GenVM Configuration

You need to set up an LLM for your node to use to provide answers to natural language prompts. You can use any LLM you wish, however the quality of its answers will affect the performance of your node.

The GenVM configuration files are located at third_party/genvm/config/

genvm-module-llm.yaml

This is the configuration file of the LLM module. In this file you can set up and configure various LLM providers, as well as the system prompts of your validator.

You should not need to modify this in general.

However, from here you can:

• turn on and off various LLMs by setting the enabled field to false: By default they all come enabled for you to use. You will get warnings in the logs for each one that's enabled and not configured. Disabling non used LLM providers will hide those warnings

Note environment variable names for LLM API keys (e.g., HEURISTKEY, COMPUT3KEY, IOINTELLIGENCE_API_KEY). You will need to ensure the appropriate key is correctly set before running the node.

genvm-module-web.yaml

This is the configuration of webdriver module that enables the GenVM to access the internet. You should not need to modify this.

genvm-manager.yaml

This is the configuration of the GenVM manager. You should not need to modify this. However, if you are running the node on macos (natively), you may want to set permits to some large value, as auto-detection is not supported there yet, such as 32.

Greyboxing LLMs

Greyboxing is a way to further customise your LLM setup to improve its performance as well as security.

genvm-modules llm provides user with ability to customize greyboxing via lua scripting. Right now users can customize prompt templates, specify temperature and system prompt.

Related scripts are located at:

1. ./scripts/genvm-greyboxing.lua – user defined script
2. ./share/lib/genvm/greyboxing/lib-greyboxing.lua – more low-level library