fixes for local and api based streaming

stream_app.py

@@ -1,5 +1,6 @@
 import os
 import argparse
+import logging
 import gradio as gr
 from openai import OpenAI
 import whisper
@@ -8,6 +9,18 @@ import io
 import dotenv
 dotenv.load_dotenv()
 
+
+# Set up logging
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s [%(levelname)s] %(message)s",
+    handlers=[
+        logging.StreamHandler()
+    ]
+)
+logger = logging.getLogger(__name__)
+
+
 whisper_model = None
 
 
@@ -18,11 +31,18 @@ def run_gradio(config:dict):
     # transcription of audio
     def audio_transcribe(input_audio_model:str, input_audio:str, input_text:str):
         global whisper_model
+        global logger
+
         if "offline" in input_audio_model.lower():
             if whisper_model is None:
                 whisper_model = whisper.load_model("base")
             audio = whisper.load_audio(input_audio)
             result = whisper_model.transcribe(audio)
+            result["no_speech_prob"] = 0
+            prob_scores = [x['no_speech_prob'] for x in result['segments']]
+            if len(prob_scores) > 0:    # average the probs
+                result["no_speech_prob"] = sum(prob_scores)/len(prob_scores)
+
         elif "online" in input_audio_model.lower():
             with open(input_audio, 'rb') as file_audio:
                 result = client.audio.translations.create(
@@ -32,8 +52,11 @@ def run_gradio(config:dict):
                     return ""
                 result = result.to_dict()
         prompt = result["text"]
-        print(f"Transcribe: {result}")
-        return input_text + " " + prompt
+        logger.warning(f"Transcription: {result}")
+        
+        if result["no_speech_prob"] < (1 - config['speech_threshold']):      # threshold to avoid bad output
+            return input_text + " " + prompt
+        return input_text
 
     # reset transcribed text
     def audio_reset(input_text):
@@ -46,6 +69,7 @@ def run_gradio(config:dict):
         if not prompt:
             return "Please enter a prompt for interaction."
 
+        logger.warning(f"Prompt: {prompt}")
         response = client.chat.completions.create(model=config['model'],
             stream=True,
             temperature=config['temperature'],
@@ -64,7 +88,7 @@ def run_gradio(config:dict):
             partial_response += token
             yield partial_response
 
-    with gr.Blocks() as demo:
+    with gr.Blocks(css="footer{display:none !important}") as demo:
         gr.Markdown("""
         # GPT-4 Gradio Demo
         Enter your prompt below and see the AI-generated response.
@@ -91,8 +115,7 @@ def run_gradio(config:dict):
                 output_text = gr.Textbox(
                     label="Output",
                     interactive=False,
-                    lines=5,
-                    max_lines=10,
+                    lines=10,
                 )
         submit_button = gr.Button("Submit", variant='primary')
         input_audio.stream(audio_transcribe,
@@ -100,27 +123,47 @@ def run_gradio(config:dict):
             outputs=input_text)
         input_audio.clear(audio_reset, inputs=input_text, outputs=input_text)
         input_audio.start_recording(audio_reset, inputs=input_text, outputs=input_text)
-        # input_audio.stop_recording(get_ai_response, 
-        #     inputs=[input_text, input_audio], 
-        #     outputs=output_text)
+        input_audio.stop_recording(get_ai_response, 
+            inputs=[input_text, input_audio], 
+            outputs=output_text)
         submit_button.click(get_ai_response, 
             inputs=[input_text, input_audio], 
             outputs=output_text)
 
-
+    # demo.set_api_mode(enabled=False)   # Disable API exposure
+    # demo.set_footer(enabled=False)   # Disable Gradio footers
+    
     demo.queue()
     demo.launch(share=False, debug=True, server_port=config["port"])
 
 
 def parse_args() -> dict:
     parser = argparse.ArgumentParser()
-    parser.add_argument("--model", type=str, default="gpt-4o")
-    parser.add_argument("--temperature", type=float, default=0.7)
-    parser.add_argument("--max_tokens", type=int, default=100)
-    parser.add_argument("--port", type=int, default=7860)
+    opt_group = parser.add_argument_group("Model  Configuration")
+    opt_group.add_argument("--model", type=str, default="gpt-4o",
+        help="Model to use for chat completion.")
+    opt_group.add_argument("--temperature", type=float, default=1.0,
+        help="Temperature for chat completion. ")
+    opt_group.add_argument("--max_tokens", type=int, default=2000,
+        help="Maximum number of tokens to generate in chat completion.")
     
+    opt_group = parser.add_argument_group("Speech Processing")
+    opt_group.add_argument("--speech_threshold", type=float, default=0.5,
+        help="Speech threshold for recognition to add text to a prompt. ")
+
+    opt_group = parser.add_argument_group("App Settings")
+    opt_group.add_argument("--port", type=int, default=7860, 
+        help="Port to run Gradio server on.")
+    opt_group.add_argument("--log_file", type=str, 
+        help="Path to log file to write to.  Empty will prevent any logging.")
+
     args = parser.parse_args()
-    return vars(args)
+    dict_vars = vars(args)
+    if dict_vars['log_file']:       # create new logger to output
+        logger.addHandler(
+            logging.FileHandler(dict_vars['log_file']),
+        )
+    return dict_vars
 
 
 if __name__ == "__main__":