Update app.py

app.py

@@ -4,15 +4,19 @@ import traceback
 import gradio as gr
 import requests
 import pandas as pd
+from openai import OpenAI
 from smolagents import CodeAgent, DuckDuckGoSearchTool, tool
-from smolagents.models import BaseModel
+from smolagents.models import OpenAIServerModel
 
 logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
 logger = logging.getLogger(__name__)
 
 SUBMISSION_URL = "https://agents-course-unit4-scoring.hf.space"
-POLLINATIONS_API = "https://text.pollinations.ai/"
-MODEL_ID = os.getenv("MODEL_ID", "openai-large")
+GITHUB_TOKEN = os.getenv("GITHUB_TOKEN")
+if not GITHUB_TOKEN:
+    raise ValueError("CRITICAL: GITHUB_TOKEN environment variable not set.")
+GITHUB_ENDPOINT = "https://models.github.ai/inference"
+MODEL_ID = os.getenv("MODEL_ID", "openai/gpt-4o-mini")
 
 try:
     search_tool_instance = DuckDuckGoSearchTool()
@@ -66,8 +70,8 @@ def wikipedia_lookup(page_title: str) -> str:
                 logger.warning(f"Wikipedia page '{title}' is disambiguation.")
                 return f"Wikipedia Error: '{title}' is a disambiguation page. Try a more specific title."
             else:
-                logger.warning(f"Wikipedia page '{title}' found but has no summary.")
-                return f"Wikipedia Error: Page '{title}' found but has no summary."
+                 logger.warning(f"Wikipedia page '{title}' found but has no summary.")
+                 return f"Wikipedia Error: Page '{title}' found but has no summary."
     except requests.exceptions.HTTPError as e:
         if e.response.status_code == 404:
             logger.warning(f"Wikipedia page not found: {page_safe}")
@@ -98,51 +102,19 @@ Formatting Rules for FINAL ANSWER:
 Let's begin!
 """
 
-# Benutzerdefiniertes Modell für Pollinations API
-class PollinationsModel(BaseModel):
-    def __init__(self, model_id="openai-large", max_tokens=8196, seed=42):
-        self.model_id = model_id
-        self.max_tokens = max_tokens
-        self.seed = seed
-        self.api_url = POLLINATIONS_API
-        logger.info(f"Initialized PollinationsModel with model_id={model_id}")
-
-    def generate(self, prompt, **kwargs):
-        logger.info(f"Generating with PollinationsModel. Prompt length: {len(prompt)}")
-        try:
-            payload = {
-                "messages": [
-                    {"role": "user", "content": prompt}
-                ],
-                "model": self.model_id,
-                "max_tokens": self.max_tokens,
-                "seed": self.seed,
-                "jsonMode": False,
-                "private": True
-            }
-            response = requests.post(self.api_url, json=payload, timeout=120)
-            response.raise_for_status()
-            result = response.json()
-            if "content" in result:
-                return result["content"]
-            else:
-                logger.error(f"Unexpected response structure: {result}")
-                return "Error: Unexpected API response format"
-        except Exception as e:
-            logger.exception(f"PollinationsModel generate failed: {e}")
-            return f"Error generating response: {str(e)}"
-
-logger.info(f"Initializing Pollinations LLM connection: {MODEL_ID}")
+logger.info(f"Initializing LLM connection: {MODEL_ID} @ {GITHUB_ENDPOINT}")
 try:
-    llm_model = PollinationsModel(
+    logger.info("Attempting to configure OpenAIServerModel with 'api_base' (and no request_timeout)...")
+    llm_model = OpenAIServerModel(
         model_id=MODEL_ID,
-        max_tokens=8196,
-        seed=42
+        api_key=GITHUB_TOKEN,
+        api_base=GITHUB_ENDPOINT
+        # Removed request_timeout=60
     )
-    logger.info("LLM connection configured using Pollinations API.")
+    logger.info("LLM connection configured using 'api_base'.")
 except Exception as e:
-    logger.exception("CRITICAL: Failed to configure PollinationsModel")
-    raise RuntimeError(f"Could not configure PollinationsModel: {e}") from e
+    logger.exception("CRITICAL: Failed to configure OpenAIServerModel (tried with api_base)")
+    raise RuntimeError(f"Could not configure SmolAgents model using api_base: {e}") from e
 
 logger.info("Initializing CodeAgent...")
 try:
@@ -197,7 +169,7 @@ def evaluate_and_submit():
         logger.info(f"Processing Q {i+1}/{len(questions)} (ID: {task_id})...")
         raw_agent_output = run_agent_on_question(question_text)
         final_answer = "AGENT_ERROR: No 'FINAL ANSWER:' marker."
-        marker = "FINAL ANSWER:"
+        marker = "FINAL ANSWER:";
         if marker in raw_agent_output: final_answer = raw_agent_output.split(marker, 1)[1].strip()
         elif "AGENT_ERROR:" in raw_agent_output: final_answer = raw_agent_output
         results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": final_answer, "Full Output": raw_agent_output})
@@ -224,7 +196,7 @@ def evaluate_and_submit():
 
 logger.info("Setting up Gradio interface...")
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 🚀 Agent Evaluation Runner 🚀\nEnsure all configurations are set. Click Run to start.")
+    gr.Markdown("# 🚀 Agent Evaluation Runner 🚀\nEnsure `GITHUB_TOKEN` secret is set. Click Run to start.")
     run_button = gr.Button("▶️ Run Evaluation & Submit All Answers", variant="primary")
     status_textbox = gr.Textbox(label="📊 Status", lines=4, interactive=False)
     results_df_display = gr.DataFrame(label="📋 Detailed Log", headers=["Task ID", "Question", "Submitted Answer", "Full Output"], wrap=True, column_widths=["10%", "25%", "20%", "45%"])
@@ -234,4 +206,4 @@ logger.info("Gradio interface setup complete.")
 if __name__ == "__main__":
     logger.info("Launching Gradio application...")
     demo.launch(debug=True, share=False)
-    logger.info("Gradio application launched.")
+    logger.info("Gradio application launched.") __name__ == "__main__": logger.info("Launching Gradio application...") demo.launch(debug=True, share=False) logger.info("Gradio application launched.")