Subhanzxz/Prodigy-Task4-Chatbot

1import gradio as gr2import os3from langchain_community.document_loaders import WikipediaLoader4from langchain_text_splitters import RecursiveCharacterTextSplitter5from langchain_community.embeddings import HuggingFaceEmbeddings6from langchain_community.vectorstores import Chroma7from langchain_google_genai import ChatGoogleGenerativeAI8from langchain.chains import create_retrieval_chain9from langchain.chains.combine_documents import create_stuff_documents_chain10from langchain_core.prompts import ChatPromptTemplate11 12# Global variable to hold our Vector Database13vector_store = None14 15def load_wikipedia(topic):16    global vector_store17    if not topic.strip():18        return "⚠️ Please enter a topic."19    20    try:21        # Fetch data from Wikipedia22        loader = WikipediaLoader(query=topic, load_max_docs=1)23        docs = loader.load()24        if not docs:25            return "❌ Could not find a Wikipedia page for that topic."26 27        # Chunk the text28        text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)29        chunks = text_splitter.split_documents(docs)30 31        # Create Embeddings and Vector Store32        embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")33        vector_store = Chroma.from_documents(chunks, embeddings)34 35        return f"✅ Success! Loaded Knowledge Base for: **{topic}**. You can now start chatting below!"36    except Exception as e:37        return f"❌ Error loading Wikipedia: {str(e)}"38 39def chat_function(message, history):40    global vector_store41    42    # Check if a document is loaded43    if vector_store is None:44        return "⚠️ Please load a Wikipedia topic at the top of the page first!"45 46    try:47        # Connect to Gemini48        llm = ChatGoogleGenerativeAI(model="gemini-1.5-flash", temperature=0.3)49        50        # Setup RAG Prompt51        system_prompt = (52            "You are a helpful assistant. Use the following pieces of retrieved context to answer "53            "the question. If you don't know the answer based on the context, say that you don't know.\n\n"54            "{context}"55        )56        prompt = ChatPromptTemplate.from_messages([57            ("system", system_prompt),58            ("human", "{input}"),59        ])60 61        # Create the RAG Chain62        retriever = vector_store.as_retriever(search_kwargs={"k": 3})63        question_answer_chain = create_stuff_documents_chain(llm, prompt)64        rag_chain = create_retrieval_chain(retriever, question_answer_chain)65 66        # Get Answer67        response = rag_chain.invoke({"input": message})68        return response["answer"]69    except Exception as e:70        return f"❌ Error generating response: {str(e)}"71 72# Build the Gradio Interface73with gr.Blocks(theme=gr.themes.Soft()) as demo:74    gr.Markdown("# 🧠 Context-Aware Wikipedia Chatbot")75    gr.Markdown("Type a topic below to scrape Wikipedia, build a vector database, and chat with the document using Gemini!")76 77    with gr.Row():78        topic_input = gr.Textbox(label="1. Enter Wikipedia Topic", placeholder="e.g., Quantum Mechanics, The Eiffel Tower...", scale=4)79        load_btn = gr.Button("Load Knowledge Base", variant="primary", scale=1)80 81    status_output = gr.Markdown("⏳ Waiting for topic...")82 83    load_btn.click(fn=load_wikipedia, inputs=topic_input, outputs=status_output)84 85    gr.Markdown("---")86    gr.Markdown("### 2. Chat with the Document")87    88    # Gradio's built in Chat Interface handles memory and history automatically89    gr.ChatInterface(fn=chat_function)90 91# Launch the app92if __name__ == "__main__":93    demo.launch()