gogo2/agent-py-bot/agent.py

import logging
import asyncio, nest_asyncio
from telegram import Bot, Message, Update
from telegram.constants import ParseMode
from telegram.ext import Application, CommandHandler, MessageHandler, filters, ContextTypes, CallbackContext
import os
import requests
import json
import base64
from selenium import webdriver
from selenium.webdriver.chrome.options import Options
from io import BytesIO
from PIL import Image
from datetime import datetime, timedelta

# Apply nest_asyncio
nest_asyncio.apply()

# Set up logging
logging.basicConfig(format='%(asctime)s - %(name)s - %(levelname)s - %(message)s', level=logging.INFO)
logger = logging.getLogger(__name__)

# Telegram Bot Token
# t.me/kevin_ai_robot
TOKEN = '6805059978:AAHNJKuOeazMSJHc3-BXRCsFfEVyFHeFnjw'
# t.me/artitherobot 6749075936:AAHUHiPTDEIu6JH7S2fQdibwsu6JVG3FNG0

# This can be your own ID, or one for a developer group/channel.
# You can use the /start command of this bot to see your chat id.
DEVELOPER_CHAT_ID = "777826553"

# LLM API Endpoint
LLM_ENDPOINT = "http://192.168.0.11:11434/api/chat"

APPEND_RESULTS = os.getenv('APPEND_RESULTS', 'True') == 'True'

async def start(update: Update, context: CallbackContext):
    await context.bot.send_message(chat_id=update.effective_chat.id, text="Hi! I'm your AI bot. Ask me aything with /ask")

async def echo(update: Update, context: CallbackContext):
        # Check if in ask mode
    if context.chat_data.get('ask_mode'):
        await context.bot.send_chat_action(chat_id=update.effective_chat.id, action="typing")       
        # Process as if it's an ask command
        context.chat_data['messages'].append(update.message.text)
        # Process the concatenated messages
        user_message = ' '.join(context.chat_data['messages'])
        llm_response = await query_llm(user_message)
        await update.message.reply_text("[ask]"+llm_response)
    else:
        # Regular echo behavior
        await update.message.reply_text(update.message.text)
    # await context.bot.send_message(chat_id=update.effective_chat.id, text=update.message.text)

async def ask(update: Message, context: CallbackContext):
    try:
        context.chat_data['ask_mode'] = True
        context.chat_data['messages'] = []
        # Send typing action
        #await context.bot.send_chat_action(chat_id=update.effective_chat.id, action=telegram.ChatAction.TYPING)
        await context.bot.send_chat_action(chat_id=update.effective_chat.id, action="typing")       

        user_message = ' '.join(context.args)
        llm_response = await query_llm(user_message)
        await update.message.reply_text(llm_response)
    except Exception as e:
        # Log the exception
        logger.error(f"An error occurred: {e}")
        # Optionally, send a message to the user about the error
        await update.message.reply_text("An error occurred while processing your request.")

async def ok(update: Update, context: CallbackContext):
    context.chat_data['ask_mode'] = False
    context.chat_data['messages'] = []
    await update.message.reply_text("Exiting ask mode.")

async def brief(update: Update, context: CallbackContext):


    await update.message.reply_text("Exiting ask mode.")

#                CODE RUNNER
import re
from agents.runner import execute_python_code

#https://github.com/jmorganca/ollama/blob/main/docs/api.md#generate-a-completion
async def query_llm(user_message, model=None):
    """Query the LLM with the user's message."""
    # use the model if provided, otherwise use the default llama2
    if model is None:
        model = "llama2:latest"

    data = {
        "model": model,
        "messages": [{"role": "user", "content": user_message}],
        "stream": False
    }
    
    response = requests.post(LLM_ENDPOINT, json=data)
    if response.status_code == 200:
            response_data = response.json()
            if "error" in response_data:
                error_message = response_data.get('error', 'Unknown error')
                # Log the error
                logger.error(f"LLM Error: {error_message}")
                # Return a user-friendly error message
                return "Sorry, there was an error processing your request."
            # handle response
            content = response_data.get('message', {}).get('content', 'No response')

            # Find and execute all code blocks
            code_blocks = re.findall(r"```(.*?)```", content, re.DOTALL)
            if code_blocks:
                for code in code_blocks:
                    execution_result = execute_python_code(code.strip())
                    if APPEND_RESULTS:
                        # Append the result after the code block
                        content = content.replace(f"```{code}```", f"```{code}```\n```{execution_result}```")
                    else:
                        # Replace the code block with its result
                        content = content.replace(f"```{code}```", f"```{execution_result}```")

            return content
    else:    
        logger.error(f"Error reaching LLM: {response.text}")
        return "Error: Unable to reach the AI agent."
    
