tobiasfabian/nova-ollama

Nova extension to generate and complete code with ollama

This Nova extension integrates ollama to help you complete code or generate code based on a prompt. This extension was developed to use code completion without external services, to ensure privacy and to be independent of online services.

The extension comes with two commands, Assist and Complete code.

nova-ollama-assist.mp4

nova-ollama-complete-code.mp4

Requirements

ollama must be installed and running on your Mac. The model you want to use must be installed with ollama.

You can start ollama with ollama serve (or start the desktop application).
If you don't already have the model you want to use installed, you can do this with ollama pull llama2 (installs llama2).

Usage

The Assist command opens an input request where you can type your prompt. Ollama will generate code based on your input. The generated code will be inserted at the current cursor position.

The Complete code command takes the selected code and tries to continue the code. If you don’t select any code, the code from the beginning of the file to the current cursor position will be used for the input prompt.

To run one of these commands

• Select the Editor → ollama menu item and click on Assist or Complete code
• Open Command Palette … (shift + cmd + p) and type Assist or Complete code

The Assist command is bind to cmd + opt + ctrl + p. To change the key binding, open the Settings, got to Key Bindings and search for Assist.

Configuration

To configure global preferences, open Extensions → Extension Library… and select ollama's Preferences tab.

You can also configure preferences on a per-project basis in Project → Project Settings….

Ollama URL

Default: http://localhost:11434

Set the origin URL where ollama is running. If you are not sure which URL ollama is running on, check the output of ollama serve. (e.g. Listening on 127.0.0.1:11434 (version 0.1.24))

Model name

Default: llama2
Example: openhermes2.5-mistral:7b-q4_K_M

The name of the model you want to use. The model must be installed. To install a new model run ollama pull llama2 ollama CLI Reference

Model names follow a model:tag format, where model can have an optional namespace such as example/model. The tag is optional and, if not provided, will default to latest. The tag is used to identify a specific version. ollama API Docs

Model parameters

Default: {}
Example: { "num_predict": 256, "stop": "\n" }

A JSON object containing model parameters.
Description of the parameters

The example parameters will predict 256 tokens (default 128) and stop the output on a line break \n.

Model keep alive

Default: 5m

Controls how long the model will stay loaded into memory following the request

System message “Complete code”

Default:

You are a code autocompletion engine. Respond with a continuation of the code provided and nothing else. Code should not be in a code block. Anything that is not code should be written as a code comment.

This system message is used when you run the “Complete code” command. You can try different messages to fine-tune the output.

System message “Assist”

Default:

You are an assistant helping with coding. Respond with code only. Don’t format code. Don’t put code in code blocks. Don’t repeat the input.

This system message is used when you run the “Assist” command. You can try different messages to fine-tune the output.

Privacy

This extension needs Send Network Requests permission to send and receive requests to the locally running ollama server. No data will be sent to external servers.