Update app.py

app.py

@@ -1,52 +1,98 @@
-import os, time, wave
+import time
 import openai
 import gradio as gr
 import requests
 from pydub import AudioSegment as am
 from xml.etree import ElementTree
 
-aoai_url = "https://mvp-azureopenai.openai.azure.com/"
-aoai_key = ""
-stts_key = ""
+aoai_url, aoai_key, stts_key, stts_region = "", "", "", ""
 
 openai.api_type = "azure"
-openai.api_base = "https://mvp-azureopenai.openai.azure.com/"
+# openai.api_base = "https://mvp-azureopenai.openai.azure.com/"
 openai.api_version = "2023-03-15-preview"
 # openai.api_key = aoai_key
 
+prompts = ""
+model_gpt = ""
 messages_gpt = []
+model_chat = ""
 messages_chat = [
     {"role": "system", "content": "You are an AI assistant that helps people find information."},
 ]
-prompts = ""
 response_walle = []
+model_vchat = ""
 messages_vchat = [
-    {"role": "system", "content": "You are an AI assistant that helps people find information and just response with SSML."},
+    {"role": "system", "content": "You are an AI assistant that helps people find information and just respond with SSML."},
 ]
 
-def get_aoai_key(txt_inpout):
-    if txt_inpout:
-        openai.api_key = txt_inpout
-    return gr.update(value=txt_inpout)
+def get_aoai_set(get_aoai_url, get_aoai_key, get_aoai_API):
+    if get_aoai_url:
+        openai.api_base = get_aoai_url
+    if get_aoai_key:
+        openai.api_key = get_aoai_key
+    if get_aoai_API:
+        openai.api_version = get_aoai_API
+    return gr.update(value=get_aoai_url), gr.update(value=get_aoai_key), gr.update(value=get_aoai_API)
+
+def get_stts_set(get_stts_key, get_stts_region):
+    global stts_key, stts_region
+    if get_stts_key:
+        stts_key = get_stts_key
+    if get_stts_region:
+        stts_region = get_stts_region
+    return gr.update(value=get_stts_key), gr.update(value=get_stts_region)
 
