# Loading required libraries
import gradio as gr              # Gradio: A library for building web interfaces
import requests                  # requests: A library for sending HTTP requests.
from openai import OpenAI        # OpenAI: Clients compatible with the Upstage Solar API

# ------------------------------
# 🔍 Defining a document parsing function
# ------------------------------
def parse_document(file, api_key):
    """
    Function to convert uploaded PDF document to HTML (using Upstage Document Parse API)
    """
    url = "https://api.upstage.ai/v1/document-ai/document-parse"  # API request URL
    headers = {'Authorization': f'Bearer {api_key}'}              # Authentication header settings정
    files = {"document": open(file.name, "rb")}                   # read file
    data = {
        "base64_encoding": "['table']",  # Table data is encoded in base64
        "model": "document-parse"        # Specify usage model
    }

    response = requests.post(url, headers=headers, files=files, data=data)  # POST request
    result = response.json()                                                # Parsing response results
    html_text = result.get("content", {}).get("html", "")                  # HTML Extraction
    return html_text


# ------------------------------
# 💬 Defining a document-based Q&A function
# ------------------------------
def chat_with_document(history, html_text, user_question, api_key):
    """
    Solar LLM functions to answer user questions based on document content
    """
    if not html_text.strip():
        return history, history, "⚠️ Please convert the document first."  # Guidance if there is no document

    # OpenAI client initialization (Upstage Solar LLM)
    client = OpenAI(
        api_key=api_key,
        base_url="https://api.upstage.ai/v1"
    )

    # Reset previous conversation history
    history = history or []

    # System prompt: Request a response based on the contents of the HTML document
    system_prompt = f"""The following is a financial statement document extracted in HTML format.
                    Please answer user questions accurately and concisely in Korean, based on the text within HTML tags.
                    
                    Document:
                    {html_text}
                    """

    # Message composition (System → User/bot conversation → Current question)
    messages = [{"role": "system", "content": system_prompt}]
    for user, bot in history:
        messages.append({"role": "user", "content": user})
        messages.append({"role": "assistant", "content": bot})
    messages.append({"role": "user", "content": user_question})

    # Solar LLM Call
    try:
        response = client.chat.completions.create(
            model="solar-pro",     # Model name to use
            messages=messages,     # Deliver the entire message
            temperature=0,         # Minimize creativity
            max_tokens=1024        # maximum response length
        )
        bot_reply = response.choices[0].message.content  # Extract response message
    except Exception as e:
        bot_reply = f"⚠️ An error occurred: {str(e)}"  # Error handling

    # Return after updating conversation history
    history.append((user_question, bot_reply))
    return history, history, ""


# ------------------------------
# 🔁 HTML View Toggle Function
# ------------------------------
def toggle_html_view(current_html, is_visible):
    """
    Function to toggle the HTML view/hide state
    """
    return (
        gr.update(value=current_html, visible=not is_visible),  # Hide/Show Text Box
        gr.update(value=current_html, visible=is_visible),      # HTML rendering opposite behavior
        not is_visible                                          # state inversion
    )


# ------------------------------
# 📦 Gradio UI configuration
# ------------------------------
with gr.Blocks() as demo:
    # Show title and description
    gr.Markdown("# 📄 Financial Statement Analysis Chatbot")
    gr.Markdown("1. Convert PDF documents to HTML using the Document Parse API.\n"
                "2. Answer document-based questions with Solar LLM.")

    # 🔑 API Key input window (entered directly by the user)
    api_key_input = gr.Textbox(label="🔑 Upstage API Key", type="password", placeholder="Paste your API key here")

    # 📎 File Upload + Document Conversion Button
    with gr.Row():
        file_input = gr.File(label="📎 Upload financial statements")
        parse_btn = gr.Button("Document HTML conversion")

    # 📘 HTML output area (text + HTML toggle view)
    html_output = gr.Textbox(label="📘 document content", lines=10, visible=True, elem_id="scrollable-html")
    html_display = gr.HTML(visible=False, elem_id="scrollable-html-display")
    toggle_html_btn = gr.Button("🔁 Switch HTML views")
    html_visible_state = gr.State(False)  # Save view state

    # Click the Convert Document button → Generate HTML
    parse_btn.click(
        fn=parse_document,
        inputs=[file_input, api_key_input],
        outputs=html_output
    )

    # When clicking the HTML view switch button → Execute the toggle action
    toggle_html_btn.click(
        fn=toggle_html_view,
        inputs=[html_output, html_visible_state],
        outputs=[html_output, html_display, html_visible_state]
    )

    # 💬 Chatbot Interface
    chatbot = gr.Chatbot(label="💬 Document-based Q&A", height=400)
    user_question = gr.Textbox(label="❓ Please enter your question", lines=2)
    answer_btn = gr.Button("Generate Answer")
    chat_state = gr.State([])  # Save conversation state

    # 💡 Example Question Button Configuration
    with gr.Row():
        gr.Markdown("💡example questions:")
        ex1 = gr.Button("Which company's financial statements are these?")
        ex2 = gr.Button("What was the total net sales for the third quarter?")

    # Click the example question button → Run question + answer
    for btn, question in [(ex1, "Which company's financial statements are these?"), (ex2, "What was the total net sales for the first quarter?")]:
        btn.click(
            fn=lambda q=question: q,        # Pass question text
            inputs=[],
            outputs=user_question
        ).then(
            fn=chat_with_document,
            inputs=[chat_state, html_output, user_question, api_key_input],
            outputs=[chatbot, chat_state, user_question],
            show_progress=True
        )

    # Submit a User Question → Solar LLM Answers
    answer_btn.click(
        fn=chat_with_document,
        inputs=[chat_state, html_output, user_question, api_key_input],
        outputs=[chatbot, chat_state, user_question],
        show_progress=True
    )

# ------------------------------
# 🎨 Styling a scrollable HTML box
# ------------------------------
demo.css = """
#scrollable-html, #scrollable-html-display {
    max-height: 400px;
    overflow: auto;
    border: 1px solid #ccc;
    padding: 10px;
}
"""

# 🚀 run app
if __name__ == "__main__":
    demo.launch()