Update app.py

app.py

@@ -1,47 +1,50 @@
 import gradio as gr
-from langchain.llms import HuggingFacePipeline
-from langchain import LLMChain, PromptTemplate
-from langchain.memory import ConversationBufferMemory
-from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+from openai import OpenAI
 
-# Load a free model from Hugging Face
-model_name = "microsoft/DialoGPT-medium"  # Or try "tiiuae/falcon-rw-1b" or "gpt2"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name)
-
-# Create pipeline
-pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_length=1000, do_sample=True)
-
-# Wrap with HuggingFacePipeline
-llm = HuggingFacePipeline(pipeline=pipe)
-
-template = """You are a helpful assistant to answer user queries.
-{chat_history}
-User: {user_message}
-Chatbot:"""
-
-prompt = PromptTemplate(
-    input_variables=["chat_history", "user_message"], template=template
+# NVIDIA-compatible OpenAI client
+client = OpenAI(
+    base_url="https://integrate.api.nvidia.com/v1",
+    api_key="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"
 )
 
-memory = ConversationBufferMemory(memory_key="chat_history")
-
-llm_chain = LLMChain(
-    llm=llm,
-    prompt=prompt,
-    verbose=True,
-    memory=memory,
-)
+# System message
+system_prompt = {
+    "role": "system",
+    "content": "You are a helpful assistant to answer user queries."
+}
 
+# Main chat function with memory from Gradio (OpenAI-style history)
 def get_text_response(user_message, history):
-    response = llm_chain.predict(user_message=user_message)
+    # Convert Gradio message history (OpenAI format) + new user message
+    messages = [system_prompt] + history + [{"role": "user", "content": user_message}]
+
+    # Stream response
+    response = ""
+    completion = client.chat.completions.create(
+        model="nvidia/llama-3.1-nemotron-70b-instruct",
+        messages=messages,
+        temperature=0.5,
+        top_p=1,
+        max_tokens=1024,
+        stream=True
+    )
+
+    for chunk in completion:
+        delta = chunk.choices[0].delta
+        if delta and delta.content:
+            response += delta.content
+
     return response
 
+# Gradio Chat UI
 demo = gr.ChatInterface(
-    get_text_response,
+    fn=get_text_response,
+    title="🧠 Nemotron 70B Assistant",
+    theme="soft",
+    chatbot=gr.Chatbot(height=400, type="messages"),  # <-- important: type="messages"
+    textbox=gr.Textbox(placeholder="Ask me anything...", container=False),
     examples=["How are you doing?", "What are your interests?", "Which places do you like to visit?"]
 )
 
-
 if __name__ == "__main__":
     demo.queue().launch(share=True, debug=True)