telegram-phd-bot

Bot does almost nothing and notifies about that gladly. No thanks, stonks pls.

#telegramPhdBot #aiogram==3.2.0 #ChatGPT #OpenAI

---

• Feature
  • Actors
  • Features to Actors
    • OpenAI Response Trigger
• Run
  • Cronjob
  • Run PhD Task Once
• TODO
• Develop

---

Feature

Actors

Bot features depends on the next Telegram actors:

• [priority chats] - with triggers described below; (+ channels)
• [chats] - to where bot was merely added
• [chat-admin-rights] - group chats with admin rights for the bot
• [from superadmin] - in any chat, but not in a channel (where is impossible to identify message sender)
• [contributor chat] - the phd bot runs with its own OpenAI token (defined in .env). However, anyone could supply to bot his own token and thus, activate openAI features for yourself or even for the chat where bot has already existed (even if the chat not in [priority chats]). Ref to command add_openai_token in the bot menu.
• [chats] bot is alive and could appreciate when you add PhD bot to the chat

and for only 1 feature there is

• phd work excluded chats (check TG_PHD_WORK_EXCLUDE_CHATS)

Features to Actors

Below is features and actors map:

• [priority chats, chats] send work result via cronjob to recently active chats & priority chats according to bot_chat_messages_cache. To exclude you should be in [phd work excluded chats]
• [priority chats, chats] echo to some messages
• [chats, chat-admin-rights] greeting new bots, new members.
• [priority chats, from superadmin, contributor chat] send OpenAI completion to chats on some triggers by rotating all tokens about which bot knows. Here it uses special TokenApiRequestManager with relay on main token and tokens of contributors.
• [priority chats, from superadmin, contributor chat] support dialog with help of OpenAI gpt-3.
• [priority chats, chats] because of the above: it stores message model in redis (redis.pipe transactions)
  • log bot messages
  • log other messages
• [chats] get chat id by command
• [chats] even random user could get access to the OpenAI features by providing his token to the bot and open access to OpenAI ChatGPT feature in telegram.
• [from superadmin] superadmin commands like: commands/admin
• [priority chats, from superadmin, contributor chat] Now it is possible to use OpenaAI DellE model to generate image from prompt.

| Additionally, bot stores all tokens cryptographically in the Redis db.

OpenAI Response Trigger

It responses when:

• chat id in the [priority chats] list and:
  • text length > 350 symbols,
  • ends with ('...', '..', ':'),
  • with bot mentioned via @,
  • replied on a bot message,
  • text consists of question mark (?)
• superadmin messages into [priority chats, chats] and:
  • bot is mentioned (e.g. @MiptPhDBot, SUSY is not exist anymore),
  • replied on a bot message,

Under the hood it uses completion model and chatGPT as chat completion model. The last one is chosen only when there is a dialog context exists, i.e. it is possible to get previous context (message has replay_to and this source message is in the redis cache).

| You could control depth of context fetching by according env (check env section).

| From OpenAI you may need API access token: https://platform.openai.com/account/usage (18-5 USD for a fresh new account usage).

Run

Prepare .env as in .example.env and start with docker compose orchestrator.

full env that is used you could find in bot settings.py.

Cronjob

docker-compose up

Run PhD Task Once

docker-compose run bot --phd-work-notification-run-once

TODO

• store tokens with ciphering
• reuse fetched data in handlers from filters
• notify user when token does not work
• add request admin right on join
• add superadmin stats fetch
• broadcast message from superadmin (ignore chats?)
• TODO: aiogram.exceptions.TelegramBadRequest: Telegram server says - Bad Request: message is too long
• add service task to delete all keys with expired updated_chat_ttl.

Develop

To Develop you may use the same docker compose, merely do not forget to rebuild always after changes, e.g. docker-compose up --build. Or write your own docker-compose with volume mounting.