Skip to main content

Token Usage & Cost Monitoring

Qarion tracks token usage across all AI interactions — Copilot chat, description generation, and field documentation. This helps administrators monitor consumption, estimate costs, and set budget controls.

Overview​

Every LLM interaction is logged with prompt and completion token counts, the model used, and the feature that triggered the call. This data powers the usage dashboard in the admin panel.

Viewing Usage​

AI Logs Tab​

Navigate to Administration → AI Logs to view a paginated log of all AI interactions:

ColumnDescription
FeatureWhich AI feature was used (chat, description, field description)
ModelLLM model identifier (e.g., claude-sonnet-4-20250514)
Prompt TokensNumber of tokens sent to the LLM
Completion TokensNumber of tokens in the LLM response
TimestampWhen the interaction occurred

Usage Summary​

The Usage tab provides aggregated views:

  • Daily breakdown by model and feature
  • Trend visualization over configurable time periods (7, 14, 30 days)
  • Cost estimation based on model pricing

Cost Estimation​

Qarion estimates costs based on standard model pricing. The estimates are calculated from tracked token counts and displayed alongside usage metrics. Actual costs depend on your LLM provider's billing.

Best Practices​

Monitor Regularly​

Review the usage dashboard weekly to identify unexpected consumption patterns. Sudden spikes may indicate excessive automated generation or unexpected user behavior.

Optimize Token Usage​

  • Use specific prompts for description generation to reduce token consumption
  • Leverage cached descriptions — regeneration is only needed when underlying metadata changes significantly