;
+}
+
+.message.user .message-bubble {
+  background: var(--user-message-bg);
+  color: var(--text-primary);
+  border-bottom-right-radius: 0.25rem;
+  box-shadow: 0 4px 15px rgba(102, 126, 234, 0.3);
+  border: 1px solid rgba(255, 255, 255, 0.1);
+}
+
+.message.assistant .message-bubble {
+  background: var(--ai-message-bg);
+  color: var(--text-primary);
+  border-bottom-left-radius: 0.25rem;
+  box-shadow: 0 4px 15px rgba(240, 147, 251, 0.3);
+  border: 1px solid rgba(255, 255, 255, 0.1);
+}
+
+.message-text {
+  margin-bottom: 0.5rem;
+}
+
+.message-text p {
+  margin-bottom: 0.5rem;
+}
+
+.message-text p:last-child {
+  margin-bottom: 0;
+}
+
+.message-text strong {
+  font-weight: 600;
+  color: var(--text-primary);
+}
+
+/* Source Badge */
+.source-badge {
+  display: inline-block;
+  background-color: var(--accent-color);
+  color: white;
+  padding: 0.25rem 0.5rem;
+  border-radius: 0.375rem;
+  font-size: 0.75rem;
+  font-weight: 500;
+  margin-bottom: 0.5rem;
+}
+
+/* Message Actions */
+.message-actions {
+  display: flex;
+  gap: 0.5rem;
+  margin-top: 0.5rem;
+  opacity: 0;
+  transition: opacity 0.2s;
+}
+
+.message:hover .message-actions {
+  opacity: 1;
+}
+
+.action-btn {
+  background: none;
+  border: none;
+  color: var(--text-muted);
+  cursor: pointer;
+  padding: 0.25rem;
+  border-radius: 0.25rem;
+  transition: color 0.2s;
+}
+
+.action-btn:hover {
+  color: var(--text-primary);
+}
+
+/* Loading Message */
+.loading-message {
+  display: flex;
+  align-items: center;
+  gap: 0.5rem;
+  color: #75ff6b;
+  font-style: italic;
+  font-weight: 500;
+  text-shadow: 0 0 10px rgba(255, 107, 107, 0.4);
+}
+
+.typing-indicator {
+  display: flex;
+  gap: 0.25rem;
+}
+
+.typing-indicator span {
+  width: 0.5rem;
+  height: 0.5rem;
+  background: #ff6b6b;
+  border-radius: 50%;
+  animation: typing 1.4s infinite ease-in-out;
+  box-shadow: 0 0 8px rgba(255, 107, 107, 0.4);
+}
+
+.typing-indicator span:nth-child(2) {
+  animation-delay: 0.2s;
+}
+
+.typing-indicator span:nth-child(3) {
+  animation-delay: 0.4s;
+}
+
+@keyframes typing {
+  0%, 60%, 100% {
+    transform: translateY(0);
+    opacity: 0.5;
+  }
+  30% {
+    transform: translateY(-0.5rem);
+    opacity: 1;
+  }
+}
+
+/* Input Area */
+.chat-input-container {
+  padding: 1rem;
+  background-color: var(--bg-secondary);
+  border-top: 1px solid var(--border-color);
+}
+
+.chat-input-wrapper {
+  display: flex;
+  align-items: center;
+  gap: 0.5rem;
+  background-color: var(--input-bg);
+  border: 1px solid var(--input-border);
+  border-radius: 1rem;
+  padding: 0.75rem 1rem;
+  max-width: 48rem;
+  margin: 0 auto;
+  transition: border-color 0.2s;
+}
+
+.chat-input-wrapper:focus-within {
+  border-color: var(--accent-color);
+}
+
+.chat-input {
+  flex: 1;
+  background: none;
+  border: none;
+  outline: none;
+  color: var(--text-primary);
+  font-size: 1rem;
+  line-height: 1.5;
+}
+
+.chat-input::placeholder {
+  color: var(--text-muted);
+}
+
+.input-btn {
+  background: none;
+  border: none;
+  color: var(--text-muted);
+  cursor: pointer;
+  padding: 0.5rem;
+  border-radius: 0.375rem;
+  transition: color 0.2s;
+}
+
+.input-btn:hover {
+  color: var(--text-primary);
+}
+
+.send-btn {
+  background-color: var(--accent-color);
+  border: none;
+  color: white;
+  cursor: pointer;
+  padding: 0.5rem;
+  border-radius: 0.375rem;
+  transition: background-color 0.2s;
+}
+
+.send-btn:hover {
+  background-color: var(--accent-hover);
+}
+
+.send-btn:disabled {
+  background-color: var(--bg-tertiary);
+  color: var(--text-muted);
+  cursor: not-allowed;
+}
+
+/* Scroll Indicator */
+.scroll-indicator {
+  text-align: center;
+  margin-top: 0.5rem;
+  color: var(--text-muted);
+  cursor: pointer;
+  transition: color 0.2s;
+}
+
+.scroll-indicator:hover {
