What problem does it solve? Tracking LLM token consumption across multiple models and providers is error-prone when relying on hardcoded model lists, and infrastructure stats like Pi-hole ad-blocking become inaccessible when web APIs fail. This Skill provides dynamic monitoring of Hermes LLM usage alongside server and network observability. ## Core Features & Use Cases - Dynamic Model Discovery: Parses ~/.hermes/config.yaml to retrieve the active model chain instead of relying on hardcoded lists, and sorts usage by consumption percentage with unconfigured models last. - Token Usage Reporting: Runs monitor scripts and reports a calculated total of tokens used across all models for clear consumption visibility. - Pi-hole Stats via FTL Database: Queries /etc/pihole/pihole-FTL.db directly with SQLite when the Pi-hole API returns 404, correctly counting GRAVITY (status 1) blocks while excluding GRAVITY_CNAME (status 17). - Grafana Alloy Deployment: Guides metrics and log scraping setup on standalone Linux (including Alpine) with node_exporter-style metrics and journal logs. - Use Case: When a user asks for a model usage report, run the monitor script, sum the used tokens, and include the total in the summary so consumption state is immediately visible. ## Quick Start Ask the assistant to show current LLM token usage across all configured models including a total consumption summary.
