Update app.py

app.py

@@ -24,6 +24,7 @@ def web_search(query: str) -> str:
     except Exception as e:
         return f"Search error: {e}"
 
+
 def wikipedia_search(query: str) -> str:
     try:
         search_url = "https://en.wikipedia.org/w/api.php"
@@ -48,6 +49,7 @@ def wikipedia_search(query: str) -> str:
     except Exception as e:
         return f"Wikipedia error: {e}"
 
+
 def run_python(code: str) -> str:
     import sys
     from io import StringIO
@@ -63,37 +65,47 @@ def run_python(code: str) -> str:
     finally:
         sys.stdout = old_stdout
 
+
 class SmartAgent:
     def __init__(self):
-        self.api_key = os.getenv("GEMINI_API_KEY")
+        self.api_key = os.getenv("GROQ_API_KEY")
         if not self.api_key:
-            raise ValueError("GEMINI_API_KEY environment variable not set!")
-        # Valid Gemini 1.5 Flash endpoint string
-        self.api_url = f"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key={self.api_key}"
-        print("SmartAgent initialized with Gemini 1.5 Flash (FREE - 1500 req/day)")
+            raise ValueError("GROQ_API_KEY environment variable not set! Please add it to your Space secrets.")
+        
+        # Groq's OpenAI-compatible endpoint
+        self.api_url = "https://api.groq.com/openai/v1/chat/completions"
+        self.model = "llama-3.3-70b-versatile" 
+        print(f"SmartAgent initialized with Groq ({self.model})")
 
     def call_llm(self, prompt: str) -> str:
+        headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json"
+        }
         payload = {
-            "contents": [{"parts": [{"text": prompt}]}],
-            "generationConfig": {
-                "temperature": 0.1,
-                "maxOutputTokens": 1024,
-            }
+            "model": self.model,
+            "messages": [{"role": "user", "content": prompt}],
+            "temperature": 0.1,
+            "max_tokens": 1024
         }
-        # Built-in retry mechanism for Rate Limiting
-        for attempt in range(3):
+        
+        # Exponential backoff retry mechanism 
+        wait_times = [15, 30, 60] 
+        for attempt, wait_time in enumerate(wait_times):
             try:
-                response = requests.post(self.api_url, json=payload, timeout=60)
+                response = requests.post(self.api_url, headers=headers, json=payload, timeout=60)
                 response.raise_for_status()
                 data = response.json()
-                return data["candidates"][0]["content"]["parts"][0]["text"].strip()
+                return data["choices"][0]["message"]["content"].strip()
             except requests.exceptions.HTTPError as e:
-                if response.status_code == 429:
-                    print(f"Rate limited (429)! Waiting 10 seconds... (Attempt {attempt+1}/3)")
-                    time.sleep(10)
+                status = response.status_code
+                if status == 429 or status == 503:
+                    print(f"Groq API Error ({status})! Waiting {wait_time} seconds... (Attempt {attempt+1}/3)")
+                    time.sleep(wait_time)
                 else:
                     raise e
-        raise Exception("Failed to call LLM after 3 attempts due to rate limiting.")
+                    
+        raise Exception("Failed to call LLM after 3 attempts due to API limits or server errors.")
 
     def __call__(self, question: str) -> str:
         print(f"\nQuestion: {question[:100]}...")
@@ -118,8 +130,8 @@ Rules for the answer:
         conversation = f"{system}\n\nQuestion: {question}"
 
         for iteration in range(6):
-            # Slowing down the ReAct loop to respect Gemini's 15 RPM Free Tier limit
-            time.sleep(4.5)
+            # Groq is fast, but we pause slightly to respect free tier limits
+            time.sleep(2.5)
 
             response = self.call_llm(conversation)
             print(f"  LLM [{iteration}]: {response[:200]}")
@@ -139,7 +151,6 @@ Rules for the answer:
             if not python_match:
                 python_match = re.search(r'PYTHON:\s*(.+?)(?:\nSEARCH|\nWIKIPEDIA|\nANSWER|$)', response, re.DOTALL)
 
-            # Execution logic changed to 'elif' to avoid overlapping tool executions
             if search_match:
                 query = search_match.group(1).strip()
                 print(f"  Tool: web_search({query})")
@@ -212,8 +223,8 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             print(f"Error on task {task_id}: {e}")
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"ERROR: {e}"})
         
-        # Pausing between questions to ensure global RPM limit safety
-        time.sleep(5)
+        # Pausing between questions to ensure safety against rate limits
+        time.sleep(3)
 
     if not answers_payload:
         return "Agent did not produce any answers.", pd.DataFrame(results_log)
@@ -247,10 +258,10 @@ with gr.Blocks() as demo:
     gr.Markdown("# 🤖 Smart Agent — GAIA Benchmark Runner")
     gr.Markdown(
         """
-        **Powered by Google Gemini 1.5 Flash (FREE - 1500 requests/day)**
+        **Powered by Groq (Llama 3.3 70B)**
 
         **Instructions:**
-        1. Make sure `GEMINI_API_KEY` is set in your Space secrets
+        1. Make sure `GROQ_API_KEY` is set in your Space secrets
         2. Log in with your Hugging Face account below
         3. Click **Run Evaluation & Submit All Answers**
         """