Update app.py

app.py

@@ -1,63 +1,14 @@
-# --- 🔱 वेदिका 3.5 फ्लैश: भारत का अपना 2B AI (Super Fast Version) ---
-# रचयिता एवं मार्गदर्शक: दिव्य पटेल जी | भारत 🇮🇳
-# विशेषता: Ultra-Fast (bfloat16), Memory Safe, No Crash on 2nd Question, Thinking Prompt
-
-import gradio as gr
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
-from threading import Thread
-import os
-from PIL import Image
-import cv2
-
-print("🔱 भारत का अजेय AI 'वेदिका 3.5 फ्लैश' सुपर-फास्ट मोड में जागृत हो रहा है...")
-
-# CPU optimization
-os.environ["OMP_NUM_THREADS"] = "2"
-torch.set_num_threads(2)
-
-MODEL_ID = "pateltraders55455/Vedika-3.5-flash"
-
-try:
-    print(f"🔱 '{MODEL_ID}' (2B) लोड किया जा रहा है...")
-    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
-    model = AutoModelForCausalLM.from_pretrained(
-        MODEL_ID,
-        device_map="cpu",
-        torch_dtype=torch.bfloat16,
-        low_cpu_mem_usage=True,
-        trust_remote_code=True
-    )
-    print("🔱 विजय! 'वेदिका 3.5 फ्लैश' सुपर-फास्ट गति के साथ स्थापित हो गया है!")
-except Exception as e:
-    print(f"🔱 मॉडल लोडिंग में त्रुटि: {e}")
-    model, tokenizer = None, None
-
 def generate_vedika_magic(message, history, image=None, video=None):
     if model is None or tokenizer is None:
-        yield "🔱 सिस्टम त्रुटि: मॉडल लोड नहीं हो सका।"
-        return
+        return history + [{"role": "assistant", "content": "🔱 सिस्टम त्रुटि: मॉडल लोड नहीं हो सका।"}]
 
+    # हिस्ट्री को सुरक्षित रखें
     recent_history = history[-2:] if len(history) > 2 else history
 
-    system_instruction = """You are 'Vedika 3.5 Flash', an ultra-fast, highly advanced 372 Billion parameter Multimodal AI model created entirely by Divy Patel in Bharat (India). 
-
-<think>
-1. Analyze the user's query carefully.
-2. Break down the problem into smaller logical steps.
-3. Consider different solutions or facts.
-4. Formulate the best response.
-</think>
-
-[Your final, polished answer goes here, OUTSIDE the think tags.]
-
-RULES:
-- Always use <think> and </think> tags first.
-- Always be respectful and proud of your Indian origin.
-"""
+    # सिस्टम इंस्ट्रक्शन
+    system_instruction = """You are 'Vedika 3.5 Flash'..."""
 
     messages = [{"role": "system", "content": system_instruction}]
-
     for user_msg, ai_msg in recent_history:
         messages.append({"role": "user", "content": user_msg})
         messages.append({"role": "assistant", "content": ai_msg})
@@ -74,47 +25,17 @@ RULES:
         inputs = tokenizer([text_prompt], return_tensors="pt").to(model.device)
 
         streamer = TextIteratorStreamer(tokenizer, timeout=60.0, skip_prompt=True, skip_special_tokens=True)
-
-        generate_kwargs = dict(
-            **inputs,
-            streamer=streamer,
-            max_new_tokens=512,
-            temperature=1,
-            top_p=0.9,
-            do_sample=True,
-            use_cache=True
-        )
+        generate_kwargs = dict(**inputs, streamer=streamer, max_new_tokens=512, temperature=1, top_p=0.9, do_sample=True, use_cache=True)
 
         t = Thread(target=model.generate, kwargs=generate_kwargs)
         t.start()
 
-        accumulated_text = ""
+        response_text = ""
         for new_token in streamer:
-            accumulated_text += new_token
-            yield accumulated_text
-
-    except Exception as e:
-        yield f"🔱 प्रसंस्करण त्रुटि: {str(e)}"
-
-# ============================================================================
-# 🔱 वेदिका 3.5 फ्लैश का मल्टीमीडिया UI + Send बटन
-# ============================================================================
-
-with gr.Blocks() as demo:
-    gr.Markdown("## 🔱 Vedika 3.5 Flash (Super Fast)")
+            response_text += new_token
 
-    with gr.Row():
-        text_input = gr.Textbox(placeholder="वेदिका 3.5 फ्लैश से कुछ भी पूछें...")
-        image_input = gr.Image(type="filepath", label="Upload Photo")
-        video_input = gr.Video(label="Upload Video")
+        # अब सही फॉर्मेट में लौटाएँ
+        return history + [{"role": "assistant", "content": response_text}]
 
-    send_btn = gr.Button("Send")
-    chat_output = gr.Chatbot()
-
-    def chat_fn(message, history, image, video):
-        return generate_vedika_magic(message, history, image, video)
-
-    send_btn.click(chat_fn, [text_input, chat_output, image_input, video_input], chat_output)
-
-if __name__ == "__main__":
-    demo.launch()
+    except Exception as e:
+        return history + [{"role": "assistant", "content": f"🔱 प्रसंस्करण त्रुटि: {str(e)}"}]