Update app.py

app.py

@@ -1,9 +1,7 @@
 import gradio as gr
-from transformers import pipeline
+from huggingface_hub import InferenceClient
 
-# Load your model
-model_name = "Atulit23/meta-llama-indian-constitution"
-pipe = pipeline("text-generation", model=model_name)
+client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
 def respond(
     message,
@@ -13,9 +11,13 @@ def respond(
     temperature,
     top_p,
 ):
+    # Limit the number of historical messages to manage memory
+    max_history_length = 10
     messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
+    
+    # Add recent history
+    recent_history = history[-max_history_length:]  # Only keep the most recent messages
+    for val in recent_history:
         if val[0]:
             messages.append({"role": "user", "content": val[0]})
         if val[1]:
@@ -23,15 +25,38 @@ def respond(
 
     messages.append({"role": "user", "content": message})
 
-    response = pipe(messages[-1]["content"], max_length=max_tokens, temperature=temperature, top_p=top_p)
-    return response[0]['generated_text']
+    response = ""
+
+    try:
+        for response_chunk in client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            stream=True,
+            temperature=temperature,
+            top_p=top_p,
+        ):
+            token = response_chunk.choices[0].delta.content
+            response += token
+
+        # Implement a basic check for relevance
+        if not is_constitution_related(response):
+            response = "Sorry, I can only answer questions related to the Constitution of India."
+
+        yield response
+
+    except MemoryError:
+        yield "Error: Memory limit exceeded. Please try again later."
+
+def is_constitution_related(response):
+    # Perform a simple check to see if the response seems related to the Constitution
+    # This can be improved based on specific needs and feedback
+    return "constitution" in response.lower() or "article" in response.lower()
 
-# Create a Gradio ChatInterface
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot focused on the Indian Constitution.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=50, step=1, label="Max new tokens"),
+        gr.Textbox(value="You are a knowledgeable assistant specializing in the Constitution of India. Only provide answers related to the Constitution. If the question is not related, inform the user accordingly.", label="System message"),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(
             minimum=0.1,
@@ -41,8 +66,6 @@ demo = gr.ChatInterface(
             label="Top-p (nucleus sampling)",
         ),
     ],
-    title="Constitutional Chatbot",
-    description="Ask questions related to the Indian Constitution."
 )
 
 if __name__ == "__main__":