 with gr.Blocks() as page:
-    with gr.Row():
-        with gr.Column(scale=0.4):
-            gr.Markdown("Your Key will not be saved or viewed by anyone.")
-        with gr.Column(scale=0.6):
-            ui_aoai_key = gr.Textbox(placeholder="Please enter your Azure OpenAI API key here.", lines=1, show_label=False, type='password').style(container=False)
-            ui_aoai_key.submit(get_aoai_key, ui_aoai_key, ui_aoai_key)
-        
     with gr.Tabs():
-        with gr.TabItem("GPT Playground"):
+        with gr.TabItem("Settings"):
+            gr.HTML("""
+                <p>Please read guide and set parameters before switch to other tab.</p> <br>Your Azure OpenAI Key and other Azure Cognitive Service Keys
+                 will not be saved or viewed by anyone. <br> You need input parameters to use Azure OpenAI Service. Please input these settings and hit 
+                 'Enter' key. <br><br>
+                 You can find these parameters in Azure Portal. Select Azure OpenAI resource or Cognitive Service resource like Speech, and then select 
+                 'Keys and Endpoint' from left panel. <br> For Azure OpenAI service, you need to provide the resource URL and key for REST API. You also 
+                 need to set the API version or just use the default value. The Azure OpenAI model which be deployment need to set in each tab. Because 
+                 you may need run different models at same time. Don't forget hit 'Enter' by every input. <br> For Azure Cognitive services, you need to 
+                 provide Key for REST API, and also need to provide service region, for example like 'westus'. App will create the endpoint URL by itself.
+                 <br><br>
+                 Thank you.
+                 """)
+            with gr.Row():
+                with gr.Column(scale=0.6):
+                    ui_aoai_url = gr.Textbox(placeholder="Like https://your-url-base.openai.azure.com , etc.", 
+                                             label="- Azure OpenAI service API endpoint:", lines=1).style(container=False)
+                with gr.Column(scale=0.2):
+                    ui_aoai_key = gr.Textbox(placeholder="Please enter your Azure OpenAI API key here.", 
+                                             label="- Azure OpenAI service API Key: ", lines=1, type='password').style(container=False)
+                with gr.Column(scale=0.2):
+                    ui_aoai_api = gr.Textbox(value="2023-03-15-preview", label="· Azure OpenAI service API version: ", 
+                                             lines=1, interactive=True).style(container=False)
+            gr.HTML("Azure Cognitive Speech Service parameters to use VoiceChat. ")
+            with gr.Row():    
+                with gr.Column(scale=0.6):
+                    ui_stts_key = gr.Textbox(placeholder="Please enter your speech service API key if you want to try VoiceChat. " + 
+                                             "Please input these settings and hit 'Enter' key.", 
+                                             label="· Azure Cognitive Speech service API Key: ", interactive=True, type='password').style(container=False)
+                with gr.Column(scale=0.4):
+                    ui_stts_loc = gr.Textbox(placeholder="Please enter your speech service region.", 
+                                             label="· Azure Cognitive Speech service region: ", interactive=True).style(container=False)
+            ui_aoai_url.submit(get_aoai_set, [ui_aoai_url, ui_aoai_key, ui_aoai_api], [ui_aoai_url, ui_aoai_key, ui_aoai_api])
+            ui_aoai_key.submit(get_aoai_set, [ui_aoai_url, ui_aoai_key, ui_aoai_api], [ui_aoai_url, ui_aoai_key, ui_aoai_api])
+            ui_aoai_api.submit(get_aoai_set, [ui_aoai_url, ui_aoai_key, ui_aoai_api], [ui_aoai_url, ui_aoai_key, ui_aoai_api])
+            ui_stts_key.submit(get_stts_set, [ui_stts_key, ui_stts_loc], [ui_stts_key, ui_stts_loc])
+            ui_stts_loc.submit(get_stts_set, [ui_stts_key, ui_stts_loc], [ui_stts_key, ui_stts_loc])
+
+        with gr.TabItem("GPT-3.5 Playground"):
             ui_chatbot_gpt = gr.Chatbot(label="GPT Playground:")
             with gr.Row():
                 with gr.Column(scale=0.9):
                     ui_prompt_gpt = gr.Textbox(placeholder="Please enter your prompt here.", show_label=False).style(container=False)
                 with gr.Column(scale=0.1, min_width=100):
                     ui_clear_gpt = gr.Button("Clear Input", )
-            with gr.Accordion("Expand to config parameters:", open=False):
-                gr.Markdown("Look at me...")
+            with gr.Accordion("Expand to config parameters:", open=True):
+                ui_memo_gpt = gr.HTML("GPT-3.5 playground use Completion(). So you just need to provide model name as engine parameter.")
+                ui_model_gpt = gr.Textbox(placeholder="Azure OpenAI GPT model deployment name. ", 
+                                          label="- Azure OpenAI deployment name:", lines=1).style(container=False)
                 with gr.Row():
                     ui_temp_gpt = gr.Slider(0.1, 1.0, 0.9, step=0.1, label="Temperature", interactive=True)
                     ui_max_tokens_gpt = gr.Slider(100, 4000, 1000, step=100, label="Max Tokens", interactive=True)
@@ -61,6 +107,11 @@ with gr.Blocks() as page:
                 ui_top_p_gpt.value = slider_3
                 print("Log - Updated GPT parameters: Temperature=", ui_temp_gpt.value,
                     " Max Tokens=", ui_max_tokens_gpt.value, " Top_P=", ui_top_p_gpt.value)
+                
+            def get_engine_gpt(get_aoai_model):
+                global model_gpt
+                model_gpt = get_aoai_model
+                return gr.update(value=get_aoai_model)
             
             def select_response_gpt(radio):
                 if radio == "Response from OpenAI Model":
@@ -76,8 +127,9 @@ with gr.Blocks() as page:
             
             def bot_gpt(history):
                 global gpt_x
+                print(ui_model_gpt.value)
                 gpt_x = openai.Completion.create(
-                    engine="mvp-text-davinci-003",
+                    engine=model_gpt,
                     prompt=prompts,
                     temperature=0.6,
                     max_tokens=1000,
@@ -92,17 +144,17 @@ with gr.Blocks() as page:
                 history[-1][1] = gpt_reply
                 return history
             
+        ui_model_gpt.submit(get_engine_gpt, ui_model_gpt , ui_model_gpt) 
         ui_temp_gpt.change(get_parameters_gpt, [ui_temp_gpt, ui_max_tokens_gpt, ui_top_p_gpt])
         ui_max_tokens_gpt.change(get_parameters_gpt, [ui_temp_gpt, ui_max_tokens_gpt, ui_top_p_gpt])
         ui_top_p_gpt.change(get_parameters_gpt, [ui_temp_gpt, ui_max_tokens_gpt, ui_top_p_gpt])
-
         ui_prompt_gpt.submit(user_gpt, [ui_prompt_gpt, ui_chatbot_gpt], [ui_prompt_gpt, ui_chatbot_gpt], queue=False).then(
             bot_gpt, ui_chatbot_gpt, ui_chatbot_gpt
                 )
         ui_clear_gpt.click(lambda: None, None, ui_chatbot_gpt, queue=False)
         ui_res_radio_gpt.change(select_response_gpt, ui_res_radio_gpt, ui_response_gpt)
 
