Project support first
Verity JE links a maintainer Ollama tutorial, and Revamped lists Ollama as a provider. Follow the selected project's configuration fields; a generic Ollama installation does not prove the mod points to it correctly.
Install and start Ollama
Use the official Ollama distribution for the operating system, start the local service, and install a model that the computer can run. Test that model through Ollama itself before connecting Minecraft.
Connect the project
Select Ollama or the project's local provider option and use the documented local endpoint/model field. Keep the service on localhost unless you intentionally understand and secure remote exposure.
Failure map
Connection refused normally means the service is not running or the endpoint is wrong. Model-not-found means the configured name is unavailable locally. Very slow or failed responses can indicate insufficient memory, an oversized model, or a long context.
Performance baseline
Begin with a small model and one text prompt in a clean instance. Record model name, available memory, response time, project release, and whether voice is disabled. Add larger models and voice only after text is stable.
Privacy boundary
Local inference reduces reliance on a cloud provider, but the Minecraft project, logs, configs, and any optional voice services remain separate data paths. Do not expose the Ollama port to the public internet without authentication and network controls.