Logging and statistics for the platform's AI usage: ElevenLabs text to speech (TTS), ElevenLabs speech to text (STT) and Claude LLM summaries. Every AI request is logged with cost (USD and DKK) and reseller attribution frozen at write time.
Access scoping applies to all usage endpoints: Telecom X staff (ADMIN and RESELLER_ADMIN) see everything including cost; white label resellers (RESELLER) are locked to their own customer tree (sub-resellers included) and never receive cost fields; everyone else (minimum MANAGER) is locked to their own customer.
AI is also rationed and billed: a customer that may use AI has an AI product (Products, type AI) with a monthly included amount of transcription seconds, summary tokens and text-to-speech characters, and it names the models the customer's requests run on. Usage is counted per customer per month in AIUsageCounters, and a request is refused with 402 quota_exceeded when the customer has no AI product or has used up that resource. The quota endpoints below show the status, the history and sell top-ups - by hand, or automatically every time a resource reaches 80 %.
The services are TTS, STT and LLM. The contexts (which platform feature triggered the request) are AUDIO_FILE, TTS_STREAM, STT_API, VOICEMAIL, VOICEMAIL_SUMMARY, RECORDING_TRANSCRIPTION, RECORDING_SUMMARY, VOICEMAIL_TO_EMAIL_TRANSCRIPTION, VOICEMAIL_TO_EMAIL_SUMMARY and UNKNOWN (legacy records migrated from the old TTSUsage collection).