Collony: AI Moderator for Telegram

A group with a thousand members lives by the laws of spam: casino ads in the morning, "quick money" at noon, a fake wallet link at night. The admin cannot keep up, and members leave. Collony is an AI moderator for Telegram that reads messages and clears the junk around the clock. 🛡️

It is a tool for chat owners and admins, not entertainment for an ordinary member.

What Collony does in your chat

• Clears spam and scams: removes ads, fraud offers and phishing links.
• Answers member questions: pulls replies from the knowledge base you upload.
• Watches the mood: shows what people argue about and what annoys them.

The platform also works in Discord, and the Telegram bot is tied to your account on the service's site through the /linkweb command.

Setup without a ten-page manual

1. Press Start in the bot, and it shows the first steps.
2. Add the bot to your group and make it an admin, otherwise it cannot delete messages.
3. Upload your material: FAQ, chat rules, a project description. The AI draws its answers from them.
4. Review early decisions and tune the settings to your community's tone.

If something goes wrong during setup, /cancel drops the current step, and /menu hints at what to do next. 🧭

How the AI decides what to delete

Simple filters look for stop words and trip over a typo. Here a model weighs the text, links, images and the author's behavior together, so a new scam scheme has a chance of being caught the first time.

Per the platform's description, the verdict can vary: allow, warn, delete or ban. Borderline cases go to a review queue where your team can forgive or ban with one click.

What to teach the bot so it earns its place

Automatic answers are only as good as the knowledge base behind them. Upload nothing and you get a polite bot that knows little.

• Chat rules: what is banned and what follows.
• Frequent questions: how to join, where to download, whom to contact.
• Tone of voice: strict, friendly or playful, to suit your community.

Once a month, skim what the bot told members. An outdated answer wears down trust fast. 💬

Sentiment tracking: what it gives you

With hundreds of members you cannot read every remark. Sentiment tracking helps you notice that people started complaining after a product update, or that one topic suddenly irritates the group.

Use this data as a hint. The decision about what to change in your project is still a human one.

Auto-moderation risks worth knowing

Any automation errs sometimes. A harmless message may get deleted, and a clever spammer may slip through. So start with mild settings and watch the action log.

Give members a clear way to appeal a wrong removal, and say in the rules that automatic moderation runs in the chat so people know what to expect. ⚖️