async def execute_code(code_block):
    """
    Execute the given Python code in a separate, sandboxed environment.
    Returns the output or any errors encountered.
    """
    try:
        # Example: Using subprocess to run code in an isolated environment
        # This is a basic example and not secure. Use a proper sandbox setup.
        result = subprocess.run(['python', '-c', code_block], 
                                capture_output=True, text=True, timeout=5)
        return result.stdout or result.stderr
    except subprocess.TimeoutExpired:
        return "Execution timed out."
    except Exception as e:
        return f"An error occurred: {str(e)}"


async def main():
    """Start the bot."""
    # Create an Application instance
    application = Application.builder().token(TOKEN).build()

    # Add handlers to the application
    # Command handlers should be registered before the generic message handler
    application.add_handler(CommandHandler("start", start))
    # application.add_handler(CommandHandler("screenshot", screenshot)) # Ensure screenshot function is async
    application.add_handler(CommandHandler("ask", ask))
    application.add_handler(CommandHandler("ok", ok))
    
    application.add_handler(CommandHandler("bad_command", bad_command))
    # This handler should be last as it's the most generic
    application.add_handler(MessageHandler(filters.TEXT, echo))

    # ...and the error handler
    application.add_error_handler(error_handler)

    # Run the bot
    await application.run_polling()


import html
import traceback
async def error_handler(update: object, context: ContextTypes.DEFAULT_TYPE) -> None:
    """Log the error and send a telegram message to notify the developer."""
    # Log the error before we do anything else, so we can see it even if something breaks.
    logger.error("Exception while handling an update:", exc_info=context.error)

    # traceback.format_exception returns the usual python message about an exception, but as a
    # list of strings rather than a single string, so we have to join them together.
    tb_list = traceback.format_exception(None, context.error, context.error.__traceback__)
    tb_string = "".join(tb_list)

    # Build the message with some markup and additional information about what happened.
    # You might need to add some logic to deal with messages longer than the 4096 character limit.
    update_str = update.to_dict() if isinstance(update, Update) else str(update)
    message = (
        "An exception was raised while handling an update\n"
        f"<pre>update = {html.escape(json.dumps(update_str, indent=2, ensure_ascii=False))}"
        "</pre>\n\n"
        f"<pre>context.chat_data = {html.escape(str(context.chat_data))}</pre>\n\n"
        f"<pre>context.user_data = {html.escape(str(context.user_data))}</pre>\n\n"
        f"<pre>{html.escape(tb_string)}</pre>"
    )

    # Finally, send the message
    await context.bot.send_message(
        chat_id=DEVELOPER_CHAT_ID, text=message, parse_mode=ParseMode.HTML
    )

async def bad_command(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None:
    """Raise an error to trigger the error handler."""
    await context.bot.wrong_method_name()  # type: ignore[attr-defined]

#------------------------- webagent --------------------------# 
import schedule
import time
from agents.webagent import run_web_agent, save_data

async def run_web_agent_and_process_result(topic, folder):
    news_data = run_web_agent(topic, folder)

    print(f"[{datetime.now()}] Doing summarisation and sentiment analysis with an AI model.")
    
    user_message = f"Summarize these news and make sentiment analysis on each news and one overall: {news_data}"
    start = time.time()
    query_result = await query_llm(user_message, "openhermes")
    print(f"[{datetime.now()}] AI call returned in {time.time() - start} seconds.")
    news_data["summary"] = query_result

    user_message = f"do sentiment analysis on theese news and report overall sentiment for the day from 1 to 100. Here's the current news articles: {news_data}"
    start = time.time()
    query_result = await query_llm(user_message, "openhermes")
    print(f"[{datetime.now()}] AI call returned in {time.time() - start} seconds.")
    news_data["sentimen"] = query_result
    
    save_data(news_data, folder)
    
    with open(os.path.join(folder, "summary_log.txt"), 'a') as log_file:
        log_file.write(f"\n\n\n{datetime.now()}: {query_result}\n")

    # Process the query_result as needed

async def async_main():
    topic = "tesla news"
    interval = 1  # in hours
    folder = "agent-py-bot/scrape/raw"

    while True:
        await run_web_agent_and_process_result(topic=topic, folder=folder)
        await asyncio.sleep(interval * 3600)  # Convert hours to seconds

if __name__ == '__main__':
    asyncio.run(async_main())