Cabi.Chat Search — Keywords, meaning and intentions
A seeker rarely knows your words. They know their intention: "someone to fix my bike tonight", "a math tutor for my kid". Search reads every query in three ways.
Exact. A card's title, its @handle, or a cabi.chat link pasted as is. Exact hits always come first, so a person who knows your name finds you at once.
Keywords. The words of the query are matched against titles and texts. A title that holds every word of the query ranks highest. The last word may be unfinished, "plumb" finds plumbers, and small typos are forgiven.
Meaning. The query is compared by meaning with what each card says: its title and description. "Someone to fix my bike" finds a bicycle workshop that never wrote the word "fix".
Cabi.Chat Search, the default engine, runs keywords and meaning together and fuses both lists. A card found by both ranks above a card found by one; keywords weigh more than meaning; Telegram audiences weigh less, so projects and articles lead. The search page lets you switch the engine: Keywords only, or Vector, meaning only, which needs a prompt.
Filters narrow, they never widen:
- Card Types: projects, articles, audiences, posts, boards, docs, sheets, workspaces, rubrics.
- Rubrics: Cabi.Chat Search — Rubrics and discovery
- Exclude words: hide results that contain any of them.
A card type or a filter alone is enough to start a search, no query needed. A geo preference orders the results instead of narrowing them: Cabi.Chat Search — Geo filter and preference
Write to be found. Put what you do in the title: the service, not a slogan. Open the description with the problem your client comes with, in the words they use for it. Search by meaning then connects their query to your card even when the words differ. Search engines read the same text: Cabi.Chat SEO — Getting indexed by search engines
Search as a whole: link