Update app.py

app.py

@@ -2,6 +2,7 @@ from openai import OpenAI
 import gradio as gr
 import numpy as np
 
+# 初始化 OpenAI 客戶端
 client = OpenAI(api_key="你的OPENAI_API_KEY")
 
 # ============ 1. 定義主要專業領域 ============ #
@@ -14,42 +15,60 @@ PROFESSIONS = {
     "設計": "你是一位設計師，回答必須專業、詳細，提供設計步驟與案例。"
 }
 
-# ============ 2. 預先建立職業向量庫 ============ #
+# ============ 2. Embedding 工具 (UTF-8 安全) ============ #
 def get_embedding(text, model="text-embedding-3-small"):
+    if not isinstance(text, str):
+        text = str(text)
+    text = text.encode("utf-8", "ignore").decode("utf-8")
     emb = client.embeddings.create(input=text, model=model)
     return np.array(emb.data[0].embedding)
 
-profession_embeddings = {field: get_embedding(field) for field in PROFESSIONS.keys()}
+profession_embeddings = {}  # 延遲初始化
+
+def ensure_embeddings():
+    global profession_embeddings
+    if not profession_embeddings:
+        profession_embeddings = {field: get_embedding(field) for field in PROFESSIONS.keys()}
 
 # ============ 3. NLP 判斷最接近職業 ============ #
 def detect_profession(detail: str) -> str:
+    ensure_embeddings()
     if not detail.strip():
         return "你是一個專業顧問，回答必須專業、詳細、可操作。"
     
     detail_emb = get_embedding(detail)
-    
-    scores = {}
-    for field, emb in profession_embeddings.items():
-        scores[field] = np.dot(detail_emb, emb) / (np.linalg.norm(detail_emb) * np.linalg.norm(emb))
-    
+    scores = {
+        field: np.dot(detail_emb, emb) / (np.linalg.norm(detail_emb) * np.linalg.norm(emb))
+        for field, emb in profession_embeddings.items()
+    }
     best_field = max(scores, key=scores.get)
     return PROFESSIONS[best_field]
 
-# ============ 4. AI Agent 回答 ============ #
-def professional_agent(user_input, main_field, detail, chat_history=[]):
-    if main_field:
-        system_prompt = PROFESSIONS.get(main_field, "你是一個專業顧問，回答必須專業、詳細、可操作。")
-    else:
-        system_prompt = detect_profession(detail)
-    
-    messages = [{"role": "system", "content": system_prompt}]
-    
-    for h in chat_history:
+# ============ 4. AI Agent 回答邏輯 ============ #
+def professional_agent(user_input, state):
+    """
+    state 結構:
+    {
+        "chat_history": [(user, assistant), ...],
+        "profession_prompt": None or str
+    }
+    """
+
+    # 如果尚未設定專業 → 第一輪一定是輸入職業
+    if state.get("profession_prompt") is None:
+        profession_prompt = detect_profession(user_input)
+        state["profession_prompt"] = profession_prompt
+        answer = f"✅ 已設定你的專業領域。接下來請提出問題。"
+        state["chat_history"].append((user_input, answer))
+        return answer, state["chat_history"], state
+
+    # 正常對話
+    messages = [{"role": "system", "content": state["profession_prompt"]}]
+    for h in state["chat_history"]:
         messages.append({"role": "user", "content": h[0]})
         messages.append({"role": "assistant", "content": h[1]})
-    
     messages.append({"role": "user", "content": user_input})
-    
+
     try:
         response = client.chat.completions.create(
             model="gpt-4o-mini",
@@ -59,28 +78,25 @@ def professional_agent(user_input, main_field, detail, chat_history=[]):
         answer = response.choices[0].message.content
     except Exception as e:
         answer = f"發生錯誤: {str(e)}"
-    
-    chat_history.append((user_input, answer))
-    if len(chat_history) > 10:
-        chat_history = chat_history[-10:]
-    
-    return answer, chat_history
+
+    state["chat_history"].append((user_input, answer))
+    if len(state["chat_history"]) > 10:
+        state["chat_history"] = state["chat_history"][-10:]
+
+    return answer, state["chat_history"], state
 
 # ============ 5. Gradio 介面 ============ #
 with gr.Blocks() as demo:
     gr.Markdown("## 🧑‍💼 全職業專業 AI 顧問 (NLP 智能判斷)")
-    
-    with gr.Row():
-        main_field = gr.Dropdown(["", *list(PROFESSIONS.keys())], label="主要領域 (可選)", value="")
-        detail = gr.Textbox(label="細部說明 (例如: 會計師、骨科醫師、前端工程師)")
-    
+    gr.Markdown("👉 第一次請先輸入你的職業，例如：`會計師`、`骨科醫師`、`前端工程師`")
+
     chatbot = gr.Chatbot()
-    msg = gr.Textbox(label="輸入你的問題")
-    state = gr.State([])
-    
-    def chat(user_input, main_field, detail, chat_history):
-        return professional_agent(user_input, main_field, detail, chat_history)
-    
-    msg.submit(chat, [msg, main_field, detail, state], [chatbot, state])
+    msg = gr.Textbox(label="輸入訊息")
+    state = gr.State({"chat_history": [], "profession_prompt": None})
+
+    def chat(user_input, state):
+        return professional_agent(user_input, state)
+
+    msg.submit(chat, [msg, state], [chatbot, state, state])
 
 demo.launch(server_name="0.0.0.0", server_port=7860)