For the complete documentation index, see llms.txt. This page is also available as Markdown.

Analytics

Admin Analytics show data about consumption, usage and configurations. Use it to analyse activity metrics over time.

Overview

Analytics events are aggregated into hourly buckets ready for analysis. Deleted records are consolidated into "Deleted" rows so data-retention rules are always honoured, and where your organisation requires it, personal details such as user names and e-mail addresses can be obfuscated based on permissions.

Every row expands to reveal further detail, and reports only cover periods where activity was actually recorded.

Consumption

Everything related to LLM consumption, grouped into the categories you already know from working in Blockbrain.

Bots — Filter and rank bots by compute blocks, active users and usage trend. Each bot also reveals its configuration, the models it has drawn on over time, model share by queries, and engagement metrics such as active users, conversations and messages.

Users — Model usage per user, measured both by consumption and by requests, and broken down by action, bot, API key or agent.

Knowledge Bases — Configuration details that explain what drove consumption, alongside ingest activity over time.

LLMs — Compute blocks, requests, and input and output tokens over time, with share attributed to bot, agent and action.

Agents — Model share by queries, plus compute blocks, requests and active users over time.

Actions — The detail adapts to the action itself: the agent or bot behind it, the knowledge base involved, and usage broken down over time.

API Keys — Usage per key, surfaced through whichever of the categories above reflects what the key has been used for.

Usage

Statistics on invocations that don't necessarily consume tokens. Tools, MCP, Custom API and Skills each share the same view: invocation share by bot, by agent and by user, with matching breakdowns over time.


The biggest change is collapsing the four identical Usage subsections into one sentence — repeating the same two lines four times reads as an oversight. If your docs structure needs those four headings to exist separately for navigation, I can restore them with a pointer back to the shared description instead.

Last updated