Costs

Query compute spend, approval queue, and cost breakdown by GPU type and experiment.

Costs API

The Costs API surfaces lab compute spend: GPU hours, LLM API calls, and storage. It also provides the live-burn view (pods currently running) and the approval queue (high-cost experiment requests awaiting Captain sign-off).

Get Cost Summary

labId · string · required

Convex ID of the lab.

curl

curl "https://www.hubify.com/api/v1/costs?labId=$LAB_ID" \
  -H "Authorization: Bearer $HUBIFY_TOKEN"

summary · object · required

Aggregated cost data.

Show Summary fields

totalUsd · number

All-time spend in USD.

totalGpuHours · number

All-time GPU hours.

byGpu · object

Per-GPU-type breakdown: { "H200 SXM": { count, costUsd } }.

liveRunning · object

Live pods currently burning compute.

Show Live running

podCount · number

Number of active pods.

costUsd · number

Accrued cost since each pod started.

hours · number

Total GPU hours running right now.

byGpu · object

Per-GPU live breakdown.

experiments · object[] · required

Per-experiment cost records (most recent 200).

Show Cost record

experimentId · string

Linked experiment ID.

estimatedCostUsd · number

Estimated cost when submitted.

actualCostUsd · number

Final cost after completion.

gpuHours · number

GPU hours used.

approvalStatus · string

One of pending_approval, approved, denied, auto_approved.

createdAt · number

Unix ms.

The liveRunning.costUsd field is derived at query time from running pods’ hourly rates. It shows your current exposure even before experiments complete.