pleaseyrab

app.py

@@ -1,103 +1,187 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
 import os
-import traceback
+import re
 
-# Get token
+# Load token
 HF_TOKEN = os.getenv("HF_TOKEN") or os.getenv("HUGGING_FACE_HUB_TOKEN")
 
-print("\n" + "="*60)
-print("🔍 DIAGNOSTIC MODE - ENHANCED ERROR REPORTING")
-print("="*60)
-print(f"Token exists: {bool(HF_TOKEN)}")
-if HF_TOKEN:
-    print(f"Token preview: {HF_TOKEN[:20]}...{HF_TOKEN[-10:]}")
-print("="*60 + "\n")
+print("="*50)
+if not HF_TOKEN:
+    print("❌ HF_TOKEN NOT FOUND")
+else:
+    print(f"✅ HF_TOKEN Found: {HF_TOKEN[:15]}...{HF_TOKEN[-8:]}")
+print("="*50)
 
-def respond(message, history):
+OFF_TOPIC_REGEX = r"(http|www|buy now|discount|subscribe|follow me|click here)"
+
+def is_safe_to_process(text: str) -> bool:
+    if len(text.strip()) < 2:
+        return False
+    if re.search(OFF_TOPIC_REGEX, text.lower()):
+        return False
+    return True
+
+def respond(
+    message,
+    history: list[dict[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+    request: gr.Request,
+):
+    user_name = "User"
+    if request:
+        user_name = request.query_params.get("username", "User")
+    
+    if not is_safe_to_process(message):
+        yield f"Hello {user_name}, I'm here to support your emotional well-being. How can I help you today?"
+        return
+    
     if not HF_TOKEN:
-        return "❌ ERROR: No HF_TOKEN found in secrets."
+        yield "❌ Configuration Error: HF_TOKEN not found in Space secrets. Please add it in Settings."
+        return
     
-    models = [
-        "mistralai/Mistral-7B-Instruct-v0.1",
-        "google/gemma-2b-it",
-        "microsoft/phi-2",
+    personalized_system = (
+        f"{system_message} "
+        f"The user's name is {user_name}. Address them naturally by name. "
+        f"Be conversational, not formal."
+    )
+    
+    messages = [{"role": "system", "content": personalized_system}]
+    messages.extend(history[-10:])
+    messages.append({"role": "user", "content": message})
+    
+    # Create client once
+    client = InferenceClient(token=HF_TOKEN)
+    
+    # Try chat_completion with these models
+    chat_models = [
+        "meta-llama/Meta-Llama-3-8B-Instruct",
+        "mistralai/Mistral-7B-Instruct-v0.2",
+        "HuggingFaceH4/zephyr-7b-beta",
     ]
     
-    errors_log = []
+    response = ""
     
-    for model_name in models:
+    # Try chat_completion first
+    for model_name in chat_models:
         try:
-            print(f"\n{'='*60}")
-            print(f"🔄 Attempting: {model_name}")
-            print(f"{'='*60}")
-            
-            # Create client
-            client = InferenceClient(token=HF_TOKEN)
-            
-            # Try chat completion
-            messages = [
-                {"role": "system", "content": "You are a helpful assistant."},
-                {"role": "user", "content": message}
-            ]
-            
-            print(f"Calling chat_completion with model={model_name}")
+            print(f"\n{'='*50}")
+            print(f"🔄 Trying chat_completion: {model_name}")
+            print(f"{'='*50}")
             
-            response = ""
             for msg in client.chat_completion(
                 messages,
                 model=model_name,
-                max_tokens=100,
+                max_tokens=max_tokens,
                 stream=True,
+                temperature=temperature,
+                top_p=top_p,
             ):
                 token = msg.choices[0].delta.content or ""
                 response += token
+                yield response
             
             print(f"✅ SUCCESS with {model_name}")
-            print(f"Response: {response[:100]}")
-            return response
+            return
             
         except Exception as e:
-            # Capture FULL error details
-            error_type = type(e).__name__
-            error_msg = str(e)
-            error_trace = traceback.format_exc()
-            
-            print(f"❌ FAILED with {model_name}")
-            print(f"Error Type: {error_type}")
-            print(f"Error Message: {error_msg}")
-            print(f"Full Traceback:\n{error_trace}")
-            print("="*60)
+            error_str = str(e)
+            error_repr = repr(e)
+            print(f"❌ Failed with {model_name}")
+            print(f"Error str: {error_str}")
+            print(f"Error repr: {error_repr}")
+            print(f"Error type: {type(e).__name__}")
             
