v0.7201 - perplexity handling improvements & fixes

README.md

@@ -57,6 +57,7 @@ yt-dlp>=2024.3.10
 - Use the `configmerger.py` to update old configuration files into a newer version's `config.ini`. You can do this by creating a copy of your existing config to i.e. a file named `myconfig.txt` and including in it the lines you want to keep for the newer version. Then, just run `python configmerger.py config.ini myconfig.txt` and all your existing config lines will be migrated to the new one. Works in most cases, but remember to be careful and double-check any migration issues with i.e. `diff`!
 
 # Changelog
+- v0.7201 - added Perplexity API model configuration to `config.ini`
 - v0.72 - improved error catching + messaging with Perplexity's API
 - v0.71 - holiday mentions via Python's `holidays` module (can be extended)
 - v0.708 - improved astronomy data combining via WeatherAPI

api_perplexity_search.py

@@ -11,22 +11,41 @@
 import os
 import httpx
 import asyncio
+import configparser
 
 from langdetect import detect
 from telegram import constants
 
+# Load the configuration file
+config = configparser.ConfigParser()
+config.read('config.ini')
+
 # ~~~~~~~~~
 # variables
 # ~~~~~~~~~
-
-# Global variable for chunk size
-# Set this value as needed
-CHUNK_SIZE = 500
+# Define fallback default values
+DEFAULT_PERPLEXITY_MODEL = "llama-3-sonar-large-32k-online"
+DEFAULT_PERPLEXITY_MAX_TOKENS = 1024
+DEFAULT_PERPLEXITY_TEMPERATURE = 0.0
+DEFAULT_PERPLEXITY_MAX_RETRIES = 3
+DEFAULT_PERPLEXITY_RETRY_DELAY = 25
+DEFAULT_PERPLEXITY_TIMEOUT = 30
+# chunk sizes are for translations
+DEFAULT_CHUNK_SIZE = 500
+
+# Perplexity API settings from config with fallback defaults
+PERPLEXITY_MODEL = config.get('Perplexity', 'Model', fallback=DEFAULT_PERPLEXITY_MODEL)
+PERPLEXITY_MAX_TOKENS = config.getint('Perplexity', 'MaxTokens', fallback=DEFAULT_PERPLEXITY_MAX_TOKENS)
+PERPLEXITY_TEMPERATURE = config.getfloat('Perplexity', 'Temperature', fallback=DEFAULT_PERPLEXITY_TEMPERATURE)
+PERPLEXITY_MAX_RETRIES = config.getint('Perplexity', 'MaxRetries', fallback=DEFAULT_PERPLEXITY_MAX_RETRIES)
+PERPLEXITY_RETRY_DELAY = config.getint('Perplexity', 'RetryDelay', fallback=DEFAULT_PERPLEXITY_RETRY_DELAY)
+PERPLEXITY_TIMEOUT = config.getint('Perplexity', 'Timeout', fallback=DEFAULT_PERPLEXITY_TIMEOUT)
+CHUNK_SIZE = config.getint('Perplexity', 'ChunkSize', fallback=DEFAULT_CHUNK_SIZE)
 
 # Assuming you've set PERPLEXITY_API_KEY in your environment variables
 PERPLEXITY_API_KEY = os.getenv("PERPLEXITY_API_KEY")
 
-#  main perplexity function
+# Main Perplexity function
 async def fact_check_with_perplexity(question: str):
     url = "https://api.perplexity.ai/chat/completions"
     headers = {
@@ -35,10 +54,10 @@ async def fact_check_with_perplexity(question: str):
         "Accept": "application/json",
     }
     data = {
-        "model": "sonar-small-online",  # Specifying the model
+        "model": PERPLEXITY_MODEL,  # Specifying the model
         "stream": False,
-        "max_tokens": 1024,
-        "temperature": 0.0,  # Adjust based on how deterministic you want the responses to be
+        "max_tokens": PERPLEXITY_MAX_TOKENS,
+        "temperature": PERPLEXITY_TEMPERATURE,  # Adjust based on how deterministic you want the responses to be
         "messages": [
             {
                 "role": "user",
@@ -47,8 +66,8 @@ async def fact_check_with_perplexity(question: str):
         ]
     }
 
-    async with httpx.AsyncClient(timeout=30) as client:  # Increased timeout
-        for attempt in range(3):  # Retry mechanism
+    async with httpx.AsyncClient(timeout=PERPLEXITY_TIMEOUT) as client:  # Increased timeout
+        for attempt in range(PERPLEXITY_MAX_RETRIES):  # Retry mechanism
             try:
                 response = await client.post(url, json=data, headers=headers)
                 if response.status_code == 200:
@@ -69,7 +88,7 @@ async def fact_check_with_perplexity(question: str):
 
     return None
 
-# queries perplexity
+# Queries Perplexity
 async def query_perplexity(bot, chat_id, question: str):
     logging.info(f"Querying Perplexity with question: {question}")
     response_data = await fact_check_with_perplexity(question)
@@ -108,26 +127,24 @@ async def translate_response(bot, user_message, perplexity_response):
         # Directly convert and return if language detection fails; assuming English or Markdown needs HTML conversion
         formatted_response = format_headers_for_telegram(perplexity_response)
         return markdown_to_html(formatted_response)
-    
+
     # Check if the detected language is English, skip translation if it is
     if user_lang == 'en':
         logging.info("User's question is in English, converting Markdown to HTML.")
         formatted_response = format_headers_for_telegram(perplexity_response)
         return markdown_to_html(formatted_response)
     else:
-        # await context.bot.send_message(chat_id=update.effective_chat.id, text="<i>Translating, please wait...</i>", parse_mode=telegram.ParseMode.HTML)
         logging.info(f"User's question is in {user_lang}, proceeding with translation.")
-    
+
     # System message to guide the model for translating
     system_message = {
         "role": "system",
         "content": f"Translate the message to: {user_lang}."
     }
-    
+
     # Prepare the chat history with only the Perplexity's response as the assistant's message to be translated
     chat_history = [
         system_message,
-        # {"role": "user", "content": user_message},
         {"role": "user", "content": perplexity_response}
     ]
 
