What AI chat replies do
The module reads Telegram group chats and replies to people on behalf of your accounts. Each answer is a reply to a specific message: the person sees they were addressed, and everyone else sees a real conversation rather than an ad.
The module picks up new messages straight away and also looks through the chats on a schedule. Messages from your own accounts and from bots are skipped, so your accounts don't talk to each other.
Whoever messages you privately after a reply can be answered by the AI auto-reply NeuroDialogs.
When the module replies
- at intervals — now and then to fresh messages: not every one, only some, so the account doesn't look like a bot;
- on triggers — only to messages with your keywords, such as "any recommendations" or "who does".
Keywords only work in trigger mode. You can also add a condition to reply only if something is true: the model first decides whether the message fits, and stays silent when unsure.
The module doesn't reply to messages that are too short or too old, and pauses before replying to the same person again.
What the model writes
Replies follow a prompt — a ready-made one or your own: sales, support, lead qualification, short and to the point, one of the crowd, a reserved tone. The model sees the message itself, a few previous messages in the chat, the language, the topic and your product.
Product mentions are switched on separately — then the model names what you promote now and then, not in every reply. If the model writes nothing, the reply is skipped: no empty messages go into the chat.
Where chats come from
Groups come as a list — @usernames, t.me links or numeric IDs — a t.me/addlist folder, or saved lists and past campaigns. The scraper helps you find chats in your niche.
Default settings
| Setting | Default |
|---|---|
| Chat check | every 5 minutes |
| Don't reply to messages older than | 30 minutes |
| Minimum characters in a message | 15 |
| Pause per author | 120 minutes |
| Skip bots | on |
| Simulate typing | on |
| Product mentions | off |
| Working hours | off |
Limits and account safety
Replies in chats count towards the account's comment limit, shared with AI commenting. New accounts get a reduced limit that grows over the first week.
The module waits out a chat's slow mode. If a chat forbids writing, it moves to another account and is skipped after a second refusal; a ban in the chat puts it on the blacklist and sends the account for a check with @SpamBot. The module waits out FLOOD_WAIT — what it is and where it comes from, see the guide on FLOOD_WAIT.
What the module doesn't do
- it doesn't message people privately first;
- it doesn't write in channels — only in groups;
- it doesn't reply to every message — only according to the chosen mode;
- it doesn't promise the reply will stay: chat admins and anti-spam bots may delete it.
Telegram and chat rules
Off-topic or ad-like replies are treated as spam by members and admins. The more precise the prompt and the reply condition, the fewer the reports. What happens after reports — see the guide on spam blocks.
