Skip to content

[codex] Persist custom model selection#100

Merged
Cyberforker merged 1 commit into
Yapie0:mainfrom
Star-Star66:codex/persist-custom-model-selection
Jun 15, 2026
Merged

[codex] Persist custom model selection#100
Cyberforker merged 1 commit into
Yapie0:mainfrom
Star-Star66:codex/persist-custom-model-selection

Conversation

@Star-Star66

Copy link
Copy Markdown
Contributor

Summary

  • persist explicit /model selections in ~/.carboncode/config.json
  • apply the saved model consistently in CLI, ACP, desktop, QQ, and dashboard entry points
  • clear an explicit model pin when the user selects a preset, so presets regain ownership of model selection
  • keep startup precedence explicit: --model > saved model > preset default

Why

Selecting a model that is not one of the built-in flash/pro presets only changed the current process. After restarting Carbon Code, the CLI silently returned to the preset model. This was especially confusing for custom or OpenAI-compatible model IDs.

This change preserves the user's explicit model choice across restarts without expanding the scope into full multi-provider configuration. Existing configs remain compatible because the new model field is optional.

Validation

  • npm run verify
  • 338 test files passed
  • 3918 tests passed, 2 skipped

@Cyberforker Cyberforker merged commit 11db7e4 into Yapie0:main Jun 15, 2026
4 checks passed
@Cyberforker

Copy link
Copy Markdown
Collaborator

Review Summary

Reviewed & merged by Carbon Code

Tests: npm run verify - 3918/3918 passed, 2 skipped
Feat: persist explicit custom model selections across CLI, ACP, desktop, and dashboard entry points.

Thanks @Star-Star66 for the thorough model-persistence implementation and coverage across entry points.

Co-Authored-By: Carbon Code carboncode@code.ai6666.com

-- Carbon Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants