# Meet Match

> 1:1 meetings at your event, matched honestly and booked without collisions. Enter the two sides you would pair, what
> each consented to, the meeting block in the published grid and any meeting number a sponsor package already names,
> and get a decision on whether to run 1:1 meetings at all, the matching mechanism with its criterion written as a
> sentence, a booked timetable that never double-books or crosses a session, the quota a package can honestly commit to
> and the no-show handling - then the one meeting brief to publish and the confirmations to send each pair. Built on
> the SkillSafe skill @samber/event-b2b-matchmaking, from samber/dev-event-organizer-skills (MIT), by Samuel Berthe.
> Answer no by default: developer attendees are not buyers, registration is not consent, and the meeting block is a
> guest in the grid. It never quotes a vendor or platform benchmark, never invents a person, an organization, a slot,
> a fill rate, a meeting count or a consent, and labels its own numbers self-set.

https://meet-match.skillsafe.ai/

## What is free
- Whether to run the program at all (Q1, Q2): no named number and no second population is "do not run it".
- What may not move, written down: the signed meeting number, the published grid, the consent wording each side saw, the floor plan, the room block - each with the sibling it routes to.
- The consent gate: only people whose consent covers matching or contact are matchable; a directory listing, a registration or nothing is excluded and counted, never re-purposed; a side whose wording was never shown holds the verdict at "settle consent first".
- The matching mechanism, scheduling mechanic, quota commitment posture and no-show handling menus re-ranked from the interview, every deleted rung struck, and the matching criterion written as a sentence.
- A timetable laid inside the block: equal slots with a buffer, never inside a published session, tables as parallel capacity, no person or table double-booked, availability respected, every unplaced pair listed with its reason.
- Every named meeting number held against what the block can hold for that sponsor - over-commitment flagged, never a benchmark.
- A verdict: do not run it, settle consent first, fix first, ready to publish the brief.
- Exports: the timetable (.csv and a per-pair .ics), the unplaced pairs (.csv), the generic outcome record (.csv), the brief's skeleton (.md), everything (.json).

## The metered lanes
Two tasks on gpt-terra, metered per run, routed by `task`. Input string fields: `task` (design or brief), `event`, `sides`, `block`, `notes` (may be empty) and `facts`.
- `design` replies with `summary`, `verdict`, `run`, `fixed`, `consent`, `mechanism`, `scheduling`, `timetable`, `unplaced`, `quota`, `no_show` and `caveats`; the page holds every pick, count, row and posture to its own.
- `brief` replies with the one meeting brief - `title`, `who_meets_whom`, `when_and_where`, `how_to_book`, `missed_slot`, `contact`, `consent_note` - plus `confirmations` for each placed pair and a `note`.

## Links
- API tutorial: https://meet-match.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@samber/event-b2b-matchmaking
- Why a sponsor segment buys: https://sponsor-why.skillsafe.ai/
- The expo floor: https://expo-floor.skillsafe.ai/