← Back to blog

LumIntelligence is a Skill Now

LumIntelligence engine is available to try on your own device now. We are living up to our LLM-agnostic goal to let you configure your own private environment. This iteration is also built to be integrated with Hermes Desktop as a skill, so you can ingest your files and folders to be part of your personal knowledge base.

The default LLM for the engine is Hermes3 for a balanced experience between efficiency and performance, but you can always replace it with your favorite model. During querying, the agent will produce answers with citations. But wait, don’t other methods produce citations too? Then allow me to share this paper about post-hoc citation in AI hallucination: Generation-Time vs. Post-hoc Citation: A Holistic Evaluation of LLM Attribution. We traverse the network first before producing answering packages, so the agent only synthesizes a response after receiving a set of distilled information.

We also improved the system to support incremental ingestion at scale. This means your compute time doesn’t grow exponentially with your dataset. Do your files have a custom glossary that no one else knows and that causes issues with your agent’s understanding? Now you can customize your glossary so your knowledge network is tuned to your personal context. If our documentation is too long to read, I’m sure your agent is more than happy to read it for you.

The repo is live here: (https://github.com/LumIntelligence/LumIntelligence-hermes-skills).

Have fun!

Yours in illumination,

Peter @ LumIntelligence

Hermes Desktop and Hermes3 are products of Nous Research, Inc.