Home.py

import gradio as gr
from src.chatbot import chatbot, keyword_search
#from gradio_calendar import Calendar
#from datetime import datetime

# Define important variables
legislature_periods = [
    "All",
    "20. Legislaturperiode",
    "19. Legislaturperiode",
    "18. Legislaturperiode",
    "17. Legislaturperiode",
    "16. Legislaturperiode",
    "15. Legislaturperiode",
    "14. Legislaturperiode",
    "13. Legislaturperiode",
    "12. Legislaturperiode",
    "11. Legislaturperiode",
    "10. Legislaturperiode",
    "9. Legislaturperiode",
    "8. Legislaturperiode",
    "7. Legislaturperiode",
    "6. Legislaturperiode",
    "5. Legislaturperiode",
    "4. Legislaturperiode",
    "3. Legislaturperiode",
    "2. Legislaturperiode",
    "1. Legislaturperiode"
]

partys = ['All','CDU/CSU','SPD','AfD','Grüne','FDP','DIE LINKE.','GB/BHE','DRP', 'WAV', 'NR', 'BP', 'FU', 'SSW', 'KPD', 'DA', 'FVP','DP','Z', 'PDS','Fraktionslos','not found', 'Gast']



with gr.Blocks() as App:
    with gr.Tab("ChatBot"):
        with gr.Blocks(fill_height=True):
            with gr.Accordion(open=False, label="Filter database"):
                # Apply RAG using chatbut function from local file ChatBot.py
                db_inputs = gr.Dropdown(choices=legislature_periods, value="All", multiselect=True, label="Legislature", info="Select a combination of legislatures as basis for the chatbot's replies", show_label=True)
                prompt_language = gr.Dropdown(choices=["DE", "EN"], value="DE",label="Language", info="Choose output language", multiselect=False)
           
            gr.ChatInterface(chatbot,
                        title="PoliticsToYou",
                        description= "Ask anything about your favorite political topic from any legislature period",
                        examples=[
                            ["Wie steht die CDU zur Cannabislegalisierung?", "All", "DE"],
                            ["Wie steht die FDP zur Rente?", "All", "DE"],
                            ["Was sagten die Parteien in der ersten Legislaturperiode über die nazi Vergangenheit?", "1. Legislaturperiode", "DE"],
                            ["Wie wird die Ehe für alle diskutiert?", "18. Legislaturperiode", "DE"],
                            ["How is the GDR perceived?", "11. Legislaturperiode", "EN"]
                            ], 
                        cache_examples=True,  #true increases loading time
                        additional_inputs = [db_inputs, prompt_language],
                        )
        
    with gr.Tab("KeywordSearch"):
        
        with gr.Blocks() as Block:
            # Keyword Input
            keyword_box = gr.Textbox(label='keyword')

            #Additional Input (hidden)
            with gr.Accordion('Detailed filters', open=False):
                # Row orientation
                with gr.Row() as additional_input:
                    n_slider = gr.Slider(label="Number of Results",info="Other filters reduces the returned results", minimum=1, maximum=100, step=1, value=10)
                    party_dopdown = gr.Dropdown(value='All', choices=partys, label='Party') 
                    # ToDo: Add date or legislature filter as input
                    #start_date = Calendar(value="1949-01-01", type="datetime", label="Select start date", info="Click the calendar icon to bring up the calendar.", interactive=True)
                    #end_date = Calendar(value=datetime.today().strftime('%Y-%m-%d'), type="datetime", label="Select end date", info="Click the calendar icon to bring up the calendar.", interactive=True)

            search_btn = gr.Button('Search')

            with gr.Column(visible=False) as output_col:
                results_df = gr.Dataframe(label='Results', interactive=False)

                # Download results from keyword search
                with gr.Accordion('Would you like to download your results?', open=False) as download_row:
                    with gr.Row():
                        ftype_dropdown = gr.Dropdown(choices=["csv","excel","json"], label="Format")
                        export_btn = gr.Button('Export')
                        file = gr.File(file_types=[".xlsx", ".csv", ".json"], visible=False)
    
            # Keyword Search on click
            def search(keyword, n, party): # ToDo: Include party and timedate
                return {
                    output_col: gr.Column(visible=True),
                    results_df: keyword_search(query=keyword, n=n, party_filter=party),
                }

            search_btn.click(
                fn=search,
                inputs=[keyword_box, n_slider, party_dopdown],
                outputs=[output_col, results_df],
            )
           
            # Export data to a downloadable format
            def export(df, keyword, ftype=None):
                if ftype == "csv":
                    file = f'{keyword}.csv'
                    df.to_csv(file, index = False)
                    return gr.File(value=file,visible=True)
                elif ftype == "json":
                    file = f'{keyword}.json'
                    df.to_json(file, index = True)
                    return gr.File(value=file,visible=True)
                else:
                    file = f'{keyword}.xlsx'
                    df.to_excel(file, index = True)
                    return gr.File(value=file,visible=True)
            
            export_btn.click(
                fn=export,
                inputs=[results_df, keyword_box, ftype_dropdown],
                outputs=[file],
            )
            
    with gr.Tab("About"):
        gr.Markdown("""
                    <h2>Welcome to <strong>PoliticsToYou</strong> - your playground for investigating the heart of politics in Germany</h2>
                    <ul>
                    <p>Would you like to gain insights into political debates or reveal party positions on specific topics from any legislature?</p>
                    <p>You can use the ChatBot to ask all your questions or search for related speech content in the Keyword Search section.</p>
                    </ul>
                    <p>Looking forward to your feedback!</p>

                    <h3>Further improvements & Ideas:</h3>
                    <ul>
                        <li>Experiment with different LLMs and Templates</li>
                        <li>Include chat history in RAG</li>
                        <li>Add a date or legislature filter to KeywordSearch</li>
                        <li>Exclude short document splits when creating the vectorstore</li>
                        <li>Improve inference time</li>
                        <li>Add analytic tools for party manifestos</li>
                        <li>Expand the scope to different countries</li>
                        <li>Update vector databases with new speech entries every n months (currently static)</li>
                    </ul>

                    <p>Big thank you to the OpenDiscourse team for creating the underlying speeches corpus. Check out their website <a href="https://opendiscourse.de/">here</a>.</p>

                    """
                    )
        
if __name__ == "__main__":
    App.launch(share=False) # true not supported on hf spaces