_predict_recommend
POST /api/v1/_predict
{
"from": "tasks",
"where": {
"project_type": "construction",
"phase": "mep",
"task_name": "HVAC commissioning",
"region": "Helsinki",
"season": "summer"
},
"predict": "subcontractor"
}
How it works
Project Plan combines Aito flows on a single editable surface. The generative side runs aito.._predict per task on the tasks table for assignee_kind, subcontractor or assignee_person, planned_days, planned_cost_eur — drafting a plan from precedent rather than a template. Beneath each task it then fans out on the purchases table: a _search with $has overlap finds the product lines whose vocabulary matches the task name ("Steel erection" → "Steel erection batch"), and two _predicts give the historically right supplier and the estimated EUR amount for that exact line. Click any supplier chip to open the swap dropdown: Aito's history-ranked candidates (with the full $why popover) plus newly-registered suppliers piped in from the acquired supplier management system — a sales/distribution channel that lets suppliers reach the planner directly. Every cell is editable; the live KPI strip recomputes in place.
Learn more