[ MODEL ALIASES ]

[ HARNESS CONFIGS ]

Every model in llama-swap is included, with its real context window and output cap, plus default, which follows the model loaded from this dashboard. Replace YOUR_API_KEY (pi) or set LLAMACPP_API_KEY (opencode). On the LAN, http://192.168.2.61:9292/v1 needs no key.

[ GPU ELECTRICITY ]

Estimated from the GPU's own board-power sensor, sampled every 5 s: the card only, not the CPU, the rest of vader or power-supply losses (add roughly 10% for those). Time the dashboard wasn't running is left out; covered says how much of the month has readings.

[ LLM USAGE ]

[ BOOT ARGS ]