app.py

import streamlit as st
import requests
import re
import os
import urllib.parse
from dotenv import load_dotenv
from bs4 import BeautifulSoup
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain.docstore.document import Document
import chromadb
from sentence_transformers import SentenceTransformer
import google.generativeai as genai

# Page configuration
st.set_page_config(layout="wide")

# 載入 .env
load_dotenv()
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")

# Initialize Gemini API
genai.configure(api_key=GEMINI_API_KEY)

# Initialize ChromaDB
CHROMA_PATH = "chroma_db"
chroma_client = chromadb.PersistentClient(path=CHROMA_PATH)

# Initialize session state
if 'scraped' not in st.session_state:
    st.session_state.scraped = False
if 'collection_name' not in st.session_state:
    st.session_state.collection_name = "default_collection"
if 'chat_history' not in st.session_state:
    st.session_state.chat_history = []

# Initialize embedding model (中文優化)
embedding_model = SentenceTransformer("shibing624/text2vec-base-chinese")

def clean_text(text):
    return re.sub(r'\s+', ' ', re.sub(r'http\S+', '', text)).strip()

def split_content_into_chunks(content):
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200, length_function=len)
    return text_splitter.split_documents([Document(page_content=content)])

def add_chunks_to_db(chunks, collection_name):
    collection = chroma_client.get_or_create_collection(name=collection_name)
    documents = [chunk.page_content for chunk in chunks]
    embeddings = embedding_model.encode(documents, convert_to_list=True)
    collection.upsert(documents=documents, ids=[f"ID{i}" for i in range(len(chunks))], embeddings=embeddings)

def scrape_text(url, max_depth=1, same_domain=True):
    visited = set()
    base_domain = urllib.parse.urlparse(url).netloc

    def _scrape(u, depth):
        if depth > max_depth or u in visited:
            return
        visited.add(u)
        try:
            response = requests.get(u)
            response.raise_for_status()
            soup = BeautifulSoup(response.text, 'html.parser')
            
            text = clean_text(soup.get_text())
            chunks = split_content_into_chunks(text)
            add_chunks_to_db(chunks, st.session_state.collection_name)
            
            # 遞迴爬取新連結
            if depth < max_depth:
                for link in soup.find_all('a', href=True):
                    next_url = urllib.parse.urljoin(u, link['href'])
                    next_domain = urllib.parse.urlparse(next_url).netloc
                    if same_domain and next_domain != base_domain:
                        continue
                    if next_url.startswith('mailto:') or next_url.startswith('javascript:'):
                        continue
                    _scrape(next_url, depth + 1)
        except requests.exceptions.RequestException:
            pass  # 忽略單一頁面錯誤

    _scrape(url, 1)
    st.session_state.scraped = True
    return "Scraping and processing complete. You can now ask questions!"

def ask_question(query, collection_name):
    collection = chroma_client.get_or_create_collection(name=collection_name)
    query_embedding = embedding_model.encode(query, convert_to_list=True)
    results = collection.query(query_embeddings=[query_embedding], n_results=2)
    top_chunks = results.get("documents", [[]])[0]
    
    system_prompt = f"""
    You are a helpful assistant. Answer only from the provided context. 
    If you lack information, say: "I don't have enough information to answer that question."
    Context:
    {str(top_chunks)}
    """
    
    model = genai.GenerativeModel('gemini-2.0-flash')
    response = model.generate_content(system_prompt + "\nUser Query: " + query)
    return response.text

# Sidebar
with st.sidebar:
    st.header("Database Management")
    if st.button("Clear Chat History"):
        st.session_state.chat_history = []
        st.rerun()
    
    st.header("Step 1: Scrape a Website")
    url = st.text_input("Enter URL:")
    max_depth = st.selectbox("Recursion Depth (層數)", options=[1,2,3,4,5], index=0, help="選擇要遞迴爬幾層，預設1層")
    same_domain = st.checkbox("只允許同網域遞迴", value=True, help="預設只爬同一網域的連結")
    if url and st.button("Scrape & Process"):
        with st.spinner("Scraping..."):
            st.success(scrape_text(url, max_depth=max_depth, same_domain=same_domain))

# Main content
st.title("Web Scraper & Q&A Chatbot")
if st.session_state.scraped:
    st.subheader("Step 2: Ask Questions")
    for message in st.session_state.chat_history:
        with st.chat_message(message["role"]):
            st.write(message["content"])
    
    user_query = st.chat_input("Ask your question here")
    if user_query:
        st.session_state.chat_history.append({"role": "user", "content": user_query})
        with st.spinner("Searching..."):
            answer = ask_question(user_query, st.session_state.collection_name)
        st.session_state.chat_history.append({"role": "assistant", "content": answer})
        
        # Limit chat history to 6 messages
        st.session_state.chat_history = st.session_state.chat_history[-6:]
        st.rerun()
else:
    st.info("Please scrape a website first.")