-            errors_log.append({
-                "model": model_name,
-                "type": error_type,
-                "message": error_msg,
-                "trace": error_trace
-            })
+            # Check for auth errors
+            if any(x in error_str for x in ["401", "403", "Unauthorized", "Forbidden"]):
+                print("🚨 AUTHENTICATION ERROR")
+                yield (
+                    f"🔐 Authentication Problem!\n\n"
+                    f"Your token isn't working. Please:\n"
+                    f"1. Create a NEW token at https://huggingface.co/settings/tokens\n"
+                    f"2. Add it in Space Settings → Secrets as 'HF_TOKEN'\n"
+                    f"3. Restart this Space\n\n"
+                    f"Error: {error_str}"
+                )
+                return
             
             continue
     
-    # All models failed - return detailed error report
-    error_report = "❌ ALL MODELS FAILED\n\n"
+    # If chat_completion failed, try text_generation as fallback
+    print("\n" + "="*50)
+    print("⚠️ All chat models failed, trying text_generation fallback")
+    print("="*50)
     
-    for idx, err in enumerate(errors_log, 1):
-        error_report += f"Model {idx}: {err['model']}\n"
-        error_report += f"Error: {err['type']}\n"
-        error_report += f"Message: {err['message'][:200]}\n"
-        error_report += "-"*40 + "\n\n"
+    text_models = [
+        "microsoft/phi-2",
+        "google/flan-t5-large",
+        "bigscience/bloom-560m",
+    ]
+    
+    # Build a simple prompt from the conversation
+    prompt = f"{personalized_system}\n\nUser: {message}\nAssistant:"
     
-    error_report += "\n🔍 Check the Logs tab for full stack traces."
+    for model_name in text_models:
+        try:
+            print(f"🔄 Trying text_generation: {model_name}")
+            
+            result = client.text_generation(
+                prompt,
+                model=model_name,
+                max_new_tokens=max_tokens,
+                temperature=temperature,
+                top_p=top_p,
+                stream=True,
+            )
+            
+            response = ""
+            for token in result:
+                response += token
+                yield response
+            
+            print(f"✅ SUCCESS with text_generation: {model_name}")
+            return
+            
+        except Exception as e:
+            print(f"❌ Failed text_generation with {model_name}: {repr(e)}")
+            continue
     
-    return error_report
+    # Everything failed
+    yield (
+        f"I'm here for you, {user_name}, but I'm experiencing connection issues with the AI service. "
+        f"This could be due to:\n\n"
+        f"• High demand on Hugging Face servers\n"
+        f"• Models are loading (cold start)\n"
+        f"• Network connectivity issues\n\n"
+        f"Please try again in 30-60 seconds. If this persists, check the Space logs for details."
+    )
 
-# Simple interface
-demo = gr.Interface(
-    fn=respond,
-    inputs=gr.Textbox(label="Test message", value="Hello, how are you?"),
-    outputs=gr.Textbox(label="Response", lines=15),
-    title="🔍 Enhanced Diagnostic Mode",
-    description="Testing all models with detailed error logging. Check Logs tab for full details.",
+# Gradio Interface
+demo = gr.ChatInterface(
+    respond,
+    type="messages",
+    title="💙 Mariam - Mental Health Support",
+    description="A compassionate AI assistant for emotional well-being",
+    additional_inputs=[
+        gr.Textbox(
+            value=(
+                "You are Mariam, a compassionate mental health assistant. "
+                "Respond directly and naturally to the user. "
+                "Be warm, empathetic, and professional. "
+                "Never use formal letter formats."
+            ),
+            label="System Message",
+            lines=4,
+        ),
+        gr.Slider(128, 1024, value=512, label="Max Tokens", step=32),
+        gr.Slider(0.1, 1.5, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p"),
+    ],
+    theme=gr.themes.Soft(primary_hue="blue"),
+    cache_examples=False,
 )
 
 if __name__ == "__main__":