Spaces:
Sleeping
Sleeping
| import asyncio | |
| import gradio as gr | |
| import pandas as pd | |
| from data_access import get_questions, get_source_finders, get_run_ids, get_baseline_rankers, \ | |
| get_unified_sources | |
| # Initialize data at the module level | |
| questions = [] | |
| source_finders = [] | |
| questions_dict = {} | |
| source_finders_dict = {} | |
| question_options = [] | |
| baseline_rankers = [] | |
| baseline_rankers_dict = {} | |
| baseline_ranker_options = [] | |
| run_ids = [] | |
| finder_options = [] | |
| finder_labels = {"All": "All Source Finders"} | |
| previous_run_id = None | |
| run_id_dropdown = None | |
| # Get all questions | |
| # Initialize data in a single async function | |
| async def initialize_data(): | |
| global questions, source_finders, questions_dict, source_finders_dict, question_options, finder_options, finder_labels, baseline_rankers, baseline_ranker_options | |
| questions = await get_questions() | |
| source_finders = await get_source_finders() | |
| baseline_rankers = await get_baseline_rankers() | |
| baseline_rankers_dict = {f["id"]: f["name"] for f in baseline_rankers} | |
| # Convert to dictionaries for easier lookup | |
| questions_dict = {q["id"]: q["text"] for q in questions} | |
| source_finders_dict = {f["id"]: f["name"] for f in source_finders} | |
| # Create formatted options for dropdowns | |
| question_options = [f"{q['id']}: {q['text']}" for q in questions] | |
| finder_options = [str(f["id"]) for f in source_finders] | |
| finder_labels = {str(f["id"]): f["name"] for f in source_finders} | |
| baseline_ranker_options = [f["id"] for f in baseline_rankers] | |
| baseline_ranker_labels = {str(f["id"]): f["name"] for f in source_finders} | |
| def update_sources_list(question_option, source_finder_id, baseline_ranker_id: str, run_id: str, evt: gr.EventData = None): | |
| global previous_run_id | |
| if evt and evt.target and evt.target.elem_id == "run_id_dropdown": | |
| if run_id == previous_run_id: | |
| return gr.update(), gr.update(), gr.update(), gr.update() | |
| # Store the current run_id for future comparison | |
| previous_run_id = run_id | |
| return asyncio.run(update_sources_list_async(question_option, source_finder_id, baseline_ranker_id, run_id)) | |
| # Main function to handle UI interactions | |
| async def update_sources_list_async(question_option, source_finder_id, baseline_ranker_id: str, run_id: str): | |
| if not question_option: | |
| return gr.update(), gr.update(), gr.update(), "No question selected" | |
| # Extract question ID from selection | |
| question_id = int(question_option.split(":")[0]) | |
| available_run_ids = await get_run_ids(question_id) | |
| run_id_options = [str(r_id) for r_id in available_run_ids] | |
| if run_id not in run_id_options: | |
| run_id = run_id_options[0] | |
| run_id_int = int(run_id) | |
| finder_id_int = None if len(source_finder_id) == 0 else int(source_finder_id) | |
| if type(baseline_ranker_id) == list: | |
| baseline_ranker_id = baseline_ranker_id[0] | |
| baseline_ranker_id_int = 1 if len(baseline_ranker_id) == 0 else int(baseline_ranker_id) | |
| source_runs = None | |
| stats = None | |
| # Get source runs data | |
| if finder_id_int: | |
| source_runs, stats = await get_unified_sources(question_id, finder_id_int, run_id_int, baseline_ranker_id_int) | |
| # Create DataFrame for display | |
| df = pd.DataFrame(source_runs) | |
| if not source_runs: | |
| return None, None, run_id_options, "No results found for the selected filters", | |
| # Format table columns | |
| columns_to_display = ['sugya_id', 'in_baseline', 'baseline_rank', 'in_source_run', 'source_run_rank', 'tractate', | |
| 'folio', 'reason'] | |
| df_display = df[columns_to_display] if all(col in df.columns for col in columns_to_display) else df | |
| # CSV for download | |
| # csv_data = df.to_csv(index=False) | |
