What do you want to change?
Add first-class support for Command Code as a provider, including dynamic model discovery instead of requiring users to manually maintain models.json entries.
Why?
Command Code GOAT exposes an OpenAI-compatible provider API and model catalog. Senpi can already talk to custom OpenAI-compatible endpoints, but users currently need to copy/model IDs into local config by hand. That gets stale quickly as Command Code adds, removes, or renames models.
This would make Command Code usable like other built-in providers from /model, especially for OMO/Senpi setups that switch models frequently.
How? (optional)
Register a commandcode provider backed by the OpenAI-compatible API, fetch its /models catalog during provider refresh, normalize the returned IDs/capabilities, and keep a static/manual config path only as fallback. Ideally credentials would come from an env var or normal provider auth flow.
What do you want to change?
Add first-class support for Command Code as a provider, including dynamic model discovery instead of requiring users to manually maintain
models.jsonentries.Why?
Command Code GOAT exposes an OpenAI-compatible provider API and model catalog. Senpi can already talk to custom OpenAI-compatible endpoints, but users currently need to copy/model IDs into local config by hand. That gets stale quickly as Command Code adds, removes, or renames models.
This would make Command Code usable like other built-in providers from
/model, especially for OMO/Senpi setups that switch models frequently.How? (optional)
Register a
commandcodeprovider backed by the OpenAI-compatible API, fetch its/modelscatalog during provider refresh, normalize the returned IDs/capabilities, and keep a static/manual config path only as fallback. Ideally credentials would come from an env var or normal provider auth flow.