Ollama Create NodeJS

Introduction

NodeJS CLI script which simplifies adding GGUF models to Ollama by creating symlinks and downloading necessary metadata from the Ollama Registry.

Benefits:

• Avoids model duplication within Ollama.
• Easy integration of GGUF models.
• No dependencies besides NodeJS.

Installation

Requires NodeJS version 18.11.0 or higher. Install it globally using npm:

npm install -g ollama-create

Usage

Usage: ollama-create [OPTIONS] [MODEL...]

Create a new Ollama model based on a base model from the Ollama registry.
MODEL can be a local GGUF file or the name of a model in the Ollama registry.

All specified files will be symlinked to Ollama to avoid duplication.
If a file is supplied which doesn't exist, it will be created from the base model in the specified location.
To remove a setting from the base model, use --no-<option>.

OPTIONS:
  -f, --from <from>           Defines the base model from the Ollama registry to use.
                              Defaults to the model architecture from the GGUF metadata.
  -n, --name <name>           Set the name for the new model.
                              Defaults to <name>-<size>-<finetune>-<version> from the GGUF metadata
                              or the base model name if no model file is specified.
  -d, --dir <dir>             Download all base model files to the specified directory and create symlinks for Ollama.

  -p, --params <json file>    Specify a JSON file containing parameters for Ollama.
  -m, --messages <json file>  Provide a JSON file containing the message history.
  -t, --template <txt file>   Define a file containing the full prompt template.
  -s, --system <txt file>     Specify a file containing the system message.
  -a, --adapter <gguf file>   Apply (Q)LoRA adapters to the model.
  -j, --projector <gguf file> Define multimodal projectors.
  -l, --license <txt file>    Specify a file containing the legal license.

  -g, --show                  Print the GGUF metadata of the model.
  -h, --help                  Display this help and exit.

Example

Download a model from the Ollama Registry

ollama-create gemma2

This will download the Gemma 2 model from the Ollama Registry and configure it in Ollama (same as ollama pull gemma2).

Use existing GGUF model file

ollama-create --from llama3.1 my-model.gguf

This will use the local my-model.gguf file and configure it in Ollama by creating a symlink. The Ollama metadata (template, params etc.) is taken from the Llama 3.1 model.

If my-model.gguf doesn't exist, the llama3.1 model will be downloaded and saved as my-model.gguf.

Use custom template

ollama-create --template my-template.txt gemma2

This will download the Gemma 2 model but use the local my-template.txt file as prompt template.

If my-template.txt doesn't exist, the template from the Gemma 2 model will be downloaded and saved as my-template.txt.

Remove parameters

ollama-create --no-params gemma2

This will download the Gemma 2 model but removes all parameter instructions.

Save all model files in custom location

ollama-create --dir ./models/gemma gemma2

This will download the Gemma 2 model and save all Ollama artifacts to the ./models/gemma/ directory. Symlinks in the Ollama blob store are created for all files.

Print GGUF metadata

ollama-create --metadata my-model.gguf

Prints the GGUF metadata of the model file as JSON.