import gradio as gr import os from typing import List, Tuple import requests from bs4 import BeautifulSoup from langchain_text_splitters import RecursiveCharacterTextSplitter from langchain_openai import OpenAIEmbeddings, ChatOpenAI from langchain_community.vectorstores import Chroma from langchain_core.prompts import ChatPromptTemplate from langchain_core.runnables import RunnablePassthrough from langchain_core.output_parsers import StrOutputParser # Global variables to store state vector_store = None qa_chain = None wiki_url = None api_key = None def fetch_wiki_content(wiki_url: str) -> List[str]: """Fetch content from a wiki URL.""" try: response = requests.get(wiki_url, timeout=10) response.raise_for_status() soup = BeautifulSoup(response.content, 'html.parser') # Extract text content # Remove script and style elements for script in soup(["script", "style"]): script.decompose() # Get text text = soup.get_text() # Clean up text lines = (line.strip() for line in text.splitlines()) chunks = (phrase.strip() for line in lines for phrase in line.split(" ")) text = '\n'.join(chunk for chunk in chunks if chunk) return [text] except Exception as e: raise Exception(f"Error fetching wiki content: {str(e)}") def fetch_wiki_pages(wiki_url: str, max_pages: int = 20) -> List[str]: """Fetch multiple pages from a wiki by following links.""" pages_content = [] visited = set() links_to_visit = [] try: # Start with the main page response = requests.get(wiki_url, timeout=10) response.raise_for_status() soup = BeautifulSoup(response.content, 'html.parser') # Get main page content for script in soup(["script", "style"]): script.decompose() text = soup.get_text() lines = (line.strip() for line in text.splitlines()) chunks = (phrase.strip() for line in lines for phrase in line.split(" ")) text = '\n'.join(chunk for chunk in chunks if chunk) if text: pages_content.append(text) visited.add(wiki_url) # Collect all links from the main page base_url = '/'.join(wiki_url.split('/')[:3]) links = soup.find_all('a', href=True) for link in links: href = link.get('href') if not href: continue # Build full URL if href.startswith('/'): full_url = base_url + href elif href.startswith('http'): # Only include links from the same domain if base_url in href: full_url = href else: continue else: continue # Skip if already visited or in queue if full_url not in visited and full_url not in links_to_visit: # Filter out common non-content links skip_patterns = ['#', 'javascript:', 'mailto:', '.pdf', '.jpg', '.png', '.gif', 'login', 'logout', 'register', 'search', 'edit', 'history'] if not any(pattern in href.lower() for pattern in skip_patterns): links_to_visit.append(full_url) # Continue fetching pages until we reach max_pages or run out of links while len(pages_content) < max_pages and links_to_visit: current_url = links_to_visit.pop(0) if current_url in visited: continue try: page_response = requests.get(current_url, timeout=10, allow_redirects=True) if page_response.status_code == 200: page_soup = BeautifulSoup(page_response.content, 'html.parser') for script in page_soup(["script", "style", "nav", "footer", "header"]): script.decompose() page_text = page_soup.get_text() lines = (line.strip() for line in page_text.splitlines()) chunks = (phrase.strip() for line in lines for phrase in line.split(" ")) page_text = '\n'.join(chunk for chunk in chunks if chunk) # Only add if there's substantial content if page_text and len(page_text.strip()) > 100: pages_content.append(page_text) visited.add(current_url) # Optionally collect more links from this page if we need more pages if len(pages_content) < max_pages: page_links = page_soup.find_all('a', href=True) for link in page_links: href = link.get('href') if not href: continue if href.startswith('/'): new_url = base_url + href elif href.startswith('http') and base_url in href: new_url = href else: continue if new_url not in visited and new_url not in links_to_visit: skip_patterns = ['#', 'javascript:', 'mailto:', '.pdf', '.jpg', '.png', '.gif', 'login', 'logout', 'register', 'search', 'edit', 'history'] if not any(pattern in href.lower() for pattern in skip_patterns): links_to_visit.append(new_url) except Exception as e: # Skip failed pages and continue continue return pages_content except Exception as e: raise Exception(f"Error fetching