Update app.py

app.py

@@ -1,8 +1,8 @@
 import os
 import time
+import traceback
 import gradio as gr
 import requests
-import inspect
 import pandas as pd
 import spaces
 
@@ -15,19 +15,24 @@ def _keep_alive():
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-# Yeni API key'lere acik olan modeller (tercih sirasiyla).
-# Google eski modelleri yeni key'lere kapattigi icin liste onemli.
-PREFERRED_MODELS = [
-    "gemini-3.1-flash-lite",    # lite modeller çok daha az yoğun
+# TEK KAYNAK: model sirasi sadece burada. Sirasiyla denenir.
+# Not: 2.x modeller listede gorunse bile yeni API key'lere kapali (404 verir),
+# o yuzden buraya hic koymuyoruz.
+MODEL_CANDIDATES = [
+    "gemini-3.1-flash-lite",
     "gemini-3-flash-preview",
     "gemini-3.5-flash-lite",
     "gemini-flash-latest",
-    "gemini-3.5-flash"
+    "gemini-3.5-flash",
+    "gemini-3.6-flash",
 ]
 
+# 503 / asiri yuk hatalarini yakalamak icin anahtar kelimeler
+OVERLOAD_KEYS = ("503", "UNAVAILABLE", "ServiceUnavailable", "overloaded", "high demand")
 
-def pick_gemini_model(api_key: str) -> str:
-    """API key'in gercekten erisebildigi modeller arasindan uygun olani secer."""
+
+def log_available_models(api_key: str) -> None:
+    """Sadece bilgi amacli: key'in gordugu modelleri loglar."""
     try:
         r = requests.get(
             "https://generativelanguage.googleapis.com/v1beta/models",
@@ -47,65 +52,50 @@ def pick_gemini_model(api_key: str) -> str:
         print("=" * 70)
     except Exception as e:
         print(f"[MODEL LIST ERROR] {type(e).__name__}: {e}")
-        return PREFERRED_MODELS[0]
-
-    # 1) Tercih listesinden ilk eslesen
-    for p in PREFERRED_MODELS:
-        if p in available:
-            print(f"[MODEL SELECTED] {p}")
-            return p
-
-    # 2) Yoksa herhangi bir 'flash' text modeli
-    blacklist = ("image", "tts", "audio", "live", "embedding", "vision")
-    flashes = [
-        m for m in available
-        if "flash" in m and not any(b in m for b in blacklist)
-    ]
-    if flashes:
-        print(f"[MODEL SELECTED - fallback] {flashes[0]}")
-        return flashes[0]
-
-    # 3) Son care
-    if available:
-        print(f"[MODEL SELECTED - last resort] {available[0]}")
-        return available[0]
-
-    return PREFERRED_MODELS[0]
 
 
 # --- Basic Agent Definition ---
 class BasicAgent:
-    FALLBACKS = ["gemini-3.5-flash", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-flash-latest"]
-
     def __init__(self):
         self.api_key = os.environ["GEMINI_API_KEY"]
-        pick_gemini_model(self.api_key)          # sadece listeyi logla
+        log_available_models(self.api_key)
         self.model_idx = 0
         self._build_agent()
 
-    def _build_agent(self):
-        from smolagents import CodeAgent, PythonInterpreterTool
-        from smolagents import LiteLLMModel
-        name = self.FALLBACKS[self.model_idx]
-        print(f"[MODEL] {name} ile agent kuruluyor")
+    def _build_agent(self) -> None:
+        from smolagents import CodeAgent, LiteLLMModel, PythonInterpreterTool
+
+        name = MODEL_CANDIDATES[self.model_idx]
+        print(f"[MODEL] agent kuruluyor -> gemini/{name}")
+
         model = LiteLLMModel(
             model_id=f"gemini/{name}",
             api_key=self.api_key,
             num_retries=5,
         )
+
         try:
             from smolagents import WebSearchTool
             search_tool = WebSearchTool()
+            print("[TOOL] WebSearchTool")
         except ImportError:
             from smolagents import DuckDuckGoSearchTool
             search_tool = DuckDuckGoSearchTool()
-        self.agent = CodeAgent(tools=[search_tool, PythonInterpreterTool()], model=model, max_steps=6)
+            print("[TOOL] DuckDuckGoSearchTool")
 
