Cost
A FinOps cockpit — unified spend across compute, LLM, storage, seats, and network, with allocation, budgets, and optimization.
Cost is proxifai’s FinOps surface. It lives at /cost (older /money URLs redirect here) and presents one unified view of spend rather than a separate bill per resource type.
The unified cost model
Every priceable thing in the platform emits a cost event, and Cost rolls those events up across five kinds of spend:
- Compute — Kubernetes workloads, priced from OpenCost + Prometheus
- LLM — model usage from the Models gateway
- Storage — persistent volumes and object storage
- Seats — per-user subscription cost
- Network — egress and transfer
The unified rollup is served from /api/v1/money; the compute slice is computed by the cost engine at /api/v1/costs (/summary, /namespaces, /pods, /containers, /trend).
The cockpit
Cost has four views:
| View | What it shows |
|---|---|
| Overview | KPIs, spend trend, month-end forecast, and composition by kind/provider |
| Allocation | Showback/chargeback by service, environment, provider, kind, cluster, namespace, model, workflow, or user — with export |
| Budgets | Spend caps with burn-down, forecasting, and anomaly alerts |
| Optimize | Savings opportunities — right-sizing, placement, and model-efficiency suggestions |
Scope awareness
With a service selected in the scope picker, every Cost view filters to that service (e.g. ?service=<id>), so you can answer “what does this service cost, and where does the money go?” without leaving its context.
OSS vs Enterprise
In OSS, cost is records-only: events are tracked and rolled up, but nothing is enforced and there’s no charge. In Enterprise, credit balances and budgets are enforced through the account service (a request can be rejected with 402 when a budget is exhausted). The LLM slice — usage tracking, per-key/scope budgets, and rate limits for the Models gateway — is documented in detail under Billing & Usage.
Budgets here cover all spend kinds. The narrower per-key/per-scope spend caps that throttle the LLM gateway specifically are configured in Models → Budgets and described in Billing & Usage.