Update app.py

app.py

@@ -1,114 +1,119 @@
-# --- 🔱 Hanuman AI: Supreme English Reasoning Portal (Qwen 2.5) ---
+# --- 🔱 Hanuman AI: Google Gemma-2 Coding & Reasoning Portal ---
 # Pioneered by Divy Patel | Bharat 🇮🇳
-# Features: 100% English, Fast Inference, Free CPU Optimized, Vision Enabled
+# Features: Gated Access Fixed, Persistent History, 100% English
 
 import gradio as gr
 import torch
-from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
-from qwen_vl_utils import process_vision_info
+from transformers import AutoTokenizer, AutoModelForCausalLM, TextIteratorStreamer
+from threading import Thread
 import os
 
-# Model Upgrade: Qwen 2.5 VL 3B (Powerful & Reasoning-focused)
-model_id = "Qwen/Qwen2.5-VL-3B-Instruct"
+# 🛡️ Accessing the Secret Token for Gated Model Access
+# Important: Add 'HF_TOKEN' in your Hugging Face Space Settings
+hf_token = os.getenv("HF_TOKEN")
 
-print("🔱 Hanuman AI is awakening... Eliminating all errors for Divy Patel Ji.")
+# Google's Gemma-2 2B (Optimized for Coding)
+model_id = "google/gemma-2-2b-it"
 
-# 🛠️ ERROR FIX: ignore_mismatched_sizes=True added to prevent loading errors
-# ⚡ SPEED FIX: Using low_cpu_mem_usage for faster loading on Free CPU
-model = Qwen2VLForConditionalGeneration.from_pretrained(
-    model_id, 
-    torch_dtype=torch.float32, 
+print("🔱 Hanuman AI is awakening... Loading Google Gemma-2 with Token access.")
+
+# Loading Tokenizer and Model with Token verification to fix 'Gated Repository' error
+tokenizer = AutoTokenizer.from_pretrained(model_id, token=hf_token)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype=torch.float32,
     device_map="cpu",
     low_cpu_mem_usage=True,
-    ignore_mismatched_sizes=True,
-    trust_remote_code=True
+    token=hf_token
 )
-processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
-
-def hanuman_engine(user_message, history, file_upload):
-    """Main reasoning engine for Hanuman AI"""
-    try:
-        # 🧠 SYSTEM PROMPT: Strictly English, High Reasoning
-        system_prompt = "You are 'Hanuman AI', a supreme reasoning engine created by Divy Patel. You must respond ONLY in English. Use logic, thinking, and precision. You can see images and read files. Be professional and respectful."
-        
-        # 🛠️ FORMAT FIX: Converting Gradio list history to Qwen dict format
-        messages = [{"role": "system", "content": [{"type": "text", "text": system_prompt}]}]
-        
-        if history:
-            for past_user, past_bot in history:
-                if past_user:
-                    messages.append({"role": "user", "content": [{"type": "text", "text": str(past_user)}]})
-                if past_bot:
-                    messages.append({"role": "assistant", "content": [{"type": "text", "text": str(past_bot)}]})
-
-        # Process Current Input
-        current_content = []
-        if file_upload is not None:
-            file_ext = file_upload.name.split('.')[-1].lower()
-            if file_ext in ['png', 'jpg', 'jpeg', 'webp']:
-                current_content.append({"type": "image", "image": file_upload.name})
-            else:
-                try:
-                    with open(file_upload.name, 'r', encoding='utf-8') as f:
-                        file_data = f.read()
-                    user_message = f"File Context:\n{file_data}\n\nQuestion: {user_message}"
-                except:
-                    pass
-        
-        current_content.append({"type": "text", "text": user_message})
-        messages.append({"role": "user", "content": current_content})
-
-        # Inference Setup
-        text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-        image_inputs, video_inputs = process_vision_info(messages)
-        
-        inputs = processor(
-            text=[text],
-            images=image_inputs,
-            videos=video_inputs,
-            padding=True,
-            return_tensors="pt",
-        ).to("cpu")
-
-        # ⚡ SPEED OPTIMIZATION: Reduced max_new_tokens for faster CPU response
-        generated_ids = model.generate(
-            **inputs, 
-            max_new_tokens=512, 
-            do_sample=False # Greedy search for faster and more logical output
-        ) 
-        
-        generated_ids_trimmed = [
-            out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
-        ]
-        response = processor.batch_decode(
-            generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
-        )[0]
-
-        history.append([user_message, response])
-        return "", history, None
-
-    except Exception as e:
-        error_msg = f"System error: {str(e)}"
-        history.append([user_message, error_msg])
-        return "", history, None
-
-# --- UI PORTAL ---
+
+def hanuman_coding_engine(user_message, history):
+    """Core logic using Google Gemma-2 for English responses with history"""
+    
+    # Strictly English System Prompt
+    system_prompt = (
+        "You are 'Hanuman AI', a high-performance coding and reasoning engine pioneered by Divy Patel. "
+        "Your responses must be strictly in English. Focus on code quality and logical precision."
+    )
+
+    # 🛠️ Formatting history for Gemma-2 Template
+    full_prompt = f"<bos><start_of_turn>system\n{system_prompt}<end_of_turn>\n"
+    
+    # Adding past history to the current prompt so AI remembers previous messages
+    for past_user, past_bot in history:
+        if past_user:
+            full_prompt += f"<start_of_turn>user\n{past_user}<end_of_turn>\n"
+        if past_bot:
+            full_prompt += f"<start_of_turn>model\n{past_bot}<end_of_turn>\n"
+    
+    # Adding current user message
+    full_prompt += f"<start_of_turn>user\n{user_message}<end_of_turn>\n<start_of_turn>model\n"
+
+    # Tokenizing input for CPU
+    inputs = tokenizer(full_prompt, return_tensors="pt").to("cpu")
+    streamer = TextIteratorStreamer(tokenizer, timeout=20.0, skip_prompt=True, skip_special_tokens=True)
+
+    # Generation parameters
+    generate_kwargs = dict(
+        inputs,
+        streamer=streamer,
+        max_new_tokens=1024,
+        do_sample=False, 
+        repetition_penalty=1.2
+    )
+
+    # Threading for real-time streaming response
+    thread = Thread(target=model.generate, kwargs=generate_kwargs)
+    thread.start()
+
+    partial_text = ""
+    for new_text in streamer:
+        partial_text += new_text
+        yield partial_text
+
+def chat_logic(message, history):
+    """Ensures that the history is preserved and visible to Divy Patel Ji"""
+    # Initialize the bot response in history
+    history = history or []
+    
+    # Create an empty slot for the bot's response in history
+    history.append([message, ""])
+    
+    # Stream the response and update history
+    for chunk in hanuman_coding_engine(message, history[:-1]):
+        history[-1][1] = chunk
+        yield "", history
+
+# --- 🔱 UI PORTAL ---
 with gr.Blocks() as demo:
-    gr.HTML("<div style='text-align: center;'><h1>🔱 Hanuman AI - English Reasoning Portal</h1><p>Pioneered by Divy Patel | 100% Free CPU | Bharat 🇮🇳</p></div>")
+    gr.HTML("<div style='text-align: center; color: #ff5500;'><h1>🔱 Hanuman AI - Google Gemma 2.0</h1><p><b>Pioneered by Divy Patel | Secure Gated Access | Bharat 🇮🇳</b></p></div>")
+    
+    chatbot = gr.Chatbot(height=550, label="Chat History")
     
