Update backend/chatbot.py

backend/chatbot.py

@@ -4,10 +4,20 @@ from dotenv import load_dotenv
 from messages import krishna_blessings
 from ayush_messages import ayush_surprises
 
+# Load environment variables (Hugging Face Space secrets)
 load_dotenv()
-REPLICATE_API_TOKEN = os.getenv("REPLICATE_API_TOKEN")
+HUGGINGFACE_API_TOKEN = os.getenv("HUGGINGFACE_API_TOKEN")
 
 def get_krishna_response(user_input):
+    """
+    Generate a Krishna-themed response for the user input.
+    
+    Args:
+        user_input (str): The user's message.
+    
+    Returns:
+        str: Krishna's response.
+    """
     # Check for predefined responses
     if "birthday" in user_input.lower():
         return ayush_surprises["birthday"]
@@ -16,38 +26,30 @@ def get_krishna_response(user_input):
     if "hello" in user_input.lower() or "hi" in user_input.lower():
         return krishna_blessings["greeting"]
 
-    # Use LLaMA-3 via Replicate API
+    # Use Gemma via Hugging Face Inference API for dynamic responses
     headers = {
-        "Authorization": f"Token {REPLICATE_API_TOKEN}",
+        "Authorization": f"Bearer {HUGGINGFACE_API_TOKEN}",
         "Content-Type": "application/json"
     }
     payload = {
-        "input": {
-            "prompt": f"You are Little Krishna, playful and wise. Respond to this in a fun, Krishna-like way: '{user_input}'",
-            "max_tokens": 60,
+        "inputs": f"You are Little Krishna, playful and wise. Respond to this in a fun, Krishna-like way: '{user_input}'",
+        "parameters": {
+            "max_length": 60,
             "temperature": 0.9
         }
     }
-    response = requests.post(
-        "https://api.replicate.com/v1/predictions",
-        headers=headers,
-        json={
-            "version": "meta-llama/llama-3-8b",  # Replace with actual model version after access
-            "input": payload["input"]
-        }
-    )
-    if response.status_code == 200:
-        prediction = response.json()
-        # Wait for prediction to complete (Replicate uses async predictions)
-        prediction_id = prediction["id"]
-        while True:
-            status_response = requests.get(
-                f"https://api.replicate.com/v1/predictions/{prediction_id}",
-                headers=headers
-            )
-            status_data = status_response.json()
-            if status_data["status"] == "succeeded":
-                return status_data["output"]
-            elif status_data["status"] == "failed":
-                return "Sorry, I couldn’t generate a response. Let’s try something else!"
-    return "Error connecting to LLaMA-3. Please try again later."
+    try:
+        response = requests.post(
+            "https://api-inference.huggingface.co/models/google/gemma-2b",
+            headers=headers,
+            json=payload
+        )
+        if response.status_code == 200:
+            result = response.json()
+            return result[0]["generated_text"].strip()
+        else:
+            print(f"Error with Gemma: {response.text}")
+            return "Sorry, I couldn’t generate a response. Let’s try something else!"
+    except Exception as e:
+        print(f"Error connecting to Hugging Face Inference API: {str(e)}")
+        return "Hare Krishna! I seem to be lost in Vrindavan’s magic—let’s try again!"