Upload folder using huggingface_hub

app.py

@@ -378,58 +378,52 @@ def generate_tasks(theme, customer_profile):
     4. Emotional connection to the persona and content/styling
     
     The tasks must be in sequential order.
-    Return the tasks as a JSON list of strings in the format: {{"tasks": ["task1", "task2", ...]}}
+    
+    CRITICAL: You MUST return a JSON object with a "tasks" key containing a list of strings.
+    Example: {{"tasks": ["task 1", "task 2", ...]}}
+    Do not include any other text in your response.
     """
     
-    # Try alias-large first, then alias-fast
-    for model_name in ["alias-large", "alias-fast"]:
+    models_to_try = ["alias-large", "alias-huge", "alias-fast"]
+    
+    for attempt in range(5):
         try:
-            # Handle potential 502 with wait
-            response = None
-            for attempt in range(3):
-                try:
-                    print(f"Attempting task generation with {model_name}, attempt {attempt+1}")
-                    if attempt > 0:
-                        # Parallel retry after first failure
-                        print("Proxy error occurred previously. Attempting parallel call to alias-large and alias-huge.")
-                        response = call_llm_parallel(client, ["alias-large", "alias-huge"], [{"role": "user", "content": prompt}])
-                        if isinstance(response, Exception):
-                            raise response
-                    else:
-                        response = client.chat.completions.create(
-                            model=model_name,
-                            messages=[{"role": "user", "content": prompt}]
-                        )
-                    break
-                except Exception as e:
-                    print(f"Error during task generation with {model_name}: {e}")
-                    if "502" in str(e) or "Proxy Error" in str(e):
-                        print(f"Waiting 35 seconds before retry...")
-                        time.sleep(35)
-                    else:
-                        # For other errors, don't wait as long but still retry or move on
-                        time.sleep(1)
+            print(f"Attempt {attempt+1} for task generation...")
+            if attempt > 0:
+                print(f"Retrying in parallel with {models_to_try}")
+                # Wait 35s if it's a retry (likely Proxy Error or Rate Limit)
+                time.sleep(35)
+                response = call_llm_parallel(client, models_to_try, [{"role": "user", "content": prompt}])
+            else:
+                response = client.chat.completions.create(
+                    model="alias-large",
+                    messages=[{"role": "user", "content": prompt}]
+                )
             
-            if not response or isinstance(response, Exception):
-                print(f"Failed to get response from {model_name} after all attempts.")
-                continue
-
-            content = response.choices[0].message.content
-            # Extract JSON from content (it might be wrapped in code blocks)
-            json_str = re.search(r"\{.*\}", content, re.DOTALL)
-            if json_str:
-                tasks_json = json.loads(json_str.group())
-                if "tasks" in tasks_json:
-                    return tasks_json["tasks"]
-                elif isinstance(tasks_json, list):
-                    return tasks_json
-                else:
-                    return list(tasks_json.values())[0]
+            if response and not isinstance(response, Exception):
+                content = response.choices[0].message.content
+                # Robust extraction
+                json_match = re.search(r"\{.*\}", content, re.DOTALL)
+                if json_match:
+                    try:
+                        tasks_json = json.loads(json_match.group())
+                        tasks = tasks_json.get("tasks", [])
+                        if tasks and isinstance(tasks, list) and len(tasks) >= 5:
+                            return tasks[:10]
+                    except:
+                        pass
+                
+                # Fallback: try to extract lines that look like tasks
+                lines = [re.sub(r'^\d+[\.\)]\s*', '', l).strip() for l in content.split('\n') if l.strip()]
+                tasks = [l for l in lines if len(l) > 20 and not l.startswith('{') and not l.startswith('`')]
+                if len(tasks) >= 5:
+                    return tasks[:10]
+                    
+            print(f"Attempt {attempt+1} failed to yield valid tasks.")
         except Exception as e:
-            print(f"Error with model {model_name}: {e}")
-            continue
-
-    return [f"Task {i+1} for {theme} (Failed to generate)" for i in range(10)]
+            print(f"Error in attempt {attempt+1}: {e}")
+            
+    return [f"Task {i+1} for {theme} (Manual fallback)" for i in range(10)]
 
 def handle_generate(theme, customer_profile, num_personas):
     try:

test_remote_api_v3.py

@@ -0,0 +1,16 @@
+from gradio_client import Client
+import json
+
+client = Client("https://harvesthealth-xxg-backup.hf.space/")
+
+print("\n--- Testing /handle_generate with Persona Pool logic ---")
+try:
+    result = client.predict(
+            theme="Education",
+            customer_profile="Student",
+            num_personas=1,
+            api_name="/handle_generate"
+    )
+    print(f"Result: {result}")
+except Exception as e:
+    print(f"Error: {e}")