     with gr.Row():
-        with gr.Column(scale=4):
-            chatbot = gr.Chatbot(height=550)
-            with gr.Row():
-                msg_input = gr.Textbox(placeholder="Ask your complex question here...", scale=7)
-                file_input = gr.File(label="Upload Image/File", scale=2)
-                submit_btn = gr.Button("ASK", variant="primary", scale=1)
+        msg_input = gr.Textbox(
+            placeholder="Type your coding query and press Enter...", 
+            label="Your Message",
+            scale=8
+        )
+        submit_btn = gr.Button("ASK", variant="primary", scale=2)
 
+    # 🛡️ Fixing History Visibility: Updating chatbot with streamed history
     submit_btn.click(
-        fn=hanuman_engine, 
-        inputs=[msg_input, chatbot, file_input], 
-        outputs=[msg_input, chatbot, file_input]
+        fn=chat_logic, 
+        inputs=[msg_input, chatbot], 
+        outputs=[msg_input, chatbot]
     )
+    
+    msg_input.submit(
+        fn=chat_logic, 
+        inputs=[msg_input, chatbot], 
+        outputs=[msg_input, chatbot]
+    )
+
+    gr.Markdown("---")
+    gr.Markdown("### 🔱 Note for Divy Patel Ji:")
+    gr.Markdown("1. Ensure **HF_TOKEN** is added in your Space Secrets.\n2. This model runs locally on CPU - 100% Free and Private.")
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()