The OpenAI backend is provided and uses the davinci-codex engine by default as that is the most useful for code.
| key | default value | description |
|---|---|---|
| api_endpoint | https://api.openai.com/v1/engines/davinci-codex/completions | v1 create completion HTTP POST endpoint, replace davinci-codex with another engine name to use a different engine |
| cache | { dir_name = 'http' } |
requests/responses will be saved under to stdpath('cache') .. '/nvim-magic-openai/' .. cache.dir_name, set to nil to disable this functionality |
The API key must be provided via an OPENAI_API_KEY environment variable.
Be aware of the following:
Since the API key is passed as an environment variable to Neovim, it could be exposed to other processes. It will be visible if the environment of the Neovim process is inspected using something like htop, for example.