Local models and editor features
Everything here is the same engine the Contenox IDE runs; the IDE drives it for you, and the commands below are how you do the same from a terminal or an editor of your own.
Native models
Section titled “Native models”Native inference runs in modeld, the worker that ships with Contenox and is downloaded on first use when you installed the CLI alone.
contenox-cli pull <name> # download a registry modelcontenox-cli list # installed native models (alias: ls)contenox-cli ps # the model the worker holds in memorycontenox-cli stop <model> # unload it; the worker keeps runningcontenox-cli auto # pick models that fit this machine and save them as defaultsAutocomplete
Section titled “Autocomplete”contenox-cli autocomplete --stdio serves fill-in-the-middle completions
over JSON lines on stdin and stdout, for an editor plugin to call. It uses
the autocomplete role:
contenox-cli config set inference.autocomplete.model <model>contenox-cli config set inference.autocomplete.provider <provider>With no autocomplete model configured, every request is answered with an error that names this fix.
Workspace search
Section titled “Workspace search”Workspace search reads an index of the workspace built with an embedding model, which must be configured explicitly:
contenox-cli config set inference.embedding.model <model>contenox-cli config set inference.embedding.provider <provider>contenox-cli index # build or update the index (incremental)contenox-cli index --status # the active index generationcontenox-cli index --plan # what a build would change, without changing itGitignored paths, binaries, dependency lockfiles and oversized files are left out and reported.
Plain chat
Section titled “Plain chat”contenox-cli chat talks to an agent with no tools: a line in, an answer
out, kept in the active session so it is still there tomorrow. For work on
the machine, use contenox-cli run "<task>" or the IDE.
Updates and catalogs
Section titled “Updates and catalogs”contenox-cli update check # newest published desktop release for this platformcontenox-cli cache clear # forget cached provider model catalogs; the next refresh fetches them again