enough for today.

app.py

@@ -6,6 +6,10 @@ from json import loads as json_loads
 from pathlib import Path
 import fitz
 
+MODEL = 'gpt-4-turbo'
+PRICE_PER_M = 10.00
+LIMIT = 120000 # some space for answer
+
 api_key = os_getenv("OPENAI_APIKEY")
 client = OpenAI(api_key=api_key)
 
@@ -14,12 +18,12 @@ def get_prompt(books, question = None):
         f"Read the following books.\n" +
         f"Each book may have some pages at the beggining with data about the book, an index, or table of content, etc. " +
         f"Pages may have a header and/or a footer. Consider all this maybe present." +
-        f"Please answer all below in the suggested format, in the language of the book:\n"+
+        f"Please answer, for each book, all below in the suggested format, in the language of the book:\n"+
         f"**Title**: ...\n"
         f"**Author**: ...\n"
         f"**Chapter Names**: ...\n"
         f"**Characters**: \n"
-        f"**Detailed Summary**: \n"
+        f"**Detailed Summary of the whole book**: \n"
     )
     prompt += f"{books}\n" 
 
@@ -45,7 +49,7 @@ def chat(message, history, files):
     history_openai_format.append({"role": "user", "content": message})
 
     response = client.chat.completions.create(
-        model='gpt-4-turbo',
+        model=MODEL,
         messages= history_openai_format,
         temperature=1.0,
         stream=True)
@@ -68,21 +72,31 @@ def get_text(filename):
 
 def files_ready(filenames):
     encoder = encoding = tiktoken.encoding_for_model('gpt-4-turbo')
-    answer = ''
+    books = ''
     for i, name in enumerate(filenames):
-        answer += f"\n## Document #{i+1}\nName: {Path(name).name}\n"
-        answer += get_text(name)
+        books += f"\n## Document #{i+1}\nName: {Path(name).name}\n"
+        books += get_text(name)
+
+    prompt = get_prompt(books)
+    tokens = len(encoder.encode(prompt))
+    cost = tokens * PRICE_PER_M / 1000000 * 2 # * 2 is too much for an answer
 
-    return len(encoder.encode(answer)), [[get_prompt(answer), None]]
+    if tokens > LIMIT:
+        raise gr.Error(f"Book is too long. It's {tokens} tokens long and can't be more than {LIMIT}.")
+    return tokens, f"${cost}", [[prompt, None]]
 
 def files_changed(filenames):
-    if not filenames:
-        return 0
+    if filenames:
+        return "-", "-"
+    else:
+        return 0, "$0"
     
 with gr.Blocks(title="Book summarization and more") as demo:
     with gr.Row():
         files = gr.Files(file_types=["txt","doc","docx","pdf"] )
-        tokens = gr.Text("0", label="Tokens")
+        with gr.Column():
+            tokens = gr.Text("0", label="Tokens")
+            cost = gr.Text("0", label="Cost")
 
     chat = gr.ChatInterface(
         fn=chat,
@@ -91,8 +105,8 @@ with gr.Blocks(title="Book summarization and more") as demo:
         multimodal=False)
     
     other = gr.Button(interactive=False)
-    files.upload(files_ready, [files], [tokens, chat.chatbot_state])
-    files.change(files_changed, files, tokens)
+    files.upload(files_ready, [files], [tokens, cost, chat.chatbot_state])
+    files.change(files_changed, files, [tokens, cost])
 
 
 auth=os_getenv("APP_USERS", "null")