Usage, Cost and Logs in Torki Platform

Overview

Torki Platform shows what your API traffic did and what it cost: per day, per model, per key and per workspace, down to each request. Usage, cost and logs need the Developer role or higher.

In every response

Each response has a usage object with prompt_tokens, completion_tokens (thinking included) and total_tokens — what the request is billed on. In a stream, send stream_options.include_usage to get it in the last chunk. Each response also carries a request-id header, the id you'll find in Logs.

The pages

Dashboard

Your credit, what you have spent this month against your monthly limit, and requests by model.

Usage

Tokens and requests by day, across your keys and the Playground. Input tokens, output tokens (including reasoning tokens from thinking models, billed as output) and tokens per request, with a Breakdown by model, key or workspace.

Cost

What your API usage cost, by day, at each model's price: the total for the period, the change against the period before, a day on average, and this month's spending against your limit. Invoices links to Billing.

Caching

How much of your input was read from cache. Cache reporting is switching on soon, and cached-input pricing is coming: until then, cached tokens bill at the input price.

Limit usage

How close you came to each rate limit in the last 24 hours, the requests refused with 429, and any lower limits a workspace sets for itself. Limits and tiers shows what your tier allows.

Logs

One row per request, kept for 30 days: time, status, model, tokens, latency and cost. New requests appear as they arrive; click to pause. Filter by status (Succeeded, Errors or 429 only), model, key (or Playground), workspace, tier and date range. Click a row for Request details, with Copy request id and Filter by this key.

Filters and export

Usage and Cost share the same filters: workspace, key (or Playground), model, tier and a date range in UTC. Export CSV downloads one row per day, group and tier, in UTC, following the filters and the breakdown you picked.

Good to know

• Requests appear within a minute of each call.
• The Playground's runs are counted with your API traffic and filtered as Playground.
• Refused requests appear in Logs with their error code and aren't billed.
• Logs keep metadata only. Prompts and completions are never stored in them, and the full error message goes back only to the caller.

FAQ

I can't see Logs or Usage.

They need the Developer role or higher. Ask an admin of your organization to change your role in Members.

Can I see a request's prompt or answer?

No. Torki doesn't keep prompts or completions in logs. Record them in your own application if you need them.

How do I see one key's traffic?

Filter Usage, Cost or Logs by that key, or open one of its requests in Logs and click Filter by this key.