Introduction

The "7. NLP Detail" report from the "CDR" report group allows you to view the real and detailed consumption of credits associated with the artificial intelligence components of wolkvox Studio that are executed in your routing points. It provides traceability by component, channel, session, and date/time. This report is especially useful for auditing costs, identifying which blocks consume the most, and validating the actual use of AI per channel.

Report Information

• TYPE_NLP: Identifier/name of the AI component or resource executed from Studio (for example, values such as getvars, model-router, nlp-wolkvox-inten, conversations-chat-61074, nlp-MrWizard-chat, nlp-voc-autoqa may appear in the view). This field helps you determine which specific block/service is generating consumption.
• NLP_CHANNEL: Channel from which the component's consumption originated or was associated (values such as whatsapp, web, chat-whatsapp, and _chat_ may appear in the view). It is useful for comparing consumption by channel (e.g., WhatsApp vs Web) and identifying where AI usage is concentrated.
• CREDITS_SEND: Number of credits consumed by executions associated with "sending" (the output/action part of the component), expressed as a numerical value. Useful for detecting components that consume more during the response/output phase.
• CREDITS_ANSWER: Number of credits consumed by executions associated with "answer" (the processing/resolution part of the component), expressed as a numerical value. Useful for identifying components that consume more during the analysis/processing phase.
• TOTAL_CREDITS: Total credits consumed in the record, expressed as a numerical value. In practice, it allows you to quickly compare "how much it cost" for that execution and prioritize optimization for the highest consumptions.
• SESSION ID: Identifier of the session linked to the execution/consumption.
• DATE: Date and time of the consumption record (timestamp). Allows analysis of peak demand times, auditing by periods, and correlation with operational events (peaks, campaigns, outages, flow changes, etc.).