app.py

# app.py

import streamlit as st
from json_schema import json_schema_response
from function_calling import function_calling_response
from models import UserData

# 設定頁面標題和結構
st.set_page_config(page_title="JSON Extractor", layout="wide")

# 初始化會話狀態
if "conversation_history" not in st.session_state:
    st.session_state["conversation_history"] = []
if "selected_option" not in st.session_state:
    st.session_state["selected_option"] = "JSON Schema"

# 左側工具列選項
with st.sidebar:
    st.header("Methods")
    st.session_state["selected_option"] = st.radio(
        "Please select a method:",
        ("JSON Schema", "Function Calling")
    )

# 主界面標題和說明
st.title("JSON Extractor Using GPT")
st.write("Please input your message containing your name, email, and phone number.")

# 根據選擇的模式顯示不同的輸入框和回應邏輯
# 如果選擇 "JSON Schema" 方法
if st.session_state["selected_option"] == "JSON Schema":
    st.subheader("JSON Schema Method")
    user_input = st.text_input("Please input your message:", key="json_schema_input")
    st.markdown(
        '<p style="font-size: 0.8em; color: gray;">For example: 我的名字是Stella，請用 stelladai1028@gmail.com 聯絡我，電話是0988999999</p>', 
        unsafe_allow_html=True
    )

    if st.button("Submit", key="json_schema_submit"):
        if user_input:
            # 添加使用者輸入到對話歷史
            st.session_state["conversation_history"].append({"role": "user", "content": user_input})
            
            result = json_schema_response(user_input, st.session_state["conversation_history"])
            if isinstance(result, UserData):
                st.subheader("Answer")
                st.json(result.model_dump())
                # 將資料寫入 JSON 檔案
                with open('user-data.json', 'a', encoding='utf-8') as f:
                    f.write(result.model_dump_json() + '\n')
                st.success("Data successfully extracted and stored.")
            elif isinstance(result, str):
                # 顯示特定的錯誤（如電子郵件格式錯誤）
                st.error(result)
            else:
                # 顯示一般輸入錯誤
                st.error("Unable to extract valid data. Please check the input format.")

# 如果選擇 "Function Calling" 方法
elif st.session_state["selected_option"] == "Function Calling":
    st.subheader("Function Calling Method")
    user_input = st.text_input("Please input your message:", key="function_calling_input")
    st.markdown(
        '<p style="font-size: 0.8em; color: gray;">For example: Stella stelladai1028@gmail.com 0988999999</p>', 
        unsafe_allow_html=True
    )

    if st.button("Submit", key="function_calling_submit"):
        if user_input:
            # 添加使用者輸入到對話歷史
            st.session_state["conversation_history"].append({"role": "user", "content": user_input})
            
            result = function_calling_response(user_input, st.session_state["conversation_history"])
            if isinstance(result, UserData):
                st.subheader("Answer")
                st.json(result.model_dump())
                # 將資料寫入 JSON 檔案
                with open('user-data.json', 'a', encoding='utf-8') as f:
                    f.write(result.model_dump_json() + '\n')
                st.success("Data successfully extracted and stored.")
            elif isinstance(result, str):
                # 顯示特定的錯誤（如電子郵件格式錯誤）
                st.error(result)
            else:
                # 顯示一般輸入錯誤
                st.error("Unable to extract valid data. Please check the input format.")

# 顯示對話歷史
st.subheader("Conversation History")
if st.session_state["conversation_history"]:
    for entry in st.session_state["conversation_history"]:
        st.write(entry)
else:
    st.write("No conversation history available.")