fig: API error by using loop

app.py

@@ -9,6 +9,7 @@ from collections import deque
 from dotenv import load_dotenv
 from google import genai
 from google.genai import types
+from google.genai.errors import APIError
 
 load_dotenv()
 
@@ -192,8 +193,21 @@ def validate_and_evaluate(description: str, code: str):
     prompt = build_prompt(description, code)
     try:
         raw_output = generate_response(prompt)
+    except APIError as exc:
+        # 1. If it's a 500 error (and the retries failed), show the friendly timeout message
+        if exc.code == 500:
+            timeout_msg = (
+                "❌ API Connection Timeout: The upstream AI provider is currently overloaded "
+                "or unresponsive. Please try again in a few moments."
+            )
+            return "", "", "", timeout_msg
+            
+        # 2. If it's a different API error (e.g., 400 Bad Request, 403 Invalid Key)
+        return "", "", "", f"❌ Upstream API Error ({exc.code}): {exc}"
+        
     except Exception as exc:
-        return "", "", "", f"❌ Model error: {exc}"
+        # 3. Catch any local application bugs or JSON parsing failures here
+        return "", "", "", f"❌ Internal Application Error: {exc}"
 
     verdict, metrics, issues = format_for_display(raw_output)
     return verdict, metrics, issues, ""
@@ -243,16 +257,37 @@ Code to review:
 # ---------------------------------------------------------------------------
 # MODULE 4 — Generation Module
 # ---------------------------------------------------------------------------
-def generate_response(prompt: str) -> str:
-    response = client.models.generate_content(
-        model=LLM_MODEL, 
-        contents=prompt,
-        config=types.GenerateContentConfig(
-            response_mime_type="application/json",
-            response_schema=response_schema, 
-        )
-    )
-    return json.loads(response.text)
+def generate_response(prompt: str, retries: int = 3, initial_delay: float = 1.5) -> dict:
+    delay = initial_delay
+    
+    for i in range(retries):
+        try:
+            response = client.models.generate_content(
+                model=LLM_MODEL, 
+                contents=prompt,
+                config=types.GenerateContentConfig(
+                    response_mime_type="application/json",
+                    response_schema=response_schema, 
+                )
+            )
+            # If successful, parse the JSON and return immediately
+            print("✅ Successful response from Gemini API.")
+            return json.loads(response.text)
+            
+        except APIError as exc:
+            # Catch transient 500 Internal Server Errors and retry
+            if exc.code == 500 and i <= retries - 1:
+                print(f"⚠️ Gemini API 500 error. Retrying attempt {i + 1}/{retries} in {delay}s...")
+                time.sleep(delay)
+                delay *= 2  # Exponential backoff
+                continue
+            
+            # If out of retries, or if it's a 400/403 error, raise it up to validate_and_evaluate
+            raise exc
+            
+        except json.JSONDecodeError as exc:
+            # Failsafe: In rare cases, if the API drops a malformed payload, catch the JSON error
+            raise RuntimeError(f"API returned invalid JSON format: {exc}")
 
 # Build the Gradio UI and connect it to the validation and evaluation function
 app = build_ui(validate_and_evaluate)