import os
import gradio as gr
from langchain_core.prompts import ChatPromptTemplate 
from langchain_core.output_parsers import StrOutputParser 
from langchain_openai import ChatOpenAI 

# 讀取API Key
#openai_api_key = os.getenv(“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”)

# 初始化 API 模型
model = ChatOpenAI(
	model=”gpt-4o-mini”,
	api_key=openai_api_key,
)

# Multi-Agent 的簡單函數
def agent1(input_text:str):
	return f”Agent1:{input_text.upper()}”

def agent2(input_from_agent1:str):
	return f”Agent2: {input_from_agent1} processed!”

# LangChain prompt: 模板
prompt_template = ChatPromptTemplate.from_template(
	“你是航空公司客服助理。”
	“用戶詢問航空資訊時, 提供法蘭克福直飛行程 (來回經濟艙NT$33,000),”
	“並建議轉機至慕尼黑。請根據以下對話回答:\n{conversation}”
)

parser = StrOutputParser()
chain = prompt_template〡model〡parser 

#建立一個全域共享變數, 讓 Multi-Agent 的輸出能傳給 Chatbot
Shared_context = {“latest_output”:””}

#Chatbot 回應函數 (含歷史記憶 + shared context)
def generate_response(message, history):
	formatted_history = “\n”.join(
   		[f”User:{h[0]}\nBot:{h[1]}” for h in history if h[1] is not None]
	)
	# 加入 Multi-Agent 輸出的上下文
	If shared_context[“latest_output”]:
		formatted_history += f”\n(系統提示 : 這是來自 Multi-Agent 的資訊
{shared_context[‘latest_output’]})\n”

	full_input = formatted_history + f”\nUser:{message}\nBot:” 
	return chain.invoke({“conversation”: full_input})

# Gradio 介面
with gr.Blocks(title=Multi-Agent & Chatbot”) as demo:
	gr.Markdown(##🐮多智能體與航班客服助理”)

	with gr.Tabs():
		#------------Multi-Agent” 分頁 ------------
		with gr.Tab(“Multi-Agent”):
			with gr.Row():
				input1 = gr.Textbox(
					label = “輸入給 Agent1”,
					value = “農曆春節期間, 貴公司有沒有直飛到德國慕尼黑的行程?”
				)

			with gr.Row():
				btn1 = gr.Button (“執行 Agent1”)
				output1 = gr.Textbox(label=”Agent1 輸出”, interative=False )
			with gr.Row():
				btn2 = gr.Button(“執行 Agent2”)
				output2 = gr.Textbox(lagel=”Agent2 輸出”, interactive=False)

			# 按鈕綁定 
			btn1.click(fn=agent1, inputs=input1, outputs=output1)

			#Agent2 產出時, 同步更新 shared_context
			def run_agent2(input_text):
				out = agent2(input_text)
				shared_context[“latest_output”] = out
				return out

			btn2.click(fn=agent2, inputs=output1,outputs=output2)
			#------------Chatbot 分頁 ------------
			with gr.Tab(“牛牛機器人”):
				gr.ChatInterface(
					fn=generate_response,
					title= “Tour Inquiry Chatbot”,
					description=”詢問航班資訊, 例如訂票或轉機建議。”
				)

If_name_== ”_main_”:
	demo.launch()