Merge pull request #91 from carlsverre/gpt-4-support

gpt-4 model support

bot/openai_helper.py

@@ -6,6 +6,11 @@ import tiktoken
 
 import openai
 
+# Models can be found here: https://platform.openai.com/docs/models/overview
+GPT_3_MODELS = ("gpt-3.5-turbo", "gpt-3.5-turbo-0301")
+GPT_4_MODELS = ("gpt-4", "gpt-4-0314")
+GPT_4_32K_MODELS = ("gpt-4-32k", "gpt-4-32k-0314")
+GPT_ALL_MODELS = GPT_3_MODELS + GPT_4_MODELS + GPT_4_32K_MODELS
 
 class OpenAIHelper:
     """
@@ -23,6 +28,16 @@ class OpenAIHelper:
         self.conversations: dict[int: list] = {}  # {chat_id: history}
         self.last_updated: dict[int: datetime] = {}  # {chat_id: last_update_timestamp}
 
+    def get_conversation_stats(self, chat_id: int) -> tuple[int, int]:
+        """
+        Gets the number of messages and tokens used in the conversation.
+        :param chat_id: The chat ID
+        :return: A tuple containing the number of messages and tokens used
+        """
+        if chat_id not in self.conversations:
+            self.reset_chat_history(chat_id)
+        return (len(self.conversations[chat_id]), self.__count_tokens(self.conversations[chat_id]))
+
     async def get_chat_response(self, chat_id: int, query: str) -> Union[tuple[str, str], str]:
         """
         Gets a response from the GPT-3 model.
@@ -172,8 +187,12 @@ class OpenAIHelper:
         return response.choices[0]['message']['content']
 
     def __max_model_tokens(self):
-        if self.config['model'] == "gpt-3.5-turbo" or self.config['model'] == "gpt-3.5-turbo-0301":
+        if self.config['model'] in GPT_3_MODELS:
             return 4096
+        if self.config['model'] in GPT_4_MODELS:
+            return 8192
+        if self.config['model'] in GPT_4_32K_MODELS:
+            return 32768
         raise NotImplementedError(
             f"Max tokens for model {self.config['model']} is not implemented yet."
         )
@@ -190,7 +209,7 @@ class OpenAIHelper:
             encoding = tiktoken.encoding_for_model(model)
         except KeyError:
             encoding = tiktoken.get_encoding("gpt-3.5-turbo")
-        if model == "gpt-3.5-turbo" or model == "gpt-3.5-turbo-0301":
+        if model in GPT_ALL_MODELS:
             num_tokens = 0
             for message in messages:
                 num_tokens += 4  # every message follows <im_start>{role/name}\n{content}<im_end>\n

bot/telegram_bot.py

@@ -70,6 +70,9 @@ class ChatGPT3TelegramBot:
         transcribe_durations = self.usage[user_id].get_current_transcription_duration()
         cost_today, cost_month = self.usage[user_id].get_current_cost()
         
+        chat_id = update.effective_chat.id
+        chat_messages, chat_token_length = self.openai.get_conversation_stats(chat_id)
+
         usage_text = f"Today:\n"+\
                      f"{tokens_today} chat tokens used.\n"+\
                      f"{images_today} images generated.\n"+\
@@ -80,7 +83,11 @@ class ChatGPT3TelegramBot:
                      f"{tokens_month} chat tokens used.\n"+\
                      f"{images_month} images generated.\n"+\
                      f"{transcribe_durations[2]} minutes and {transcribe_durations[3]} seconds transcribed.\n"+\
-                     f"💰 For a total amount of ${cost_month:.2f}"
+                     f"💰 For a total amount of ${cost_month:.2f}"+\
+                     f"\n----------------------------\n\n"+\
+                     f"Current conversation:\n"+\
+                     f"{chat_messages} chat messages in history.\n"+\
+                     f"{chat_token_length} chat tokens in history.\n"
         await update.message.reply_text(usage_text)
 
     async def reset(self, update: Update, context: ContextTypes.DEFAULT_TYPE):
@@ -374,7 +381,7 @@ class ChatGPT3TelegramBot:
         """
         Handles errors in the telegram-python-bot library.
         """
-        logging.debug(f'Exception while handling an update: {context.error}')
+        logging.error(f'Exception while handling an update: {context.error}')
 
     def is_group_chat(self, update: Update) -> bool:
         """