-        with gr.TabItem("ChatGPT"):
+        with gr.TabItem("ChatGPT on GPT-4"):
             ui_chatbot_chat = gr.Chatbot(label="ChatGPT:")
             with gr.Row():
                 with gr.Column(scale=0.9):
@@ -110,9 +162,13 @@ with gr.Blocks() as page:
                 with gr.Column(scale=0.1, min_width=100):
                     ui_clear_chat = gr.Button("Clear Chat")
             with gr.Blocks():
-                with gr.Accordion("Expand to config parameters:", open=False):
-                    gr.Markdown("Here is the default system prompt, you can change it to your own prompt.")
-                    ui_prompt_sys = gr.Textbox(value="You are an AI assistant that helps people find information.", show_label=False, interactive=True).style(container=False)
+                with gr.Accordion("Expand to config parameters:", open=True):
+                    gr.HTML("ChatGPT use ChatCompletion(). Here is the default system prompt, you can change it to your own prompt.")
+                    ui_prompt_sys = gr.Textbox(value="You are an AI assistant that helps people find information.", 
+                                               label="Here is the default system prompt, you can change it to your own prompt.", 
+                                               interactive=True).style(container=False)
+                    ui_model_chat = gr.Textbox(placeholder="Azure OpenAI model deployment name. ", 
+                                              label="· Azure OpenAI GPT-3.5/4 deployment name:", lines=1).style(container=False)
                     with gr.Row():
                         ui_temp_chat = gr.Slider(0.1, 1.0, 0.7, step=0.1, label="Temperature", interactive=True)
                         ui_max_tokens_chat = gr.Slider(100, 8000, 2000, step=100, label="Max Tokens", interactive=True)
@@ -127,7 +183,12 @@ with gr.Blocks() as page:
                 ui_top_p_chat.value = slider_3
                 print("Log - Updated chatGPT parameters: Temperature=", ui_temp_chat.value,
                     " Max Tokens=", ui_max_tokens_chat.value, " Top_P=", ui_top_p_chat.value)
-            
+
+            def get_engine_chat(get_aoai_model):
+                global model_chat
+                model_chat = get_aoai_model
+                return gr.update(value=get_aoai_model)
+
             def select_response_chat(radio):
                 if radio == "Response from OpenAI Model":
                     return gr.update(value=chat_x)
@@ -141,7 +202,7 @@ with gr.Blocks() as page:
             def bot_chat(history):
                 global chat_x
                 chat_x = openai.ChatCompletion.create(
-                    engine="mvp-gpt-35-turbo", messages=messages_chat,
+                    engine=model_chat, messages=messages_chat,
                     temperature=ui_temp_chat.value,
                     max_tokens=ui_max_tokens_chat.value,
                     top_p=ui_top_p_chat.value,
@@ -165,6 +226,7 @@ with gr.Blocks() as page:
                     {"role": "system", "content": sysmsg},
                 ]
         
+        ui_model_chat.submit(get_engine_chat, ui_model_chat, ui_model_chat) 
         ui_res_radio_chat.change(select_response_chat, ui_res_radio_chat, ui_response_chat)
         ui_temp_chat.change(get_parameters_chat, [ui_temp_chat, ui_max_tokens_chat, ui_top_p_chat])
         ui_max_tokens_chat.change(get_parameters_chat, [ui_temp_chat, ui_max_tokens_chat, ui_top_p_chat])
@@ -176,8 +238,9 @@ with gr.Blocks() as page:
         ui_clear_chat.click(lambda: None, None, ui_chatbot_chat, queue=False).then(reset_sys, ui_prompt_sys)
 
 
-        with gr.TabItem("DALL·E 2"):
-            ui_prompt_walle = gr.Textbox(placeholder="Please enter your prompt here to generate image.", show_label=False).style(container=False)
+        with gr.TabItem("DALL·E 2 Painting"):
+            ui_prompt_walle = gr.Textbox(placeholder="Please enter your prompt here to generate image.", 
+                                         show_label=False).style(container=False)
             ui_image_walle = gr.Image()
             with gr.Accordion("Select radio button to see detail:", open=False):            
                 ui_response_walle = gr.TextArea(show_label=False, interactive=False).style(container=False)
@@ -212,15 +275,15 @@ with gr.Blocks() as page:
 
         ui_prompt_walle.submit(get_image_walle, ui_prompt_walle, ui_image_walle, queue=False).then(get_response_walle, None, ui_response_walle)
 
