app.py

import os, re, base64
from langchain_core.documents import Document
from langchain_chroma import Chroma
from openai import OpenAI
from langchain.embeddings.base import Embeddings
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain_community.vectorstores import FAISS
import gradio as gr
from langchain.memory import ConversationBufferMemory

# =============================================
# 1️⃣ 內建 Embedding：使用 Gemini embedding API
# =============================================
from langchain_community.embeddings import HuggingFaceEmbeddings

embedding = HuggingFaceEmbeddings(model_name="BAAI/bge-small-zh-v1.5")

# =============================================
# 2️⃣ 載入 QA 檔案並分類
# =============================================
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
path = os.path.join(BASE_DIR, "QA_v2.txt")

if not os.path.exists(path):
    raise FileNotFoundError(f"❌ 找不到 QA 檔案：{path}")

with open(path, "r", encoding="utf-8") as f:
    text = f.read()

pattern = r"(Ｑ[:：].*?)(?=Ｑ[:：]|$)"
qas = re.findall(pattern, text, flags=re.S)

qa_docs = {"證券": [], "期貨": [], "複委託": []}
for qa in qas:
    if "證券" in qa:
        qa_docs["證券"].append(Document(page_content=qa.strip(), metadata={"source": path}))
    elif "期貨" in qa:
        qa_docs["期貨"].append(Document(page_content=qa.strip(), metadata={"source": path}))
    elif "複委託" in qa:
        qa_docs["複委託"].append(Document(page_content=qa.strip(), metadata={"source": path}))

print("✅ 已成功讀取 QA 並完成分類：", {k: len(v) for k, v in qa_docs.items()})

# =============================================
# 3️⃣ 建立向量資料庫（使用 FAISS，記憶體型）
# =============================================
vectordbs = {}
for k, docs in qa_docs.items():
    vectordbs[k] = FAISS.from_documents(docs, embedding)

# =============================================
# 4️⃣ 初始化 Gemini LLM
# =============================================
API_KEY = os.getenv("GOOGLE_API_KEY")
if not API_KEY:
    raise ValueError("⚠️ 未設定 GOOGLE_API_KEY，請在 Hugging Face Secrets 中新增。")

llm = ChatGoogleGenerativeAI(model='gemini-2.5-flash', google_api_key=API_KEY)
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

# =============================================
# 5️⃣ 對話邏輯
# =============================================
def auto_detect_category(text):
    if any(k in text for k in ["股票", "證券", "開戶", "下單", "交割", "現股"]):
        return "證券"
    elif any(k in text for k in ["期貨", "選擇權", "結算", "保證金", "契約"]):
        return "期貨"
    elif any(k in text for k in ["複委託", "海外", "美股", "港股", "國外"]):
        return "複委託"
    else:
        return "證券"

def chat_fn(message, history):
    category = auto_detect_category(message)
    vectordb = vectordbs.get(category)
    if not vectordb:
        return "目前尚無此類別的知識庫。"

    docs = vectordb.similarity_search(message, k=2)
    context = "\n\n".join([d.page_content for d in docs]) if docs else "查無相關內容。"

    prompt = f"""
我是一位金融客服人員。根據以下公司規章內容回答使用者問題：
---
{context}
---
使用者問題：{message}
    """

    try:
        response = llm.invoke(prompt)
        reply = response.content.strip()
    except Exception as e:
        reply = f"⚠️ 生成錯誤：{e}"

    return reply or "請洽營業員"

# =============================================
# 6️⃣ Gradio 介面
# =============================================
logo_path = os.path.join(BASE_DIR, "mega.png")
logo_base64 = ""
if os.path.exists(logo_path):
    with open(logo_path, "rb") as f:
        logo_base64 = base64.b64encode(f.read()).decode("utf-8")

logo_path = os.path.join(BASE_DIR, "mega.png")
logo_base64 = ""
if os.path.exists(logo_path):
    with open(logo_path, "rb") as f:
        logo_base64 = base64.b64encode(f.read()).decode("utf-8")

gr.HTML("""
    <style>
        /* ====== 桌機預設：單行顯示 ====== */
        #main-title {
            font-size: 28px;
            font-weight: bold;
            text-align: center;
            line-height: 1.4;
            margin: 0;
            display: inline-block;
        }

        /* ====== 手機版：自動兩行顯示 ====== */
        @media (max-width: 768px) {
            #main-title {
                font-size: 24px;               /* 👈 手機字體略小 */
                white-space: pre-line;
            }
            #main-title::before {
                content: "👨‍💼 我是小智\\A您的金融好幫手 🫰"; /* \\A = 換行 */
                white-space: pre;               /* 保留換行格式 */
            }
            #main-title span {
                display: none;                  /* 隱藏原本的單行文字 */
            }
        }
    </style>

    <div id="main-title-wrapper" style="text-align:center; margin-top:20px;">
        <h1 id='main-title'><span>👨‍💼 我是小智&nbsp;&nbsp;您的金融好幫手 🫰</span></h1>
        <p id='sub-title' style='margin-top:10px; font-size:14px; color:#666;'>Powered by Gemini & LangChain</p>
    </div>
""")

    with gr.Row():
        with gr.Column(scale=4):
            chatbox = gr.Chatbot(label="💬 對話紀錄", type="messages")

            with gr.Row(elem_id="input-row"):
                user_input = gr.Textbox(
                    elem_id="user-input",
                    show_label=False,
                    placeholder="輸入訊息...",
                    scale=8
                )
                send_btn = gr.Button("送出", elem_id="send-btn", scale=1)

            def handle_input(message, history):
                if not message.strip():
                    return history, gr.update(value="")
                reply = chat_fn(message, history)
                history = history + [
                    {"role": "user", "content": message},
                    {"role": "assistant", "content": reply}
                ]
                return history, gr.update(value="")

            user_input.submit(handle_input, [user_input, chatbox], [chatbox, user_input])
            send_btn.click(handle_input, [user_input, chatbox], [chatbox, user_input])

        with gr.Column(scale=1):
            gr.Markdown("### 👇 快速提問")
            btns = [
                ("未成年可以開戶嗎？", "未成年可以開戶嗎?"),
                ("法人開戶要準備什麼？", "法人開戶要準備什麼?"),
                ("期貨交易保證金是什麼？", "期貨交易保證金是什麼?"),
                ("複委託要如何下單？", "複委託要如何下單?"),
                ("美股交易時間？", "美股交易時間?"),
                ("美股可以定期定額嗎？", "美股可以定期定額嗎?")
            ]
            for label, q in btns:
                gr.Button(label).click(lambda h, q=q: handle_input(q, h), [chatbox], [chatbox, user_input])

            def clear_memory():
                memory.clear()
                return [], gr.update(value="", placeholder="輸入訊息...")
            gr.Button("🧹 整理畫面").click(clear_memory, outputs=[chatbox, user_input])

    # 底部版權列
    gr.HTML("<div id='footer'>© Fintech Assistant — 僅業務使用，非官方授權</div>")

    # 手機鍵盤彈出時捲動補丁
    demo.load(None, None, None, js="""
        window.addEventListener('focusin', () => {
            document.querySelector('textarea')?.scrollIntoView({ behavior: 'smooth', block: 'center' });
        });
    """)

demo.launch()