Update app.py

gpt4 chats and all image analysis use gpt-4o-2024-08-06

app.py

@@ -187,12 +187,12 @@ def chat(prompt, user_window, pwd_window, past, response, gptModel, uploaded_ima
             response = genUsageStats(True)
             return [past, response, None, gptModel, uploaded_image_file]
         if prompt.startswith('gpt4'):
-            gptModel = 'gpt-4o'
+            gptModel = 'gpt-4o-2024-08-06'
             prompt = prompt[5:]
         if prompt.startswith("clean"):
             user = prompt[6:]
-            response = f'cleaned all .wav files for {user}'
-            final_clean_up(user)
+            response = f'cleaned all .wav and .b64 files for {user}'
+            final_clean_up(user, True)
             return [past, response, None, gptModel, uploaded_image_file]
         if prompt.startswith('files'):
             (log_cnt, wav_cnt, other_cnt, others, log_list) = list_permanent_files()
@@ -204,6 +204,7 @@ def chat(prompt, user_window, pwd_window, past, response, gptModel, uploaded_ima
             completion = client.chat.completions.create(model=gptModel,
                                             messages=past)
         else:
+            gptModel = 'gpt-4o-2024-08-06'
             (completion, msg) = analyze_image(user_window, gptModel)
             uploaded_image_file= ''
             if not msg == 'ok':
@@ -301,11 +302,18 @@ def gen_speech_file_names(user, cnt):
         rv.append(dataDir + f'{user}_speech{i}.wav')
     return rv
 
-def final_clean_up(user):
-    if user.strip().lower() == 'all':
+def final_clean_up(user, do_b64 = False):
+    user = user.strip().lower()
+    if user == 'kill':
+        flist = glob(dataDir + '*')
+    elif user == 'all':
         flist = glob(dataDir + '*_speech*.wav')
+        if do_b64:
+            flist.extend(glob(dataDir + '*.b64'))
     else:
         flist = glob(dataDir + f'{user}_speech*.wav')
+        if do_b64:
+            flist.append(dataDir + user + '_image.b64')
     for fpath in flist:
         try:
             os.remove(fpath)
@@ -384,13 +392,13 @@ def show_help():
           4.4 When done viewing image, tap the "Restart Conversation" button
     5. (OR) Analyze an Image you provide:
           5.1 Enter what you want to know about the image in the prompt window. You can include instructions
-               to write a poem about something in the image, for example.
-          5.2 Tap the "Upload & Analyze Image" button.
-          5.3 An empty image box will appear lower left. Drag or upload image into it. It offers web cam input
-               also but I have not tried that yet.
+               to write a poem about something in the image, for example.  Or just say "what's in this image?"
+          5.2 Tap the "Upload Image to Analyze" button.
+          5.3 An empty image box will appear lower left. Drag or upload image into it. It offers web cam or camera
+               input also.
           5.4 The image should appear. This can take some time with a slow internet connection and large image.
           5.5 Tap the "Submit Prompt/Question" button to start the analysis.  This initiates a chat dialog and
-               you can ask follow-up questions.
+               you can ask follow-up questions. However, the image is not re-analyzed for follow-up dialog.
 
     Hints:
         1. Better chat and image results are obtained by including detailed descriptions and instructions
@@ -409,7 +417,7 @@ def upload_image(prompt, user, password):
     return [gr.Image(visible=True, interactive=True), '']
 
 def load_image(image, user):
-    status = 'ok'
+    status = 'OK, image is ready! Tap "Submit Prompt/Question" to start analyzing'
     try:
         with open(image, 'rb') as image_file:
             base64_image = base64.b64encode(image_file.read()).decode('utf-8')
@@ -417,7 +425,7 @@ def load_image(image, user):
         with open(fpath, 'wt') as fp:
             fp.write(base64_image)
     except:
-        status = 'Unable to create base64 image'
+        status = 'Unable to upload image'
     return [fpath, status]
 
 def analyze_image(user, model):
@@ -580,7 +588,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
         # gpt_chooser=gr.Radio(choices=[("GPT-3.5","gpt-3.5-turbo"),("GPT-4o","gpt-4o-mini")],
         #                      value="gpt-3.5-turbo", label="GPT Model", interactive=True)
         button_do_image = gr.Button(value='Make Image')
-        button_get_image = gr.Button(value='Upload & Analyze Image')
+        button_get_image = gr.Button(value='Upload Image to Analyze')
         submit_button = gr.Button(value="Submit Prompt/Question")
         speak_output = gr.Button(value="Speak Dialog", visible=False)
     prompt_window = gr.Textbox(label = "Prompt or Question")