Cabi.Chat Radar

Cabi.Chat Radar is the graph's intent detection infrastructure. Seekers ask, offerers answer, and the radar is how a project finds the asks: it finds public messages in public Telegram groups where people ask for exactly what its campaign describes, and shows them to the project owner. It does not guess and it does not index for its own sake. Built as cheap text processing, so matching is never the expensive part.

Two stages

Words. Always on. You describe the message in keyword groups. Inside a group any word is enough; across groups every group must hit. One hit from the exclude list drops the message. Case, punctuation and emoji are ignored, and a keyword matches as a fragment, so rent also catches rental. Every hit lands in the radar's Matches as a post card: the text, the group, a link to the original message.

Meaning. Optional, and it needs training: at least five real matches in the radar's Canonical workspace. They set the bar: every new match is compared by meaning against them, and only the close ones are pre-qualified. An example unlike the rest is treated as an outlier and does not lower the bar. Auto outreach relies on this stage.

One request, one match. The same text pasted into twenty groups within a day is one match, not twenty, and near-duplicates fold into one.

Where it looks. The public groups you attach to the radar, one by one or as presets. Cabi.Chat Radar — Telegram group presets

What happens to a match. It is shown to you, or answered by an oracle: the radar's two modes. Cabi.Chat Radar — With or without an oracle

Lifetime. A stopped radar pauses from the next batch. Matches expire in ninety days unless caught again. Act on them while they are warm.

Writing and tuning the words: link
A radar is one kind of campaign: Cabi.Chat Campaigns