wiki pages: {str(e)}") def initialize_rag(wiki_url_input: str, api_key_input: str, max_pages: int = 20, progress=gr.Progress()) -> str: """Initialize the RAG system with wiki content.""" global vector_store, qa_chain, wiki_url, api_key if not wiki_url_input: return "❌ Please enter a wiki URL" if not api_key_input: return "❌ Please enter an API key" try: progress(0.1, desc="Setting up API key...") api_key = api_key_input wiki_url = wiki_url_input os.environ["OPENAI_API_KEY"] = api_key progress(0.2, desc="Fetching wiki content...") # Fetch wiki pages (convert to int in case it comes as float from Number input) documents = fetch_wiki_pages(wiki_url_input, max_pages=int(max_pages)) if not documents: return "❌ Failed to fetch wiki content. Please check the URL." progress(0.4, desc="Splitting documents into chunks...") # Split documents into chunks text_splitter = RecursiveCharacterTextSplitter( chunk_size=1000, chunk_overlap=200, length_function=len, ) texts = text_splitter.create_documents(documents) progress(0.6, desc="Creating embeddings...") # Create embeddings embeddings = OpenAIEmbeddings(openai_api_key=api_key) progress(0.8, desc="Building vector store...") # Create vector store vector_store = Chroma.from_documents( documents=texts, embedding=embeddings, persist_directory="./chroma_db" ) progress(0.9, desc="Setting up QA chain...") # Create QA chain using newer LangChain API llm = ChatOpenAI(temperature=0, openai_api_key=api_key, model_name="gpt-3.5-turbo") # Create prompt template prompt = ChatPromptTemplate.from_template("""Use the following pieces of context to answer the question at the end. If you don't know the answer, just say that you don't know, don't try to make up an answer. Use three sentences maximum and keep the answer concise. Context: {context} Question: {question} Answer:""") # Create retriever retriever = vector_store.as_retriever(search_kwargs={"k": 3}) # Create the chain def format_docs(docs): return "\n\n".join(doc.page_content for doc in docs) qa_chain = ( {"context": retriever | format_docs, "question": RunnablePassthrough()} | prompt | llm | StrOutputParser() ) progress(1.0, desc="Ready!") return f"✅ Successfully loaded wiki from {wiki_url_input} ({len(documents)} pages)" except Exception as e: return f"❌ Error initializing RAG: {str(e)}" def chat_with_docs(message: str, history: List[List[str]]) -> Tuple[str, List[List[str]]]: """Handle chat messages with the RAG system.""" global qa_chain if qa_chain is None: return "⚠️ Please initialize the RAG system first by entering a wiki URL and API key, then clicking 'Initialize RAG'.", history try: # Get response from QA chain answer = qa_chain.invoke(message) # Add to history history.append([message, answer]) return "", history except Exception as e: error_msg = f"❌ Error: {str(e)}" history.append([message, error_msg]) return "", history # Create Gradio interface with gr.Blocks(title="Chat with Wiki Documents") as app: gr.Markdown("# 📚 Chat with Wiki Documents") gr.Markdown("Enter your wiki URL and API key to start chatting with your documents!") with gr.Row(): with gr.Column(scale=2): wiki_url_input = gr.Textbox( label="Wiki URL", placeholder="https://example.com/wiki", value="" ) with gr.Column(scale=2): api_key_input = gr.Textbox( label="OpenAI API Key", placeholder="sk-...", type="password", value="" ) with gr.Column(scale=1): max_pages_input = gr.Number( label="Max Pages", value=20, minimum=1, maximum=100, step=1, info="Number of pages to fetch" ) with gr.Row(): init_btn = gr.Button("Initialize RAG", variant="primary", scale=1) status_output = gr.Textbox( label="Status", interactive=False, value="Waiting for initialization..." ) gr.Markdown("---") chatbot = gr.Chatbot( label="Chat", height=500, show_copy_button=True ) with gr.Row(): msg_input = gr.Textbox( label="Message", placeholder="Ask a question about the wiki...", scale=4, container=False ) submit_btn = gr.Button("Send", variant="primary", scale=1) # Event handlers init_btn.click( fn=initialize_rag, inputs=[wiki_url_input, api_key_input, max_pages_input], outputs=status_output ) submit_btn.click( fn=chat_with_docs, inputs=[msg_input, chatbot], outputs=[msg_input, chatbot] ) msg_input.submit( fn=chat_with_docs, inputs=[msg_input, chatbot], outputs=[msg_input, chatbot] ) if __name__ == "__main__": app.launch()