Analytics

Analytics shows account usage, spend, invoices, and detailed billing listings for a selected time range. Use it when you need to understand where usage came from, compare model or SKU consumption, investigate unexpected spend, or reconcile invoices with activity.

Analytics is different from the lighter Usage page. Use Usage to check current balance, storage, plan limits, and subscription-model quotas when your plan includes them. Use Analytics when you need historical charts, grouped usage, request-level listings, or invoices.

Analytics is scoped to the active account shown in the console. Before reviewing spend, adding credits, or sharing screenshots, confirm that you are in the correct account, especially if you work with multiple accounts.

When to use Analytics

Open Analytics when you need to:

Review account spend for a specific period.
Compare usage by model, SKU, or resource.
Investigate which requests produced a cost.
Check invoice history and payment status.
Add credits from the invoices area.
Prepare a usage summary for operations, finance, or support.

Analytics is not a conversation debugger. If you need the full prompt, response, tool calls, metadata, or error details for an inference run, use Conversations and Logs when those pages are available in your workspace.

Choose a time range

Use the List to date date-range selector at the top of the page to control the reporting period. The picker supports manual From and To date-time values and quick ranges:

Past 30 minutes
Past hour
Past 3 hours
Past week
Past month
Past year

Changing the range refreshes the Analytics data. Use short ranges when diagnosing a recent spike; use longer ranges when reviewing trends, model mix, invoices, or monthly usage.

The selected period cannot exceed one year. For older investigations, compare smaller ranges separately and keep notes outside AIVAX if you need a long-running finance reconciliation trail.

Usage listings may take up to 10 minutes to update. If a new request does not appear immediately, wait and refresh the selected period before assuming it was not billed or recorded.

Analytics is optimized for investigation and summaries, not complete ledger export. High-volume periods can be capped in the console: usage listings show the latest rows, model and SKU groups show top groups, and resource groups show a bounded set per resource type. If a view looks incomplete, narrow the date range around the time you are investigating.

Read Discriminations

The Discriminations tab is the main dashboard for grouped usage.

In Analytics, discriminations are grouped breakdowns of usage by model, SKU, or resource.

The Account section shows:

Storage usage: Current account storage used by stored resources.
Period spend: Total usage amount for the selected period.

The Inference section shows:

Volume: Total measured usage volume for the period.
Input tokens: Tokens sent to models.
Output tokens: Tokens generated by models.
Cached tokens: Input tokens served through cache-related billing categories.
Cache hit rate: A derived cache metric for the selected period.

Use these cards as directional indicators, then inspect the charts and tables to understand the drivers. For example, a period can show low spend but high token volume when the active models or SKUs are priced very low, discounted, included, or otherwise billed at a small amount.

Usage per model

Use Usage per model to see model spend over time and model-related token counts in the table. The chart is spend-oriented, while the table groups model name, SKU, token count, and total usage for selected inference categories. Non-token or service-specific charges may be easier to understand in Usage per SKU or Usage per resource.

Usage per SKU

Use Usage per SKU to understand the billing category behind usage. SKUs separate input, output, cached input, embeddings, RAG answer generation, and other billable resource types. This view is useful when the model name alone does not explain the cost pattern.

Usage per resource

Use Usage per resource to see usage grouped by account resource. This helps connect spend to a feature area or object, such as a gateway, collection, batch job, or other resource that generated usage.

If a resource name or ID is unfamiliar, copy only the minimum identifier you need, search the related feature page by ID or name, and keep the selected time range with your notes. When asking for support, provide redacted IDs and the time range instead of full screenshots with account, invoice, prompt, or customer data.

Read Listings

The Listings tab shows the latest balance releases in the account. Each listing can include:

Usage listing ID
Request ID
Moment
Description
SKU
Quantity
Amount

Rows are grouped by date and request. When multiple SKU rows belong to the same request, the table shows a request total after the group. Use this tab when a chart shows a spike and you need to inspect the request-level billing entries that contributed to it.

Treat request IDs and usage listings as operational identifiers. They are useful for support and internal investigation, but they should not be posted publicly with account, customer, or prompt details.

Read Invoices

The Invoices tab shows invoice history and credit-related actions.

The My invoices table includes:

Invoice ID
Date
Expiration
Description
Amount
Status

Use Add credits when you need to add balance to the account. If an invoice is expired or unpaid, create or complete a current top-up flow instead of relying on an old invoice.

Add credits safely

Add credits starts a credit addition and payment flow. It can create or update invoice records, but credits should be treated as available only after payment or processing succeeds. Confirm the result in two places: the invoice status in Analytics and the current balance on the Usage page.

Do not reuse expired invoice links as proof of available balance. If the payment flow is interrupted, return to Invoices and Usage before retrying a workload that depends on the new balance.

Investigate unexpected spend

When spend looks higher than expected:

Set List to date to the smallest period that contains the spike.
Open Discriminations and check whether the increase is concentrated in one model, SKU, or resource.
Open Listings and find request groups around the time of the spike.
Use the request time, resource, model, and SKU to continue the investigation in related pages, such as AI Gateways, RAG Collections, Batch, Conversations, or Logs.
Contain the source before retrying or rerunning work.

Do not assume every high token count is a high bill. Compare token volume with Total usage and Period spend.

Use the source-specific containment path:

Source	Immediate containment
Batch job	Pause the job before retrying or importing more items.
AI Gateway or direct model traffic	Stop the caller, review routing, and rotate or revoke the calling API key if the traffic is unauthorized.
RAG Collection	Stop the caller that is querying the collection, then review the gateway or API integration using it.
Chat Client or messaging integration	Disable the integration, delete exposed sessions, or stop the channel that is sending traffic.
Unknown source	Narrow the time range, inspect Listings request IDs, then continue in Logs or the related feature page.

Reconcile invoices and account balance

Use Analytics and Usage together:

Analytics Invoices shows invoice history and payment status.
Analytics Discriminations shows spend for the selected period.
The Usage page shows current balance, plan, included resources, and quota-style limits.

These views answer different questions. A paid invoice can add credits, but period spend depends on usage during the selected range. Current balance depends on both credits and subsequent usage.

Analytics can expose account names, invoice IDs, payment status, balance, spend patterns, request IDs, resource IDs, model usage, and operating times. Before sharing screenshots, copied table rows, browser URLs, or support notes, redact unrelated invoices, account identifiers, API keys, customer data, prompts, and any request details that are not needed for the investigation.

Troubleshooting

Problem	Likely cause	What to do
Recent usage is missing	Usage listings can take up to 10 minutes to update	Wait, refresh, and choose a range that includes the request time.
Spend looks too high	A model, SKU, tool, RAG query, batch job, or retry loop increased usage	Narrow the time range, check Discriminations, then inspect Listings and the related feature page.
Token volume is high but spend is low	The selected model or SKU may be low-cost, cached, included in plan limits, discounted, or otherwise priced differently	Verify current plan and pricing, then compare token count with Total usage and Period spend instead of judging by tokens alone.
Cache hit rate looks unusual	The selected period may include small samples, cached categories, or mixed request types	Compare a longer period and inspect Usage per SKU before making billing decisions.
An invoice is expired	The payment window for that invoice has passed	Use Add credits to start a current top-up flow.
A support case needs evidence	Screenshots or listings may expose account activity, IDs, invoices, and usage patterns	Share only the minimum IDs and time ranges needed. Redact customer data, prompts, API keys, and unrelated account details.

API reference

For programmatic account balance, plan, storage, and quota status, use the published account balance endpoint. This endpoint is for balance and quota status, not for reproducing the full Analytics tabs.

Detailed Analytics charts, invoices, and request listings are presented in the console. The console uses an authenticated internal usage endpoint for these Analytics views; it is not currently published as a public API reference endpoint. Use the console when you need the same grouped views described in this page.

Table of Contents