/coc-codellama

A completion engine for nvim.coc using the codellama LLM

Primary LanguageTypeScriptMIT LicenseMIT

coc-codellama

A completion engine for nvim.coc using the codellama LLM.

Install

:CocInstall coc-codellama

Options

Option Type Default value Description
coc-codellama.enabled boolean true Enable coc-codellama extension
coc-codellama.host string localhost The host to connect to.
coc-codellama.port number 11434 The remote port to connect to.
coc-codellama.contextAbove number 7 How many lines above the current line to include in the context for completion.
coc-codellama.contextBelow number 7 How many lines below the current line to include in the context for completion.
coc-codellama.shortcut string CL The shortcut to use for this completion source.
coc-codellama.priority number 995 How far up in the list to put completions from this source.
coc-codellama.model string codellama:code The large language model to use for completions.
coc-codellama.prefix string <PRE> The token to put in front of the fill in the middle prefix
coc-codellama.suffix string <SUF> The token to put in front of the fill in the middle suffix.
coc-codellama.middle string <MID> The token to put at the end of the fill in the middle prompt.
coc-codellama.token-limit number 24 The maximum number of tokens allowed.
coc-codellama.temperature number 0.8 How variable the output should be. Must be between 0 and 2. 0 means totally deterministic.

Using codegemma:code

:CocConfig

"coc-codellama.model": "codegemma:code",
"coc-codellama.prefix": "<|fim_prefix|>",
"coc-codellama.suffix": "<|fim_suffix|>",
"coc-codellama.middle": "<|fim_middle|>",

Demo

asciicast

License

MIT