Update app.py

app.py

@@ -1,59 +1,52 @@
-import os
-import re
-import gradio as gr
 import edge_tts
-import asyncio
-import time
 import tempfile
+import gradio as gr
 from huggingface_hub import InferenceClient
-from deep_translator import GoogleTranslator
-
-def get_translation(context_custom, language = 'en'):
-    translated = GoogleTranslator(source = 'auto', target = language).translate(text = context_custom)
-    return translated
-
-client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1",token=os.getenv("TOKEN"))
 
-system_instructions1 = "<s>[SYSTEM] You are AI assistant named DorjGPT, Keep conversation very short, clear, friendly and concise. The text provided is a request for a specific type of response from you, the virtual assistant. The request asks you to provide friendly responses as if super interlligent AI assistant. The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, your task is to generate responses that appear as if they are created by the user and not an AI-powered assistant. [USER]"
+"""
+For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
+"""
+client = InferenceClient("google/gemma-3-27b-it")
 
 global history
 history = []
-def format_prompt(message, history):
-    prompt = system_instructions1
-    for user_prompt, bot_response in history:
-        prompt += f"[INST] {user_prompt} [/INST]"
-        prompt += f" {bot_response}</s> "
-    prompt += f"[INST] {message} [/INST]"
-    return prompt
+async def respond(
+    message, 
+    history=[],
+    system_message="You are a Gemma, created by Google. You is a helpful assistant and always reply back in Mongolian, and only return Mongolian text.",
+    max_tokens=512,
+    temperature=0.001,
+    top_p=0.95,
+):
+    messages = [{"role": "system", "content": system_message}]
 
-async def generate1(prompt, history=[], b=None):
-    generate_kwargs = dict(
-        temperature=0.6,
-        max_new_tokens=256,
-        top_p=0.95,
-        repetition_penalty=1,
-        do_sample=True,
-        seed=42,
-    )
-    #formatted_prompt = system_instructions1 + prompt + "[JARVIS]"
-    prompt_en = get_translation(prompt)
-    formatted_prompt = format_prompt(f"{system_instructions1}, {prompt_en}", history) + "[DORJGPT]"
-    stream = client.text_generation(
-        formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
+    for val in history:
+        if val[0]:
+            messages.append({"role": "user", "content": val[0]})
+        if val[1]:
+            messages.append({"role": "assistant", "content": val[1]})
 
-    output = ""
-    for response in stream:
-        output += response.token.text
-    output = output.replace("</s>","")
-    history.append([prompt, output])
-    output_mn = get_translation(output, language="mn")
-    
-    communicate = edge_tts.Communicate(output_mn, voice="mn-MN-BataaNeural")
+    messages.append({"role": "user", "content": message})
+
+    response = ""
+
+    for message in client.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
+        temperature=temperature,
+        top_p=top_p,
+    ):
+        token = message.choices[0].delta.content
+        response += token
+        
+    communicate = edge_tts.Communicate(response, voice="mn-MN-BataaNeural")
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
         tmp_path = tmp_file.name
         await communicate.save(tmp_path)
     yield tmp_path
 
+
 with gr.Blocks(theme="gradio/monochrome", title="Dorj Assistant") as demo:
     gr.HTML("""
         <h1 style="text-align: center; style="font-size: 3m;">
@@ -72,7 +65,7 @@ with gr.Blocks(theme="gradio/monochrome", title="Dorj Assistant") as demo:
     with gr.Tab():
       with gr.Row():
         translate_btn = gr.Button("Submit")
-        translate_btn.click(fn=generate1, inputs=user_input,
+        translate_btn.click(fn=respond, inputs=user_input,
                             outputs=output_audio, api_name="translate")  
 
 if __name__ == "__main__":