Roster
Volunteers
| Name | Roles (comma separated) | Shift limit |
|---|
Shifts
| Name | When | Role needed | Seats |
|---|
Availability & preference
Click a cell to cycle: off → available → preferred. Only qualified volunteers (matching role) can be marked.
Reuse this roster
Save the current roster, edits included, under a new name. Next week you open the copy, adjust times and people, and solve again instead of rebuilding.
Same seats, different priorities
The solver maximises filled seats within the model you give it. The fairness weight only changes who carries the load, by making each extra shift for the same person cost more. Both runs are optimal for their stated cost model.
Assignment flow
Volunteers on the left, shifts on the right, one line per assignment (bold = preferred slot). People highlighted in amber are at their shift limit — they are the binding bottleneck the min cut reports.
Bottleneck report (min cut)
Capacity check
Before solving: for each role, seats demanded vs the most its qualified volunteers could cover, given their shift limits and the slots they actually marked.
Exports
Import validates the file first: a rejected file changes nothing, and when an import succeeds your previous work is kept under My rosters.
Audit preview
What this is
FairFlow assigns volunteers to shifts with a min-cost max-flow solver (successive shortest paths with Johnson potentials), implemented from scratch in about 200 lines with no libraries. Everything runs locally in your browser: no account, no server, no data leaves the page.
The solver is optimal for the model you give it: role requirements, one seat per person per shift, personal shift limits, and three availability states. It does not model overlapping-time conflicts between shifts, travel time, seniority rules or labour law. Those are modelling choices for you, not things the solver can guess. A heuristic warning layer flags assigned pairs whose slot labels parse to overlapping times, and lists the pairs it could not parse as unchecked — it never blocks an assignment. Ties between equally good assignments are broken deterministically by input order.