+  color: var(--text-primary);
+}
+
+/* Responsive Design */
+@media (max-width: 768px) {
+  .chat-header {
+    padding: 0.75rem 1rem;
+  }
+  
+  .chat-messages {
+    padding: 0.75rem;
+  }
+  
+  .chat-input-container {
+    padding: 0.75rem;
+  }
+  
+  .message-content {
+    max-width: 85%;
+  }
+  
+  .header-left h1 {
+    font-size: 1.125rem;
+  }
+}
+
+/* Hide scrollbar but keep functionality */
+.chat-messages::-webkit-scrollbar {
+  width: 0.25rem;
+}
+
+.chat-messages::-webkit-scrollbar-track {
+  background: transparent;
+}
+
+.chat-messages::-webkit-scrollbar-thumb {
+  background: var(--border-color);
+  border-radius: 0.125rem;
+}
+
+.chat-messages::-webkit-scrollbar-thumb:hover {
+  background: var(--text-muted);
+}
+
+/* Animation for new messages */
+@keyframes slideIn {
+  from {
+    opacity: 0;
+    transform: translateY(1rem);
+  }
+  to {
+    opacity: 1;
+    transform: translateY(0);
+  }
+}
+
+.message {
+  animation: slideIn 0.3s ease-out;
+}

templates/index.html

@@ -0,0 +1,90 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+  <meta charset="UTF-8" />
+  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+  <meta name="theme-color" content="#3b82f6" />
+  <title>Corex | AI Assistant</title>
+
+  <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.4.0/css/all.min.css" />
+  <link rel="stylesheet" href="/static/styles.css" />
+  <link rel="preconnect" href="https://fonts.googleapis.com" />
+  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
+  <link href="https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&display=swap" rel="stylesheet" />
+  <script src="https://cdnjs.cloudflare.com/ajax/libs/jspdf/2.5.1/jspdf.umd.min.js"></script>
+</head>
+<body>
+  <div class="chat-container">
+    <!-- Header -->
+    <header class="chat-header">
+      <div class="header-left">
+        <h1>Corex</h1>
+        <div class="header-dropdown">
+          <i class="fas fa-chevron-down"></i>
+        </div>
+      </div>
+      <div class="header-right">
+        <button class="header-btn" title="Share">
+          <i class="fas fa-share"></i>
+        </button>
+        <div class="dropdown-container">
+          <button class="header-btn" id="optionsBtn" title="More options">
+            <i class="fas fa-ellipsis-h"></i>
+          </button>
+          <div class="dropdown-menu" id="optionsMenu">
+            <button class="dropdown-item" id="downloadTxt">
+              <i class="fas fa-file-text"></i>
+              Download as TXT
+            </button>
+            <button class="dropdown-item" id="downloadPdf">
+              <i class="fas fa-file-pdf"></i>
+              Download as PDF
+            </button>
+            <div class="dropdown-divider"></div>
+            <button class="dropdown-item" id="clearChat">
+              <i class="fas fa-trash"></i>
+              Clear Chat
+            </button>
+          </div>
+        </div>
+      </div>
+    </header>
+
+    <!-- Chat Messages -->
+    <main class="chat-messages" id="chatMessages">
+      <div class="welcome-message">
+        <h2>Welcome to Corex!</h2>
+        <p>Ask me anything and I'll help you with accurate, document-backed answers.</p>
+      </div>
+    </main>
+
+    <!-- Input Area -->
+    <div class="chat-input-container">
+      <div class="chat-input-wrapper">
+        <button class="input-btn" title="Attach file">
+          <i class="fas fa-plus"></i>
+        </button>
+        <input
+          type="text"
+          id="queryInput"
+          placeholder="Ask anything"
+          autocomplete="off"
+          class="chat-input"
+        />
+        <button class="input-btn" title="Voice input">
+          <i class="fas fa-microphone"></i>
+        </button>
+        <button id="askButton" class="send-btn" title="Send message">
