Update app.py

app.py

@@ -1,7 +1,16 @@
-"""Gradio interface for the smolagents CodeAgent."""
+"""
+Gradio interface for the smolagents CodeAgent.
+
+By default this script will NOT launch a local Gradio server. To allow
+local runs (for testing) set the environment variable `RUN_LOCAL=1` or pass
+`--run-local` on the command line. This prevents accidental local launches
+when you intended to deploy to Hugging Face Spaces.
+"""
 import os
 import sys
 import logging
+import json
+from typing import List, Dict, Any
 from code_agent import run_agent
 
 # (Gradio update/analytics env flags removed per user request)
@@ -21,6 +30,68 @@ def respond(prompt: str) -> str:
         logger.error("Agent failed: %s", e)
         return f"Agent error: {type(e).__name__}: {str(e)[:200]}"
 
+def extract_prompt_from_question(q: Dict[str, Any]) -> str:
+    """Extract the actual question/prompt from a question dict."""
+    for key in ("question", "prompt", "input", "text", "task"):
+        if key in q and isinstance(q[key], str):
+            return q[key]
+    return str(q)
+
+def fetch_all_questions() -> List[Dict[str, Any]]:
+    """Fetch all questions from the scoring API."""
+    try:
+        from evaluation_client import ScoringAPIClient
+        client = ScoringAPIClient()
+        questions = client.get_questions()
+        return questions if questions else []
+    except Exception as e:
+        logger.error("Failed to fetch questions: %s", e)
+        return []
+
+def answer_all_questions(questions: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+    """Answer all questions and return answers in submission format.
+    
+    Args:
+        questions: List of question dicts from the scoring API.
+    
+    Returns:
+        List of dicts with keys: task_id, submitted_answer.
+    """
+    answers = []
+    total = len(questions)
+    
+    for idx, q in enumerate(questions, start=1):
+        task_id = q.get("task_id") or q.get("id") or q.get("taskId")
+        if not task_id:
+            logger.warning("Question %d/%d: Missing task_id, skipping", idx, total)
+            continue
+
+        prompt = extract_prompt_from_question(q)
+        logger.info("Question %d/%d: task_id=%s, prompt_len=%d", idx, total, task_id, len(prompt))
+        
+        try:
+            ans = run_agent(prompt)
+            ans = ans.strip()
+            logger.info("  ✓ Answer: %s", ans[:60])
+            answers.append({"task_id": task_id, "submitted_answer": ans})
+        except Exception as e:
+            logger.error("  ❌ Failed to answer: %s", type(e).__name__)
+            answers.append({"task_id": task_id, "submitted_answer": f"(error) {type(e).__name__}"})
+    
+    logger.info("✓ Answered %d/%d questions", len(answers), total)
+    return answers
+
+def submit_answers(username: str, agent_code_url: str, answers: List[Dict[str, Any]]) -> str:
+    """Submit answers to the scoring API."""
+    try:
+        from evaluation_client import ScoringAPIClient
+        client = ScoringAPIClient()
+        resp = client.submit(username=username, agent_code=agent_code_url, answers=answers)
+        return f"✓ Submission successful! Response: {resp}"
+    except Exception as e:
+        logger.error("Submission failed: %s", e)
+        return f"ERROR: Submission failed: {e}"
+
 
 _demo = None
 
@@ -65,15 +136,72 @@ def _get_demo():
 
     with gr.Blocks() as demo:
         gr.Markdown("# Agents Course — Final Agent Demo")
+
+        # Single question interface
         with gr.Row():
             prompt_box = gr.Textbox(label="Fetched Prompt (read-only)", lines=6)
         with gr.Row():
             fetch_btn = gr.Button("Fetch Random Task")
             run_btn = gr.Button("Run Agent on Fetched Task")
-        output_box = gr.Textbox(label="Agent Response", lines=12)
+        single_output = gr.Textbox(label="Agent Response", lines=6)
+
+        # Batch processing interface
+        gr.Markdown("## Batch Processing")
+        questions_state = gr.State([])
+        answers_state = gr.State([])
+        status_box = gr.Textbox(label="Status", lines=4)
+
+        with gr.Row():
+            fetch_all_btn = gr.Button("Fetch All Questions")
+            answer_all_btn = gr.Button("Answer All Questions")
+            submit_btn = gr.Button("Submit All Answers")
+
+        with gr.Row():
+            username_box = gr.Textbox(label="Hugging Face Username", placeholder="your_hf_username")
+            agent_url_box = gr.Textbox(label="Agent Code URL", placeholder="https://huggingface.co/spaces/...")
+
+        batch_output = gr.Textbox(label="Batch Results", lines=12)
 
+        # Wire up the buttons
         fetch_btn.click(fn=fetch_random_task, inputs=[], outputs=[prompt_box])
-        run_btn.click(fn=run_on_current, inputs=[prompt_box], outputs=[output_box])
+        run_btn.click(fn=run_on_current, inputs=[prompt_box], outputs=[single_output])
+
+        def fetch_all_questions_wrapper():
+            questions = fetch_all_questions()
+            status = f"Fetched {len(questions)} questions"
+            return questions, [], status
+
+        def answer_all_questions_wrapper(questions):
+            if not questions:
+                return [], "No questions to answer"
+            answers = answer_all_questions(questions)
+            status = f"Answered {len(answers)} questions"
+            return answers, status
+
+        def submit_answers_wrapper(username, agent_url, answers):
+            if not answers:
+                return "No answers to submit"
+            if not username or not agent_url:
+                return "Please provide both username and agent code URL"
+            return submit_answers(username, agent_url, answers)
+
+        fetch_all_btn.click(
+            fn=fetch_all_questions_wrapper,
+            inputs=[],
+            outputs=[questions_state, answers_state, status_box]
+        )
+
+        answer_all_btn.click(
+            fn=answer_all_questions_wrapper,
+            inputs=[questions_state],
+            outputs=[answers_state, status_box]
+        )
+
+        submit_btn.click(
+            fn=submit_answers_wrapper,
+            inputs=[username_box, agent_url_box, answers_state],
+            outputs=[batch_output]
+        )
 
     _demo = demo
     return _demo