-    def _switch_model(self):
-        if self.model_idx + 1 < len(self.FALLBACKS):
+        self.agent = CodeAgent(
+            tools=[search_tool, PythonInterpreterTool()],
+            model=model,
+            max_steps=6,
+        )
+
+    def _switch_model(self) -> bool:
+        if self.model_idx + 1 < len(MODEL_CANDIDATES):
             self.model_idx += 1
+            print(f"[FALLBACK] siradaki modele geciliyor: {MODEL_CANDIDATES[self.model_idx]}")
             self._build_agent()
             return True
+        print("[FALLBACK] denenecek baska model kalmadi")
         return False
 
     def _clean(self, text: str) -> str:
@@ -116,7 +106,6 @@ class BasicAgent:
         return text.strip().strip('"').rstrip(".")
 
     def __call__(self, question: str) -> str:
-        import traceback
         prompt = (
             "You are answering a benchmark question. Your response is graded by EXACT string match.\n"
             "Output ONLY the answer itself: no explanation, no sentence, no units unless the question asks for them, "
@@ -124,16 +113,18 @@ class BasicAgent:
             "If the answer is a number, write just the number. If it is a name, write just the name.\n\n"
             f"Question: {question}"
         )
-        for attempt in range(2):
+
+        for attempt in range(3):
             try:
                 answer = self.agent.run(prompt)
                 cleaned = self._clean(answer)
                 if cleaned:
                     return cleaned
             except Exception as e:
-                print(f"[AGENT ERROR] attempt {attempt+1} | {type(e).__name__}: {e}")
+                err = f"{type(e).__name__}: {e}"
+                print(f"[AGENT ERROR] attempt {attempt+1} | {err}")
                 traceback.print_exc()
-                if "503" in str(e) or "UNAVAILABLE" in str(e):
+                if any(k in err for k in OVERLOAD_KEYS):
                     self._switch_model()
                 time.sleep(5)
         return "unknown"
@@ -162,6 +153,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         agent = BasicAgent()
     except Exception as e:
         print(f"Error instantiating agent: {e}")
+        traceback.print_exc()
         return f"Error initializing agent: {e}", None
 
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
@@ -177,13 +169,12 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             print("Fetched questions list is empty.")
             return "Fetched questions list is empty or invalid format.", None
         print(f"Fetched {len(questions_data)} questions.")
-    except requests.exceptions.RequestException as e:
-        print(f"Error fetching questions: {e}")
-        return f"Error fetching questions: {e}", None
     except requests.exceptions.JSONDecodeError as e:
         print(f"Error decoding JSON response from questions endpoint: {e}")
-        print(f"Response text: {response.text[:500]}")
         return f"Error decoding server response for questions: {e}", None
+    except requests.exceptions.RequestException as e:
+        print(f"Error fetching questions: {e}")
+        return f"Error fetching questions: {e}", None
     except Exception as e:
         print(f"An unexpected error occurred fetching questions: {e}")
         return f"An unexpected error occurred fetching questions: {e}", None
@@ -209,6 +200,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             print(f"<<< [{idx}/{total}] answer = {submitted_answer!r}")
         except Exception as e:
             print(f"Error running agent on task {task_id}: {e}")
+            answers_payload.append({"task_id": task_id, "submitted_answer": "unknown"})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
 
         # Free tier RPM limitini asmamak icin nefes payi
@@ -220,8 +212,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
 
     # 4. Prepare Submission
     submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
-    status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
-    print(status_update)
+    print(f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'...")
 
     # 5. Submit
     print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
@@ -237,8 +228,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             f"Message: {result_data.get('message', 'No message received.')}"
         )
         print("Submission successful.")
-        results_df = pd.DataFrame(results_log)
-        return final_status, results_df
+        return final_status, pd.DataFrame(results_log)
     except requests.exceptions.HTTPError as e:
         error_detail = f"Server responded with status {e.response.status_code}."
         try:
@@ -248,23 +238,19 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             error_detail += f" Response: {e.response.text[:500]}"
         status_message = f"Submission Failed: {error_detail}"
         print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
+        return status_message, pd.DataFrame(results_log)
     except requests.exceptions.Timeout:
         status_message = "Submission Failed: The request timed out."
         print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
+        return status_message, pd.DataFrame(results_log)
     except requests.exceptions.RequestException as e:
         status_message = f"Submission Failed: Network error - {e}"
         print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
+        return status_message, pd.DataFrame(results_log)
     except Exception as e:
         status_message = f"An unexpected error occurred during submission: {e}"
         print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
+        return status_message, pd.DataFrame(results_log)
 
 
 # --- Build Gradio Interface using Blocks ---
@@ -281,7 +267,7 @@ with gr.Blocks() as demo:
         ---
         **Disclaimers:**
         Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
-        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
+        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution.
         """
     )