Cost Dashboard

The Cost Dashboard at /projects/{p}/cost answers one question: where is the money going, and how much is the cache saving?. It joins every row in gateway_logs against the per-endpoint estimated_cost_per_1k_tokens_usd rate and rolls up the results.

What it shows

Spend (24h / MTD) with last-month comparison.
Cache saved (24h / MTD) — money the cache prevented from leaving the bank.
Cache save % — share of would-be MTD cost served from cache.
Spend / day (30d) stacked chart with billed vs cache-saved bars.
By endpoint (30d) — requests, tokens, rate, 30-day cost, cache hit %.
By provider (30d) — token volume + 30-day cost.
Top tokens (30d) — which API tokens drove the spend.

How “cost” is computed

For each gateway_logs row:

billable_tokens = total_tokens   (or 0 if cache_hit)
cost            = (billable_tokens / 1000) × endpoint.estimated_cost_per_1k_tokens_usd
saved           = (cached_tokens  / 1000) × endpoint.estimated_cost_per_1k_tokens_usd  (when cache_hit)

If an endpoint has no rate set, its rows contribute zero cost — they still show up in token counts so you can spot rate gaps.

Limitations

Estimated, not invoiced — provider invoices include taxes, contracts, and rounding the dashboard doesn’t see. Use it for trend, not bookkeeping.
Per-endpoint flat rate — input vs output token pricing differences aren’t modelled (most provider price sheets do separate them). If you need that fidelity, set the rate to a blended weight.
Retention applies — Community Edition keeps 7 days of gateway_logs. The 30d window naturally truncates there until the row count grows.

© Akyros Labs LLC. All rights reserved.