| result_message = f"Found {len(source_runs)} results" | |
| return df_display, stats, run_id_options, result_message, | |
| # Create Gradio app | |
| # Ensure we clean up when done | |
| async def main(): | |
| global run_id_dropdown | |
| await initialize_data() | |
| with gr.Blocks(title="Source Runs Explorer") as app: | |
| gr.Markdown("# Source Runs Explorer") | |
| with gr.Row(): | |
| with gr.Column(scale=3): | |
| with gr.Row(): | |
| with gr.Column(scale=1): | |
| # Main content area | |
| question_dropdown = gr.Dropdown( | |
| choices=question_options, | |
| label="Select Question", | |
| value=None, | |
| interactive=True, | |
| elem_id="question_dropdown" | |
| ) | |
| with gr.Column(scale=1): | |
| baseline_rankers_dropdown = gr.Dropdown( | |
| choices=baseline_ranker_options, | |
| label="Select Baseline Ranker", | |
| interactive=True, | |
| elem_id="baseline_rankers_dropdown" | |
| ) | |
| with gr.Row(): | |
| with gr.Column(scale=1): | |
| source_finder_dropdown = gr.Dropdown( | |
| choices=finder_options, | |
| label="Source Finder", | |
| interactive=True, | |
| elem_id="source_finder_dropdown" | |
| ) | |
| with gr.Column(scale=1): | |
| run_id_dropdown = gr.Dropdown( | |
| choices=run_ids, | |
| value="1", | |
| allow_custom_value=True, | |
| label="Run id for Question and source finder", | |
| interactive=True, | |
| elem_id="run_id_dropdown" | |
| ) | |
| result_text = gr.Markdown("Select a question to view source runs") | |
| gr.Markdown("# Source Run Statistics") | |
| statistics_table = gr.DataFrame( | |
| headers=["num_high_ranked_baseline_sources", | |
| "num_high_ranked_found_sources", | |
| "overlap_count", | |
| "overlap_percentage", | |
| "high_ranked_overlap_count", | |
| "high_ranked_overlap_percentage" | |
| ], | |
| interactive=False, | |
| ) | |
| gr.Markdown("# Sources Found") | |
| results_table = gr.DataFrame( | |
| headers=['Source Finder', 'Run ID', 'Sugya ID', 'Tractate', 'Folio', 'Rank', 'Reason'], | |
| interactive=False | |
| ) | |
| # download_button = gr.DownloadButton( | |
| # label="Download Results as CSV", | |
| # interactive=True, | |
| # visible=True | |
| # ) | |
| with gr.Column(scale=1): | |
| # Sidebar area | |
| gr.Markdown("### About") | |
| gr.Markdown("This tool allows you to explore source runs for Talmudic questions.") | |
| gr.Markdown("Start by selecting a question, then optionally filter by source finder and run ID.") | |
| gr.Markdown("### Statistics") | |
| gr.Markdown(f"Total Questions: {len(questions)}") | |
| gr.Markdown(f"Source Finders: {len(source_finders)}") | |
| gr.Markdown("### Source Finders") | |
| for f in source_finders: | |
| gr.Markdown(f"**{f['id']}**: {f['name']}") | |
| gr.Markdown("### Baseline Source Rankers") | |
| for f in baseline_rankers: | |
| gr.Markdown(f"**{f['id']}**: {f['name']}") | |
| # Set up event handlers | |
| question_dropdown.change( | |
| update_sources_list, | |
| inputs=[question_dropdown, source_finder_dropdown, run_id_dropdown, baseline_rankers_dropdown], | |
| # outputs=[run_id_dropdown, results_table, result_text, download_button] | |
| outputs=[results_table, statistics_table, run_id_dropdown, result_text] | |
| ) | |
| source_finder_dropdown.change( | |
| update_sources_list, | |
| inputs=[question_dropdown, source_finder_dropdown, run_id_dropdown, baseline_rankers_dropdown], | |
| # outputs=[run_id_dropdown, results_table, result_text, download_button] | |
| outputs=[results_table, statistics_table, run_id_dropdown, result_text] | |
| ) | |
| run_id_dropdown.change( | |
| update_sources_list, | |
| inputs=[question_dropdown, source_finder_dropdown, run_id_dropdown, baseline_rankers_dropdown], | |
| outputs=[results_table, statistics_table, run_id_dropdown, result_text] | |
| ) | |
| app.queue() | |
| app.launch() | |
| if __name__ == "__main__": | |
| asyncio.run(main()) | |