Update app.py

app.py

@@ -1,54 +1,66 @@
 import os
+import torch
 import gradio as ui
-from huggingface_hub import InferenceClient
+from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
 
-# Target model ID definition
 MODEL_ID = "coder-vansh/cypher_llm_model"
-
-# Fetch secure token from space settings automatically
 HF_TOKEN = os.getenv("HF_TOKEN")
 
-# Initialize the official Hugging Face native Inference Engine
-client = InferenceClient(model=MODEL_ID, token=HF_TOKEN)
+print("⚙️ System: Initializing 4-bit Quantization engine to save Cloud RAM...")
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.float16,
+    bnb_4bit_use_double_quant=True
+)
+
+print("📥 Loading Model weights natively into memory space...")
+# Load tokenizer and model weights inside the local basic CPU instance
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, token=HF_TOKEN)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    quantization_config=bnb_config,
+    low_cpu_mem_usage=True,
+    token=HF_TOKEN
+)
+print("✅ CYPHER Brain is now natively alive inside the Space environment!")
 
 def predict(message, history):
-    # Perfect prompt engineering template layout for CYPHER format parsing
     system_prompt = "You are CYPHER, a casual, witty Hinglish AI companion by Vansh & Aditya. Tagline: \"Not just an AI — YOUR AI.\""
     
-    # Building native messaging map structure
+    # Building exact chat block matching schema
     messages = [{"role": "system", "content": system_prompt}]
-    
-    # Restoring past history blocks if any exist
     if history:
         for user_msg, bot_msg in history:
             messages.append({"role": "user", "content": user_msg})
             messages.append({"role": "assistant", "content": bot_msg})
             
-    # Appending the latest query
     messages.append({"role": "user", "content": message})
 
     try:
-        # Calling native client pipeline to automatically handle DNS mapping
-        response = client.chat_completion(
-            messages=messages,
-            max_tokens=250,
-            temperature=0.7,
-            stream=False
+        inputs = tokenizer.apply_chat_template(
+            messages, 
+            tokenize=True, 
+            add_generation_prompt=True, 
+            return_tensors="pt"
         )
         
-        # Safe extraction of response token string
-        return response.choices[0].message.content
+        with torch.no_grad():
+            outputs = model.generate(
+                input_ids=inputs,
+                max_new_tokens=250,
+                temperature=0.7,
+                do_sample=True,
+                pad_token_id=tokenizer.eos_token_id
+            )
+            
+        response = tokenizer.decode(outputs[0][inputs.shape[1]:], skip_special_tokens=True)
+        return response
 
     except Exception as e:
-        error_msg = str(e)
-        if "503" in error_msg or "loading" in error_msg.lower():
-            return "⏳ CYPHER abhi cloud network me so raha hai... 1 minute me weights spin-up ho rahe hain, ek baar dubara message bhejo bhai!"
-        elif "401" in error_msg or "403" in error_msg:
-            return "🔑 Token Issue: Bhai lagta hai tumne Space Settings me 'HF_TOKEN' naam ka Secret sahi se add nahi kiya hai!"
-        else:
-            return f"📡 Network glitch trace: {error_msg}\n\n(Bhai agar Model loading status hai toh 1 min wait karke retry karo!)"
-
-# Elegant web layout container
+        return f"⚠️ Execution Glitch: {str(e)}"
+
+# Elegant interface container
 with ui.Blocks() as demo:
     ui.Markdown("# 🤖 CYPHER AI Live Production Space")
     ui.Markdown("### *Not just an AI — YOUR AI.* | Developed by Vansh & Aditya")
@@ -56,8 +68,7 @@ with ui.Blocks() as demo:
     ui.ChatInterface(
         fn=predict,
         textbox=ui.Textbox(placeholder="Bhai se kuch bhi poocho...", container=False, scale=7),
-        description="CYPHER is running smoothly via Official Hugging Face Native Hub Core.",
+        description="CYPHER is running 100% Natively inside your cloud container.",
     )
 
-# Launching web application frame
 demo.launch()