[FEATURE]: Expose configured model eligibility to model transforms #54480

Activity

1. opencode-agent commented on Oct 11, 2026

   

   opencode-agentBot

   on Oct 11, 2026 – with OpenCode Agent

   Contributor

   More actions

   Thanks for the detailed write-up, @vmvarela!

   To restate the request: a plugin that creates model aliases with ctx.model.transform needs a supported way to tell whether a target model has been disabled in config, so that an alias like openai/gpt-sol can skip a disabled openai/gpt-6.1-sol and use the next candidate. Config precedence should still hold.

   I checked the current code and it matches what you found. The built-in config transform is registered after external model transforms on purpose, so that config wins. That same step is where disabled becomes enabled = false, so external transforms never see it. This is a gap in the plugin API, not a bug in your plugin. A few ways it could be closed:

   • a read-only view of configured overrides or eligibility on the model editor during transforms
   • a later transform or validation phase that runs after config overrides are applied
   • a built-in alias concept that picks its target only after eligibility is applied

   Maintainers will consider this. Related issues:

   • docs: add opencode-model-aliases to the V2 ecosystem #54470: docs entry for opencode-model-aliases
   • plugins: config-defined agents are missing during agent transforms #51598: the same ordering problem for config-defined agents during agent transforms
   • [FEATURE]: Add a native V2 replacement for provider model filtering #49986: native model filtering in v2
   • provider.models hook never runs for config-declared providers (ordering) #44081: hook ordering problem for config-declared providers (v1)