Update app.py

app.py

@@ -3,21 +3,20 @@ from huggingface_hub import InferenceClient
 import os
 
 # ==============================================================================
-# 🚀 NITHIN-1: SERVERLESS AGI (Runs on Cloud GPUs via API)
+# 🚀 NITHIN-1: SERVERLESS AGI (Fixed & Updated)
 # ==============================================================================
 
 # 1. SETUP CLIENTS
-# Manam Token ni 'Secrets' nundi automatic ga teesukuntam
-# Make sure you added HF_TOKEN in Settings!
 token = os.getenv("HF_TOKEN")
 
-# TEXT BRAIN: Using 'Zephyr-7b' (It's like Mistral but faster for chat)
+# TEXT BRAIN: Switched to Mistral-7B-Instruct-v0.3 (Latest & Best Free Model)
+# This model supports the "Chat" format perfectly.
 text_client = InferenceClient(
-    "HuggingFaceH4/zephyr-7b-beta",
+    "mistralai/Mistral-7B-Instruct-v0.3",
     token=token
 )
 
-# IMAGE EYES: Using 'Stable Diffusion XL' (Best Free Image Model)
+# IMAGE EYES: Stable Diffusion XL
 image_client = InferenceClient(
     "stabilityai/stable-diffusion-xl-base-1.0", 
     token=token
@@ -25,33 +24,40 @@ image_client = InferenceClient(
 
 # 2. THE BRAIN LOGIC
 def nithin_agi(message, history):
-    # A. IMAGE DETECTION (Simple Logic)
+    # A. IMAGE DETECTION
     triggers = ["draw", "generate", "image", "photo", "paint", "picture"]
     if any(word in message.lower() for word in triggers):
-        return "🎨 I am generating your image... Check the 'Image Generator' tab!"
+        return "🎨 I am generating your image... Please check the 'Image Engine' tab!"
     
-    # B. TEXT GENERATION (Streaming)
-    # Idhi Google GPUs meedha run avuthundi, so CPU slow unna parvaledu.
+    # B. TEXT GENERATION (Conversational Mode)
     system_prompt = "You are Nithin-1, an advanced AI created by Nithin. Be helpful, smart, and concise."
     
-    # Formatting prompt for Zephyr model
-    prompt = f"<|system|>\n{system_prompt}</s>\n<|user|>\n{message}</s>\n<|assistant|>"
+    # New Logic: Using 'chat_completion' instead of 'text_generation'
+    messages = [
+        {"role": "system", "content": system_prompt},
+        {"role": "user", "content": message}
+    ]
     
     partial_message = ""
     try:
-        # Stream response (Letter by letter style)
-        for token in text_client.text_generation(prompt, max_new_tokens=512, stream=True):
-            partial_message += token
-            yield partial_message
+        # Streaming the response correctly
+        stream = text_client.chat_completion(messages, max_tokens=512, stream=True)
+        
+        for chunk in stream:
+            # Extracting content from the new format
+            if chunk.choices and chunk.choices[0].delta.content:
+                content = chunk.choices[0].delta.content
+                partial_message += content
+                yield partial_message
+                
     except Exception as e:
-        yield f"⚠️ Error: Please check your HF_TOKEN in settings. Details: {e}"
+        yield f"⚠️ Brain Error: {str(e)}\n\n(Tip: Check if your HF_TOKEN is valid in Settings)"
 
 # 3. THE IMAGE LOGIC
 def generate_image(prompt):
     if not token:
         return None
     try:
-        # Calling API for Image
         image = image_client.text_to_image(prompt + ", high quality, 8k, masterpiece")
         return image
     except Exception as e:
@@ -59,7 +65,7 @@ def generate_image(prompt):
         return None
 
 # ==============================================================================
-# 🎨 THE UI (Professional Look)
+# 🎨 THE UI
 # ==============================================================================
 
 with gr.Blocks(theme="soft") as demo: