knowmindknowmind
DokumentationQuickstartAnbindungReferenzFehlerbehebung
DEEN
knowmindknowmind
DocumentationQuickstartIntegrationsReferenceTroubleshooting
DEEN

Tasks

Budget and costs

The AI usage cockpit shows your consumption across all stored BYOK keys. You can set a monthly budget and configure warning emails at 50, 80 and 100 percent of consumption. The view includes token count, list-price estimate, and a breakdown by provider and model.

Audience
All users who have stored BYOK keys, especially in a business context where AI costs are budgeted.

Prerequisites

  • At least one BYOK key stored (see: Store a BYOK key)
  • At least one AI call made (otherwise the cockpit is empty)
  • An email address for the budget warnings (optional; otherwise the sign-in address is used)
Note

Estimate, not the final invoice

Cockpit costs are an estimate based on the provider's list prices. The actual invoice comes from the provider and may differ (volume discounts, cache discount, exchange rate). knowmind does not block anything when the budget is hit — the display is informational. For a hard cap, set spending limits at your AI provider.

Steps

  1. 1

    Open the cockpit

    In the dashboard navigate to AI usage cockpit (in the header navigation under "Analytics", labelled "AI usage").

    Outcome: You see total cost, call count, input and output tokens, and a timeframe switcher (7 days, 30 days, 3 months, 12 months).

  2. 2

    Set a budget

    In the "Monthly budget" block, enter a value in euros and an optional recipient address for the warning email, then click "Save".

    Outcome: A progress bar appears showing the current monthly consumption in percent. You receive a mail at 50, 80 and 100 percent.

  3. 3

    Analyse consumption by provider

    In the "By provider" block you see costs split across Anthropic, OpenAI, Google, Mistral and so on, with a bar visualisation. The "Detail by model" block lists calls, input tokens, output tokens, cache hits and cost for each provider-model combination.

  4. 4

    Switch the timeframe

    Use the timeframe switcher in the top right. At 12 months you also see quarterly and annual trends.

Verify the result

  • The cockpit shows values for calls and cost (provided AI calls happened in the chosen timeframe).
  • The progress bar shows the monthly consumption in percent.
  • After crossing 50 percent you receive a mail at the configured address (check your spam folder if needed).

Troubleshooting

Error messageCauseResolution
Cockpit shows 0 € and 0 callsNo AI calls in the chosen timeframe, or the timeframe is too narrow.Switch to 12 months. If it stays empty, verify that a BYOK key is stored and that knowmind is actually using it (try a document upload with enrichment).
Warning email does not arriveMail in spam, recipient blocks Brevo, or typo in the address.Check the spam folder, allow mails from info@schuebeler-consulting.de, correct the recipient address in the cockpit.
Cost differs significantly from the provider invoiceExchange rate (USD/EUR), volume discount, cache discount at the provider.The estimate is based on standard list prices — differences are expected. The provider's invoice is binding; the cockpit is a guide.
Budget exceeded, knowmind keeps calling the LLMknowmind does not block hard — the budget is a warning, not a limit.Set hard limits at the AI provider (e.g. monthly spending limit in the Anthropic or OpenAI account).

Related