Skip to main content

Ollama

Official website https://ollama.ai

Opensource project to run, create, and share large language models (LLMs).

Connect Ollama Models​

  • Download and Install Ollama from the link: ollama.ai.

Download models via the console​

  • To download and use a model immediately in the terminal, run: ollama run MODEL. For example: ollama run deepseek-r1

    This command downloads the model (if not already available) and starts using it right away.

  • To only download the model for later use in CodeGPT, run: ollama pull MODEL, for example :ollama pull deepseek-r1. This command downloads the model without running it, making it available for integration in CodeGPT.

  • Please wait until the loading reaches 100% completion.

How to use Ollama

  • On the VSCode CodeGPT Extension, change the model on the chat.

  • Pick the Local LLMs and select Ollama as Provider.
  • Use the models available directly.
  • Paste the link of the server where the model is running. For localhost: htttp://localhost:11434.
  • Click outside the options and ask to chat.

Remove or edit Key

To remove the port, click on Change connection settings button.

Autocomplete models availables:​

  • codestral:latest
  • qwen2.5-coder:7b
  • deepseek-coder:base
  • qwen2.5-coder:1.5b
  • codegemma:code
  • codellama:code

API Errors​

If you are getting API errors, check the following link: Ollama Documentation

Ollama Errors​

  • If the Ollama model does not respond in the chat, consider restarting it locally by turning it off and then on again. This action should resolve the issue.
  • If the Ollama is running but not responding, please manually remove 'Ollama_Host' from the environment variables and let it revert to the default setting.