---
title: "Routing"
id: "1818"
type: "page"
slug: "routing"
published_at: "2026-09-20T22:09:00+00:00"
modified_at: "2026-09-21T01:36:29+00:00"
url: "https://xedant.com/agents/telegram/docs/routing"
markdown_url: "https://xedant.com/agents/telegram/docs/routing.md"
excerpt: "Routing is the automatic triage of what has been collected: messages from sources receive categories,…"
---

# Routing

[https://xedant.com/agents/telegram/docs/routing.md](https://xedant.com/agents/telegram/docs/routing.md)

Routing is the automatic triage of what has been collected: messages from sources receive categories, target channels and turn into ideas in the queue by themselves. It saves you from manually scrolling the feed in search of value: the content finds its own way to your channels and queues.

## Categories

Categories are colored labels for sources and messages. A source with categories passes them to all of its messages automatically; messages without categories get them from routing.

## Routers — ways to determine a category

A router is a classification rule of one of three types:

- **Keywords** — case-insensitive phrases, the “any” mode (one is enough) or “all” (every one is needed), stop phrases (a message containing one does not match); on a match the router offers its categories;
- **AI (LLM)** — an OpenAI-compatible model assigns the text to categories from a list you define; a custom prompt can be set; the model cannot step outside your list of categories. An LLM is a large language model — the same class of technology as chatbots;
- **Script** — an external program receives messages on input and returns categories; a type for those who can program.

## The router chain

Enabled routers work in order, each new message is classified once — the first match wins and the rest are not asked. One router’s failure (a wrong AI key, for example) does not stop the others: the chain simply moves on. A message with no match stays without categories — and without a route.

## “Category → channel” rules

Each category can be given a destination channel with the “add to queue” action and, optionally, a target schedule slot. One category can have several channels — an idea is then created in each channel’s queue.

## What comes out of it

A message gets categories and a list of target channels. For every “add to queue” rule, an idea is created in the channel’s queue carrying the message’s text (and the target slot, when set) — from there it travels the pipeline like any other: draft, review, schedule. Routing runs once, at the moment the message is collected — and the result is visible right away, both on the message and in the queues.

## The test panel

Any router can be checked on a sample text: whether it matches, which categories it offers, what the engine answered and how long it took. The check changes nothing in the data — experiment as much as you like.

## What is next

Where the messages come from is the [Sources](/agents/telegram/docs/sources)
 section, how to read the feed is the [Message Feed](/agents/telegram/docs/messages)
, and what happens to an idea next is the [Content Queue](/agents/telegram/docs/queue)
.

[← Back to the documentation index](/agents/telegram/docs)