+          <i class="fas fa-paper-plane"></i>
+        </button>
+      </div>
+      <div class="scroll-indicator">
+        <i class="fas fa-chevron-down"></i>
+      </div>
+    </div>
+  </div>
+
+  <script src="/static/script.js"></script>
+</body>
+</html>
+

vector_rag.py

@@ -0,0 +1,101 @@
+from langchain_community.document_loaders import PyPDFLoader
+from langchain_community.vectorstores import FAISS
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+# Use the generic HuggingFaceEmbeddings for the smaller model
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_huggingface import HuggingFacePipeline
+# Remove BitsAndBytesConfig import
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+# Set cache directories with fallback for permission issues
+os.environ.setdefault('HF_HOME', '/tmp/huggingface_cache')
+os.environ.setdefault('TRANSFORMERS_CACHE', '/tmp/huggingface_cache/transformers')
+os.environ.setdefault('HF_DATASETS_CACHE', '/tmp/huggingface_cache/datasets')
+
+# --- MODEL INITIALIZATION (Minimal Footprint) ---
+print("Loading Qwen2-0.5B-Instruct...")
+model_name = "Qwen/Qwen2-0.5B-Instruct" 
+
+# Removed: quantization_config = BitsAndBytesConfig(load_in_8bit=True) 
+
+tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+# Removed: quantization_config parameter from from_pretrained
+model = AutoModelForCausalLM.from_pretrained(
+    model_name, 
+    device_map="cpu", 
+    trust_remote_code=True
+)
+
+llm_pipeline = pipeline(
+    "text-generation", 
+    model=model, 
+    tokenizer=tokenizer, 
+    max_new_tokens=256, 
+    do_sample=True,
+    temperature=0.5,
+    top_p=0.9,
+)
+llm = HuggingFacePipeline(pipeline=llm_pipeline)
+
+# Use the lighter all-MiniLM-L6-v2 embeddings model
+embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2") 
+
+# --- DOCUMENT LOADING & CHUNKING ---
+loader = PyPDFLoader("data/sample.pdf") # Correct path for Docker: data/sample.pdf
+documents = loader.load()
+text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
+chunks = text_splitter.split_documents(documents)
+
+if not chunks:
+    raise ValueError("No document chunks found.")
+
+# Initialize FAISS and retriever
+vectorstore = FAISS.from_documents(chunks, embeddings)
+retriever = vectorstore.as_retriever()
+
+# Expose the necessary components for rag.py to import
+def query_vector_store(query: str, conversation_history: list = None) -> str:
+    """
+    Query the vector store with conversation context.
+    
+    Args:
+        query: The user's current question
+        conversation_history: List of previous messages (optional)
+    
+    Returns:
+        Answer string or None if no documents found
+    """
+    if conversation_history is None:
+        conversation_history = []
+    
+    docs = retriever.get_relevant_documents(query)
+    if docs:
+        context = "\n\n".join([doc.page_content for doc in docs])
+        
+        # Build prompt with conversation context
+        prompt = "You are a helpful assistant engaged in a conversation.\n\n"
+        
+        if conversation_history:
+            # Format conversation history
+            history_lines = []
+            for msg in conversation_history[-10:]:  # Last 10 messages
+                role = "User" if msg["role"] == "user" else "Assistant"
+                history_lines.append(f"{role}: {msg['content']}")
+            history_text = '\n'.join(history_lines)
+            prompt += f"Previous conversation:\n{history_text}\n\n"
+        
+        prompt += f"""Use the following context from documents to answer the current question:
+
+{context}
+
+Current question: {query}
+Answer:"""
+        
+        raw_output = llm.invoke(prompt)
+        answer = raw_output.replace(prompt, "").strip()
+        return answer
+    return None