Skip to content

Model Discovery

The Node discovers AI model files by scanning a designated directory on its host machine. Any GGUF file placed in the correct location is picked up automatically — no configuration is needed beyond copying the file.

Model directories

The Node uses separate subdirectories for language models and embedding models. The base path depends on the operating system:

Language models (chat / text generation)

Windows 

C:\ProgramData\Parker Software\OptimaNode\Executors\Models\lammacpp\

macOS 

/Library/Application Support/Parker Software/OptimaNode/Executors/Models/lammacpp/

Linux 

/var/lib/ParkerSoftware/OptimaNode/Executors/Models/lammacpp/

Embedding models

Windows 

C:\ProgramData\Parker Software\OptimaNode\Executors\Models\lammacpp_embed\

macOS 

/Library/Application Support/Parker Software/OptimaNode/Executors/Models/lammacpp_embed/

Linux 

/var/lib/ParkerSoftware/OptimaNode/Executors/Models/lammacpp_embed/

The Node creates these directories automatically if they do not exist.

How discovery works

The Node scans each directory recursively, finding all files with a .gguf extension. For each file found, it reads the model's metadata from the file itself (using the GGUF format's embedded header) and builds a manifest describing the model — including its name, family, size, quantisation type, and context window.

These manifests are reported to the Gateway, which displays the discovered models in the Executors panel of the relevant node's detail view.

Adding a model

  1. Copy the .gguf file into the appropriate directory for its type.
  2. The Node will detect the new file during its next scan. If the Node is already running, this happens automatically within a short time — no restart is required.
  3. The model will appear in the Gateway's Executors panel as a new, offline executor.
  4. Enable it from the Executors panel to start the model process. See Executors.

Subdirectories are supported

You can organise model files into subdirectories within the discovery directory. The Node scans recursively, so files in any subfolder will be found regardless of how deeply they are nested.

Manually triggering a rescan

If a newly added model does not appear in the Gateway after a short wait, click the refresh button in the Executors panel of the node's detail view. This prompts the Node to re-scan the directory immediately.