SubLaneSubLane

Everyday use

Check a request, read usage, and find the next relevant troubleshooting step.

Use this guide after your first request works. Viewing requests and usage does not require an allocation rule.

Check a request

Open Requests to see your own calls. Locate the time of the call, then inspect its result, model and duration. Administrators can use All requests to inspect team calls. Keep the request ID when asking your administrator to investigate.

The history contains request metadata, not prompt or response bodies. For filters and diagnostic details, see request history.

Example personal request history with synthetic records:

Personal requests table with outcomes, models and duration

See usage

Open Overview, then choose a time range in Your usage. Start with request counts and reported input/output tokens. Administrators can open Administration → Team usage to compare members, or Administration → Instance status for full gateway setup guidance and service health.

A missing token report means unknown usage, not zero. These statistics describe past calls; they are not a bill or a subscription's remaining quota. Detailed chart definitions are in usage summaries.

Example administrator team usage view with synthetic totals:

Team usage totals and daily request chart

When a call fails

What you seeNext action
Invalid, paused or expired keyCheck the key in API keys. Ask the administrator if your member account was disabled.
No authorized pool or modelAsk the administrator to check your workspace membership, direct pool grant and model policy.
Busy or rate-limitedWait for the indicated retry time. If recurring, ask the administrator to inspect request/account concurrency.
Upstream quota exhaustedAsk the administrator to inspect the account's remaining quota and reset time.
An old conversation stops after an account changeExisting conversations keep their subscription. Start a new conversation after the intended account change.
A configured allowance is exhausted or waiting for syncCheck your resource balance and ask the administrator to inspect the allowance rule. This applies only when limits were configured.

For Codex subscription exhaustion, an administrator can inspect reset cards and execution results. Card use does not replenish member allocation budgets or replay the failed model request; retry from the client after the quota recovers.

Only add controls when a need appears

If shared access works for your team, no further configuration is required.

Model prices

Open General → Model prices (/pricing) to inspect the current merged reference catalog. Administrators and members can search models and view input, cached-input and output rates in USD per million tokens, with source labels. Refresh reloads the catalog view; it does not force an upstream price synchronization. A reload failure retains the last loaded data with a warning. GET /api/pricing requires a browser session. Prices support request cost estimates and new allowance revisions; existing revisions keep their saved rates. A listed price grants no model access and is not an actual subscription charge. The offline demo shows an empty catalog.

Export CSV

Use Export CSV in request history and usage views. Request exports preserve current filters and include up to 5,000 retained records, across pages. Personal exports retain account-identity redaction. Usage exports preserve the selected reporting period, totals, daily rows and the existing top-20 breakdowns.

Files use UTF-8, stable English column names and UTC timestamps. Unknown token counts are blank and reporting-coverage counts remain available. Reference costs are estimates, not invoices. Formula-prefixed text is neutralized for spreadsheet import; prompts, responses and credentials are never exported. Downloads stop when the user/workspace changes or the page unmounts.