Log Mic Lab / Connected ideas
Understand model consumption without obscuring its accounting boundaries. These articles connect an AI request history to the usage metadata reported around it. They examine attempts, correlation, missing quantities, overlapping detail categories, and the difference between a usage observation and a cost estimate.
Start with AI logging if the application’s operation boundaries are still unclear. Then use the token guide to define the accounting record and its reconciliation rules. Reported usage, estimated usage, and unavailable usage should remain distinct enough for a later reader to explain the resulting totals.
2 connected guides
Trace a model request through timings, retries, tool calls, configuration changes, and outcomes without collecting every message.
Read AI logging: what to capture around a model request
Build interpretable LLM usage records with clear totals, missing-value handling, deduplication, and transparent cost estimates.
Read Token usage logging without storing secretsExplore a shared question across different kinds of logs. Each collection brings together relevant guides and a useful reading path.