One scenario answering in every channel
A shared beginning and separate endings: the question and the answer are one, the formatting is per channel.
What the scenario looks like
- Execute + Data
- LLM
Copying a working bot for a second channel is tempting, and a month later there are three copies whose prompts differ because they were edited one at a time. There should be no copies: only the ending differs, and the channel is an ordinary branch.
What is inside
- Question from any channel — the entry. The scenario does not know where the question came from until it is asked.
- Model’s answer — the agent. Its prompt asks for no formatting: the next step adds that, and it differs.
- Where the question came from — a branch on
data.trigger. The platform puts the door the run came through in there. - Three endings with different formatting and an exit each.
Formatting differs everywhere, which is exactly why the endings are separate: Telegram
renders _italics_, the site renders Markdown with links, and VK shows formatting
characters as typed — so there the text has none.
What to set up for yourself
- The agent’s prompt — one for all channels, and that is the whole point: it is edited in one place.
- The endings. They currently differ only in formatting. Differences of substance belong here too: a link to a form on the site, a “type operator” hint in a messenger.
- A new channel is one more branch, not another copy of the scenario.
What to connect
Nothing, to try it: it works on the site straight away. The messenger branches come alive once you connect a Telegram bot or a VK community and enable the matching targets on the Publish tab. Until then they simply never fire, and the scenario does not mind.