hermes-local-rig-accounting: model tool for Hermes Agent

What hermes-local-rig-accounting does

You describe your rig under local_rig in config.yaml: hardware cost, lifespan in years, optional GPU-only cost as the depreciation base, average power draw in watts and electricity rate per kWh, or set the rate to auto with a region for lookup. A benchmark run measures tokens per second, and the cost model divides depreciation and energy cost per hour by throughput to give dollars per million tokens. The README's example of a 1,500 dollar GPU at 450 W and 50 tokens per second works out to $0.62 per million tokens.

Slash commands /rig-benchmark, /rig-summary, /rig-cost, /rig-rates and /rig-submit, and matching LLM tools rig_cost, rig_summary, rig_benchmark, rig_rates and rig_submit, expose the data. Hooks on post_api_request, on_session_start and on_session_finalize track tokens from local providers such as localhost, LM Studio, Ollama and vLLM and ignore cloud API calls. A rigs list with hostnames supports multiple machines. The README says cost data stays local, with no telemetry.

Key features

Depreciation and energy cost model that yields dollars per million tokens
Benchmark command that measures tokens per second for a local model
Automatic electricity rate lookup by region
Multi-rig profiles selected by hostname
Tracking limited to local providers such as LM Studio, Ollama and vLLM
Optional submission of benchmarks to a community leaderboard

When to use it

Work out whether a local model is cheaper than a cloud API for your usage
Compare benchmark results for different models on the same rig
Track session cost on a desktop server and a laptop with separate profiles

Who it is for: Hermes Agent users who run local models and want to know what each token really costs them.

How it fits with Hermes Agent

Built for Hermes Agent as a plugin that hooks into API requests and session events and adds slash commands and LLM tools.

How to install hermes-local-rig-accounting

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

hermes plugins install GumbyEnder/hermes-local-rig-accounting

Note: The auto_submit setting defaults to true and, when gh is authenticated, submits benchmark results to a community leaderboard after /rig-benchmark.

FAQ

What is hermes-local-rig-accounting?

hermes-local-rig-accounting is a Hermes Agent plugin that calculates the cost per token of local LLM inference using your hardware cost, lifespan, power draw and electricity rate.

How do I install hermes-local-rig-accounting?

Run hermes plugins install GumbyEnder/hermes-local-rig-accounting, then add it to plugins.enabled and a local_rig section in config.yaml. You can also clone it into ~/.hermes/plugins/local-rig-accounting.

Is hermes-local-rig-accounting free and open source?

Yes. The repository is published under the MIT license.