micode still rejects both per-agent overrides.
It prints:
model openai/gpt-5.4 for brainstormer / commander is not available
then falls back to openai/gpt-5.2-codex
same behavior with github-copilot/gpt-5.4
So the problem is now clearly:
micode’s per-agent model validator is stricter than OpenCode’s runtime model registry.
Current state
Global OpenCode default: openai/gpt-5.4
micode per-agent overrides: set to GPT-5.4 variants, but currently ignored
actual micode fallback: openai/gpt-5.2-codex
micode still rejects both per-agent overrides.
It prints:
model openai/gpt-5.4 for brainstormer / commander is not available
then falls back to openai/gpt-5.2-codex
same behavior with github-copilot/gpt-5.4
So the problem is now clearly:
micode’s per-agent model validator is stricter than OpenCode’s runtime model registry.
Current state
Global OpenCode default: openai/gpt-5.4
micode per-agent overrides: set to GPT-5.4 variants, but currently ignored
actual micode fallback: openai/gpt-5.2-codex