chatbot1-9  Exercise 2: Which Group Messages Does the Bot Answer?
==================================================================

ex2.py (save in the project root):

  import asyncio
  from bot.channels.telegram import TelegramAdapter
  from bot.core import Bot
  from bot.providers import FakeProvider

  class FakeAPI:
      def __init__(self): self.sent = []
      async def send_message(self, chat_id, text, reply_to=None): self.sent.append((chat_id, text, reply_to))
      async def send_chat_action(self, chat_id, action="typing"): pass

  def group(text, uid=7, update_id=1, reply_to_bot=False, chat_type="supergroup"):
      msg = {"message_id": update_id, "from": {"id": uid, "is_bot": False}, "chat": {"id": -100, "type": chat_type}, "date": 1, "text": text}
      if reply_to_bot:
          msg["reply_to_message"] = {"message_id": 1, "from": {"id": 999, "is_bot": True}, "chat": msg["chat"], "date": 1, "text": "x"}
      return {"update_id": update_id, "message": msg}

  CASES = [
      ("plain chat between members",                  group("did anyone read the new novel?")),
      ("mentions the bot",                            group("@LibraryBot when do you open?")),
      ("mentions it in the middle, other case",       group("hey @librarybot, any news?")),
      ("a command without the bot's name",            group("/facts")),
      ("a command with the bot's name",               group("/facts@LibraryBot")),
      ("a reply to one of the bot's messages",        group("thanks!", reply_to_bot=True)),
      ("addresses a DIFFERENT bot",                   group("@OtherBot when do you open?")),
      ("message from another bot that mentions us",   {**group("@LibraryBot hello"), "message": {**group("@LibraryBot hello")["message"], "from": {"id": 5, "is_bot": True}}}),
  ]

  async def main():
      print(f"{'group message':<44} {'answered?':<10} text sent to the model")
      for n, (label, update) in enumerate(CASES, start=1):
          api = FakeAPI()
          bot = Bot(FakeProvider(["(model reply)"]))
          adapter = TelegramAdapter(bot, api, username="LibraryBot", bot_id=999)
          update["update_id"] = n
          await adapter.handle_update(update)
          sent = bot.provider.calls[0][-1]["content"] if bot.provider.calls else "-"
          answered = bool(api.sent)
          print(f"{label:<44} {'yes' if answered else 'no':<10} {sent if bot.provider.calls else ('(a command, no model call)' if answered else '')}")
  asyncio.run(main())

Run with:  py ex2.py

Output:

  group message                                answered?  text sent to the model
  plain chat between members                   no         
  mentions the bot                             yes        when do you open?
  mentions it in the middle, other case        yes        hey , any news?
  a command without the bot's name             no         
  a command with the bot's name                yes        (a command, no model call)
  a reply to one of the bot's messages         yes        thanks!
  addresses a DIFFERENT bot                    no         
  message from another bot that mentions us    no         

The rule I would reconsider: ignoring a bare command such as /facts. In a group with only this bot, people will type /facts and expect an answer, and the bot's silence looks like a fault. Telegram's own behaviour is that general commands go to all bots in the group (and to privacy-mode bots only in limited cases), and the update does not say which bot the person meant. The risk of accepting bare commands is that two bots in the same group both answer, or that this bot reacts to a command meant for another. A reasonable compromise: accept bare commands
when the group has a single bot, which needs a setting (or a member-count check) that I did not build.

WHY THIS WORKS AS AN ANSWER
---------------------------
Each case is a fresh bot and adapter fed one hand-written group update. Only messages that mention the bot, use /command@BotName or reply to one of the bot's messages are answered, matching case-insensitively. The mention is removed before the text reaches the model (the middle-of-sentence case leaves "hey , any news?", untidy but harmless). A command is answered by the bot's own command handler without a model call. A mention of a different bot and a message from another bot are ignored, which also avoids two bots endlessly answering each other.