added initial version of visibility feature and fixed automatic update of results every hour

app.py

@@ -5,6 +5,7 @@ from envs import RESULTS_REPO_ID, REPO_ID, API, HF_TOKEN
 from pathlib import Path
 import pandas as pd
 import os
+import json
 from utils import parse_json_files, create_scatter_plot
 from huggingface_hub import snapshot_download
 from apscheduler.schedulers.background import BackgroundScheduler
@@ -12,8 +13,84 @@ from apscheduler.schedulers.background import BackgroundScheduler
 def restart_space():
     API.restart_space(repo_id=REPO_ID, token=HF_TOKEN)
 
+# New function to download results
+def download_latest_results():
+    print("Downloading latest results...")
+    snapshot_download(RESULTS_REPO_ID, 
+                    local_dir=abs_path / "evals",
+                    repo_type='dataset',
+                    tqdm_class=None,
+                    etag_timeout=30,
+                    max_workers=4,
+                    )
+    print("Download complete.")
+
 abs_path = Path(__file__).parent
 
+
+# load task_analyses.json from evals/usaco_traces folder
+with open(os.path.join(abs_path, "evals", "usaco_traces", "task_analyses.json"), "r") as f:
+    analyzed_traces = json.load(f)
+
+
+
+def update_task_analysis(task_id):
+    if task_id not in analyzed_traces:
+        return "No analysis available for this task.", [], ""
+    
+    analysis = analyzed_traces[task_id]
+    summary = analysis['summary']
+    
+    if isinstance(summary, str):
+        try:
+            summary = json.loads(summary)
+        except json.JSONDecodeError:
+            return "Error: Unable to parse summary data.", [], ""
+    elif not isinstance(summary, dict):
+        return "Error: Summary data is in an unexpected format.", [], ""
+    
+    overview = f"# Task Overview\n\n{summary.get('overview', 'No overview available.')}\n\n"
+    overview += f"## Successes\n{summary.get('successes', 'No successes listed.')}\n\n"
+    overview += f"## Challenges\n{summary.get('challenges', 'No challenges listed.')}\n\n"
+    
+    steps = [(f"Step {i+1}", i) for i in range(len(analysis['steps']))]
+    
+    return overview, gr.Dropdown(choices=steps, label="Agent Steps"), ""
+
+def update_step_details(task_id, step_index):
+    if task_id not in analyzed_traces:
+        return "No analysis available for this task."
+    
+    if step_index is None:
+        return "Please select a step to view details."
+    
+    steps = analyzed_traces[task_id]['steps']
+    
+    if isinstance(step_index, tuple):
+        step_index = step_index[1]
+    elif isinstance(step_index, str):
+        step_index = int(step_index.split()[-1]) - 1
+    
+    if step_index < 0 or step_index >= len(steps):
+        return f"Invalid step index: {step_index}"
+    
+    step = steps[step_index]
+    analysis = step['analysis']
+    
+    if isinstance(analysis, str):
+        try:
+            analysis = json.loads(analysis)
+        except json.JSONDecodeError:
+            return "Error: Unable to parse step analysis data."
+    elif not isinstance(analysis, dict):
+        return "Error: Step analysis data is in an unexpected format."
+    
+    details = f"# Step {step_index + 1} Details\n\n"
+    details += f"## Description\n{analysis.get('description', 'No description available.')}\n\n"
+    details += f"## Assessment\n{analysis.get('assessment', 'No assessment available.')}\n\n"
+    
+    return details
+
 with gr.Blocks() as demo:
     gr.Markdown("""
     # 🥇 Agent Leaderboard
@@ -54,20 +131,27 @@ with gr.Blocks() as demo:
                                        "results_accuracy": 20,
                                        "results_total_cost": 20},
                     )
+            gr.Markdown("## USACO Task Trace Explorer")
+            with gr.Row():
+                with gr.Column(scale=1):
+                    task_dropdown = gr.Dropdown(choices=list(analyzed_traces.keys()), label="Select USACO Task")
+                    task_overview = gr.Markdown()
+                with gr.Column(scale=1):
+                    steps_dropdown = gr.Dropdown(label="Agent Steps")
+                    step_details = gr.Markdown()
+            
+            task_dropdown.change(update_task_analysis, inputs=[task_dropdown], outputs=[task_overview, steps_dropdown, step_details])
+            steps_dropdown.change(update_step_details, inputs=[task_dropdown, steps_dropdown], outputs=[step_details])
+        
         with gr.Tab("About"):
             gr.Markdown((Path(__file__).parent / "about.md").read_text())
 
 if __name__ == "__main__":
     # Download the results from the Hugging Face Hub
-    snapshot_download(RESULTS_REPO_ID, 
-                    local_dir=abs_path / "evals",
-                        repo_type='dataset',
-                        tqdm_class=None,
-                        etag_timeout=30,
-                        max_workers=4,
-                    )
+    download_latest_results()
     
     scheduler = BackgroundScheduler()
     scheduler.add_job(restart_space, "interval", hours=1) # restarted every 1h
+    scheduler.add_job(download_latest_results, "interval", hours=1) # download latest results every 1h
     scheduler.start()
     demo.launch()