@@ -158,7 +175,7 @@ async def translate_response(bot, user_message, perplexity_response):
             return translated_reply
         except Exception as e:
             logging.error(f"Error processing translation response: {e}")
-            return f"Translation failed due to an error: {e}"  
+            return f"Translation failed due to an error: {e}"
     else:
         logging.error(f"Error in translating response: {response.text}")
         return f"Failed to translate, API returned status code {response.status_code}: {response.text}"

config.ini

@@ -119,3 +119,24 @@ ElasticsearchEnabled = False
 # ~~~~~~~~~~~~~~~~~~~~~
 [HolidaySettings]
 EnableHolidayNotification = true
+
+# ~~~~~~~~~~~~~~~
+# Perplexity API
+# ~~~~~~~~~~~~~~~
+[Perplexity]
+# Model name to use with Perplexity API
+Model = llama-3-sonar-large-32k-online
+
+# Maximum tokens for Perplexity API response
+MaxTokens = 1024
+
+# Temperature for Perplexity API response
+Temperature = 0.0
+
+# Retry settings for Perplexity API
+MaxRetries = 3
+RetryDelay = 25
+Timeout = 30
+
+# Chunk size for translation
+ChunkSize = 500

main.py

@@ -5,7 +5,7 @@
 # https://github.com/FlyingFathead/TelegramBot-OpenAI-API
 #
 # version of this program
-version_number = "0.72"
+version_number = "0.7201"
 
 # experimental modules
 import requests

text_message_handler.py

@@ -474,14 +474,12 @@ async def handle_message(bot, update: Update, context: CallbackContext, logger)
                     # ~~~~~~~~~~~~~~
                     # Perplexity API
                     # ~~~~~~~~~~~~~~
-
-                    # Handling the Perplexity API call with automatic translation                    
+                    # Handling the Perplexity API call with automatic translation
                     elif function_name == 'query_perplexity':
                         arguments = json.loads(function_call.get('arguments', '{}'))
                         question = arguments.get('question', '')
 
                         if question:
-
                             # Make the asynchronous API call to query Perplexity
                             perplexity_response = await query_perplexity(context.bot, chat_id, question)
 
@@ -494,18 +492,32 @@ async def handle_message(bot, update: Update, context: CallbackContext, logger)
                                 chat_history.append({"role": "system", "content": "Perplexity API is currently unavailable due to server issues. Please inform the user about this issue."})
                             else:
                                 if perplexity_response is not None:
-                                    # Append the bot's reply to the chat history before sending it
-                                    chat_history.append({"role": "system", "content": f"[Response from Perplexity API, translate to user's language as needed] {perplexity_response}"})
-                                    context.chat_data['chat_history'] = chat_history  # Update the chat data with the new history
+                                    # Flag for translation in progress
+                                    context.user_data['active_translation'] = True
+
+                                    # Translate or process the response as necessary
+                                    bot_reply_formatted = await translate_response_chunked(bot, user_message, perplexity_response, context, update)
+
+                                    # After translation or processing is completed, clear the active translation flag
+                                    context.user_data.pop('active_translation', None)
 
-                                    await context.bot.send_message(
-                                        chat_id=update.effective_chat.id,
-                                        text=perplexity_response,
-                                        parse_mode=ParseMode.HTML
-                                    )
+                                    if bot_reply_formatted and not bot_reply_formatted.startswith("Error"):  # Check for a valid, non-error response
+                                        # Append the bot's reply to the chat history before sending it
+                                        chat_history.append({"role": "assistant", "content": bot_reply_formatted})
+                                        context.chat_data['chat_history'] = chat_history  # Update the chat data with the new history
 
-                                    response_sent = True  # Indicate that a response has been sent
-                                    break  # Exit the loop since response has been handled
+                                        await context.bot.send_message(
+                                            chat_id=update.effective_chat.id,
+                                            text=bot_reply_formatted,
+                                            parse_mode=ParseMode.HTML
+                                        )
+
+                                        response_sent = True  # Indicate that a response has been sent
+                                        break  # Exit the loop since response has been handled
+                                    else:
+                                        # Log the error and maybe send a different message or handle the error differently
+                                        logging.error("Error processing or translating the Perplexity response.")
+                                        chat_history.append({"role": "system", "content": "Fallback to base model due to processing error in Perplexity response."})
 
                                 else:
                                     logging.error("No valid response from Perplexity, Perplexity response was None or empty.")
@@ -517,6 +529,11 @@ async def handle_message(bot, update: Update, context: CallbackContext, logger)
 
                         context.chat_data['chat_history'] = chat_history
 
+
+                    # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+                    # old code; to be nuked and paved ...
+                    # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
                     # # Handling the Perplexity API call with automatic translation
                     # elif function_name == 'query_perplexity':
                     #     arguments = json.loads(function_call.get('arguments', '{}'))