-        with gr.TabItem("VoiceChat"):
+        with gr.TabItem("VoiceChat on GPT"):
             with gr.Row():
                 with gr.Column():
-                    with gr.Accordion("Expand to config parameters:", open=False):
-                        ui_prompt_sys_vchat = gr.Textbox(value="You are an AI assistant that helps people find information and just response with SSML.", show_label=False, interactive=True).style(container=False)
-                        ui_stts_key = gr.Textbox(placeholder="Input your Azure Speech service API key", show_label=False, interactive=True).style(container=False)
-                        ui_stts_key.submit(lambda x: None, ui_stts_key)
-                        stts_key = ui_aoai_key.value
-                        print(stts_key)
+                    with gr.Accordion("Expand to config parameters:", open=True):
+                        ui_prompt_sys_vchat = gr.Textbox(value="You are an AI assistant that helps people find information and just respond with SSML.", 
+                                                         label="Here is the default system prompt, you can change it to your own prompt.", 
+                                                         interactive=True).style(container=False)
+                        ui_model_vchat = gr.Textbox(placeholder="- Azure OpenAI model deployment name. ", 
+                                                   label="- Azure OpenAI GPT-3.5/4 deployment name:", lines=1).style(container=False)
 
                     ui_voice_inc_vchat = gr.Audio(source="microphone", type="filepath")
                     ui_voice_out_vchat = gr.Audio(value=None, type="filepath", interactive=False).style(container=False)
@@ -242,22 +305,24 @@ with gr.Blocks() as page:
                 print("Log - Updated chatGPT parameters: Temperature=", ui_temp_vchat.value,
                     " Max Tokens=", ui_max_tokens_vchat.value, " Top_P=", ui_top_p_vchat.value)
             
+            def get_engine_vchat(get_aoai_model):
+                global model_vchat
+                model_vchat = get_aoai_model
+                return gr.update(value=get_aoai_model)
+            
             def select_response_vchat(radio):
                 if radio == "Response from OpenAI Model":
                     return gr.update(value=vchat_x)
                 else:
                     return gr.update(value=messages_vchat)
 
-            
-
             def speech_to_text(voice_message):
                 # Downsample input voice to 16kHz
                 voice_wav = am.from_file(voice_message, format='wav')
                 voice_wav = voice_wav.set_frame_rate(16000)
                 voice_wav.export(voice_message, format='wav')
                 # STT
-                OASK = stts_key
-                service_region = "westus"
+                service_region = stts_region
 
                 base_url = "https://"+service_region+".stt.speech.microsoft.com/"
                 path = 'speech/recognition/conversation/cognitiveservices/v1'
@@ -267,7 +332,7 @@ with gr.Blocks() as page:
                         'format': 'detailed'
                         }
                 headers = {
-                        'Ocp-Apim-Subscription-Key': OASK_Speech,
+                        'Ocp-Apim-Subscription-Key': stts_key,
                         'Content-Type': 'audio/wav; codecs=audio/pcm; samplerate=16000',
                         'Accept': 'application/json;text/xml'
                         }
@@ -285,14 +350,14 @@ with gr.Blocks() as page:
                 return sst_text  
             
             def text_to_speech():
-                OASK_Speech = stts_key
-                service_region = "westus"
-
+                service_region = stts_region   
+                # test
+                print(stts_key)
                 base_url = "https://"+service_region+".tts.speech.microsoft.com/"
                 path = 'cognitiveservices/v1'
                 constructed_url = base_url + path
                 headers = {
-                        'Ocp-Apim-Subscription-Key': OASK_Speech,
+                        'Ocp-Apim-Subscription-Key': stts_key,
                         'Content-Type': 'application/ssml+xml',
                         'X-Microsoft-OutputFormat': 'riff-24khz-16bit-mono-pcm',
                         'User-Agent': 'Voice ChatGPT'
@@ -324,7 +389,7 @@ with gr.Blocks() as page:
             def bot_vchat(history):
                 global vchat_x, vchat_reply
                 vchat_x = openai.ChatCompletion.create(
-                    engine="mvp-gpt-35-turbo", messages=messages_vchat,
+                    engine=model_vchat, messages=messages_vchat,
                     temperature=ui_temp_chat.value,
                     max_tokens=ui_max_tokens_chat.value,
                     top_p=ui_top_p_chat.value,
@@ -338,8 +403,9 @@ with gr.Blocks() as page:
                 messages_vchat.append({"role": "assistant", "content": vchat_reply})
                 history[-1][1] = vchat_reply
                 return history
-            
-        ui_res_radio_chat.change(select_response_chat, ui_res_radio_chat, ui_response_chat)
+
+        ui_model_vchat.submit(get_engine_vchat, ui_model_vchat, ui_model_vchat) 
+        ui_res_radio_vchat.change(select_response_vchat, ui_res_radio_vchat, ui_response_vchat)
         ui_temp_chat.change(get_parameters_chat, [ui_temp_chat, ui_max_tokens_chat, ui_top_p_chat])
         ui_max_tokens_chat.change(get_parameters_chat, [ui_temp_chat, ui_max_tokens_chat, ui_top_p_chat])
         ui_top_p_chat.change(get_parameters_chat, [ui_temp_chat, ui_max_tokens_chat, ui_top_p_chat])