Update app.py

app.py

@@ -1,29 +1,28 @@
 import gradio as gr
+from llama_cpp import Llama
 import os
-import requests
-from cachetools import TTLCache
+from huggingface_hub import hf_hub_download
 
 # Configuration
-API_URL = "https://api.mistral.ai/v1/chat/completions"
-API_TOKEN = os.environ.get("HF_TOKEN")  # Get from environment # Replace with your token
+MODEL_REPO = "TheBloke/Mistral-7B-Instruct-v0.3-GGUF"
+MODEL_FILE = "mistral-7b-instruct-v0.3.Q4_K_M.gguf"
 MAX_TOKENS = 150
-CACHE_TTL = 3600  # 1 hour cache
+CPU_THREADS = os.cpu_count() or 4  # Use all available cores
 
-#Validate API key
-if not API_TOKEN:
-    raise gr.Error("API_KEY environment variable not set!")
-    
-# Initialize response cache
-response_cache = TTLCache(maxsize=100, ttl=CACHE_TTL)
-
-#theme
-SPIRITUAL_THEME = gr.themes.Default(
-    primary_hue="teal",     # Closest to emerald
-    secondary_hue="yellow", # Closest to gold
-    font=[gr.themes.GoogleFont("Hind Siliguri")]
+# Download model (will cache after first run)
+model_path = hf_hub_download(
+    repo_id=MODEL_REPO,
+    filename=MODEL_FILE,
+    cache_dir="models"
 )
 
-    
+# Initialize Llama
+llm = Llama(
+    model_path=model_path,
+    n_ctx=2048,
+    n_threads=CPU_THREADS,
+    n_gpu_layers=0  # CPU only
+)
 
 def format_krishna_prompt(message, history):
     """Create proper Mistral instruction prompt with Krishna context"""
@@ -35,88 +34,46 @@ Maintain a compassionate, all-knowing tone.
     
     conversation = []
     for human, assistant in history:
-        conversation.append(f"{human}[/INST] {assistant}<s>")
-    return f"{system_prompt}{''.join(conversation)}[INST] {message} [/INST]"
+        conversation.append(f"{human}[/INST] {assistant}</s>")
+    return f"{system_prompt}{''.join(conversation)}<s>[INST] {message} [/INST]"
 
 def query_krishna(message, history):
-    """Get response from Mistral API with streaming"""
-    # Check cache first
-    cache_key = (message, tuple(history))
-    if cache_key in response_cache:
-        yield response_cache[cache_key]
-        return
-
-    # Build API request
-    headers = {"Authorization": f"Bearer {API_TOKEN}"}
-    formatted_prompt = format_krishna_prompt(message, history)
-    
+    """Get response from Mistral 7B on CPU"""
     try:
-        response = requests.post(
-            API_URL,
-            headers=headers,
-            json={
-                "inputs": formatted_prompt,
-                "parameters": {
-                    "max_new_tokens": MAX_TOKENS,
-                    "temperature": 0.7,
-                    "top_p": 0.9,
-                    "return_full_text": False
-                }
-            },
-            stream=True
+        prompt = format_krishna_prompt(message, history)
+        output = llm(
+            prompt,
+            max_tokens=MAX_TOKENS,
+            temperature=0.7,
+            top_p=0.9,
+            stop=["</s>"]
         )
-        response.raise_for_status()
-        
-        # Stream response
-        full_response = ""
-        for chunk in response.iter_content(chunk_size=128):
-            if chunk:
-                full_response += chunk.decode("utf-8")
-                yield full_response.split("[/INST]")[-1].strip()
-        
-        # Cache final response
-        response_cache[cache_key] = full_response.split("[/INST]")[-1].strip()
-
+        return output['choices'][0]['text'].strip()
     except Exception as e:
-        yield f"🙏 The divine connection is weak. Please try again. ({str(e)})"
+        return f"🙏 Divine wisdom temporarily obscured: {str(e)}"
 
 # Custom UI components
-krishna_avatar = "krishna.jpg"  # Local file path
+krishna_avatar = "krishna.jpg"  # Make sure this file is in your Space repo
 
-with gr.Blocks(theme=SPIRITUAL_THEME, title="Divine Dialogue with Lord Krishna") as demo:
+with gr.Blocks(title="Divine Dialogue with Lord Krishna") as demo:
     gr.Markdown("""
-    # 🕉️ Bhagavad Gita Wisdom Chatbot
-    **Ask questions to Shree Krishna** - Embodiment of Supreme Truth and Eternal Bliss
+    # 🕉️ Bhagavad Gita Wisdom Chatbot (CPU Edition)
+    **Ask questions to Shree Krishna** - Running Mistral 7B Instruct on CPU
     """)
     
     with gr.Row():
         gr.Image(krishna_avatar, height=200, show_label=False)
     
-    chatbot = gr.ChatInterface(
+    gr.ChatInterface(
         query_krishna,
-        chatbot=gr.Chatbot(
-            height=500,
-            type="messages"
-        ),
-        textbox=gr.Textbox(
-            placeholder="Ask your spiritual question...",
-            container=False,
-            scale=7
-        ),
+        chatbot=gr.Chatbot(height=500),
         examples=[
             "What is the purpose of life?",
             "How to achieve inner peace?",
             "Explain karma yoga from Bhagavad Gita"
-        ],
-        theme="compact"
+        ]
     )
-    
-    # Footer
-    gr.Markdown("""
-    ---
-    *Built with 💙 using [Mistral-7B-Instruct](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3) and [Gradio](https://gradio.app)*
-    """)
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860)
+    demo.launch()