feat(MM):model settings export import

[Pfannkuchensack]

Summary

Add export/import functionality for model settings (default settings, trigger phrases, cpu_only). This allows users to save a model's configuration to a JSON file and re-apply it later or to another model of the same type.

Export saves the current model's default_settings, trigger_phrases, and cpu_only fields to a downloadable .json file.

Import reads a .json file and applies the settings via the existing updateModel API mutation.

Related Issues / Discussions

QA Instructions

1. Open the Model Manager and select a model
2. Configure some default settings (e.g. steps, CFG scale, dimensions) and save
3. Click the export button (download icon) — a .json file should download
4. Change or clear the settings and save
5. Click the import button (upload icon) and select the exported .json file
6. Verify the settings are restored immediately in the form (no page reload needed)
7. Verify the settings persist after a page reload
8. Test with different model types: Main models, LoRAs (trigger phrases + weight), Encoder models (cpu_only), ControlNet/T2I Adapter (preprocessor)

Merge Plan

No special merge considerations. Standard merge to main.

Checklist

• The PR has a short but descriptive title, suitable for a changelog
• Tests added / updated (if applicable)
• ❗Changes to a redux slice have a corresponding migration
• Documentation added / updated (if applicable)
• Updated What's New copy (if doing a release after this PR)

[Copilot]

Pull request overview

This PR adds export/import functionality for model settings in the Model Manager, allowing users to save and restore configurations including default settings, trigger phrases, and CPU-only flags. The changes enable users to backup model configurations and apply them to other models of the same type.

Changes:

• Added ModelSettingsExportButton component to download model settings as JSON
• Added ModelSettingsImportButton component to upload and apply settings from JSON files
• Integrated export/import buttons into ModelView header for models with settings
• Added useEffect hooks to auto-reset forms when model data changes (enabling immediate UI updates after import)

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
ModelView.tsx	Integrated export/import buttons in model header for models with settings
ModelSettingsImportButton.tsx	New component handling JSON file upload and settings application via API
ModelSettingsExportButton.tsx	New component creating and downloading JSON files with model settings
MainModelDefaultSettings.tsx	Added useEffect to reset form when settings change externally
LoRAModelDefaultSettings.tsx	Added useEffect to reset form when settings change externally
EncoderModelSettings.tsx	Added useEffect to reset form when settings change externally
ControlAdapterModelDefaultSettings.tsx	Added useEffect to reset form when settings change externally
en.json	Added localization strings for export/import functionality

Comments suppressed due to low confidence (6)

invokeai/frontend/web/src/features/modelManagerV2/subpanels/ModelPanel/ModelSettingsExportButton.tsx:66

• Consider disabling the export button when there are no settings to export (i.e., when buildExportData returns an empty object). This would provide better UX by preventing users from exporting empty files. You could check Object.keys(data).length === 0 and set isDisabled on the IconButton accordingly.

  return (
    <IconButton
      size="sm"
      icon={<PiDownloadSimpleBold />}
      aria-label={t('modelManager.exportSettings')}
      tooltip={t('modelManager.exportSettings')}
      onClick={handleExport}
    />

invokeai/frontend/web/src/features/modelManagerV2/subpanels/ModelPanel/ModelSettingsExportButton.tsx:21

• The condition modelConfig.trigger_phrases will exclude exporting when trigger_phrases is an empty array, which is a valid state (no trigger phrases). This means if a user clears all trigger phrases, an exported file won't include the trigger_phrases field at all. Consider changing to modelConfig.trigger_phrases !== null && modelConfig.trigger_phrases !== undefined to export empty arrays.

  if ('trigger_phrases' in modelConfig && modelConfig.trigger_phrases) {
    data.trigger_phrases = modelConfig.trigger_phrases;
  }

invokeai/frontend/web/src/features/modelManagerV2/subpanels/ModelPanel/ModelSettingsExportButton.tsx:17

• The condition modelConfig.default_settings will exclude exporting when default_settings is an empty object {}, which might be a valid state. Consider changing to modelConfig.default_settings !== null && modelConfig.default_settings !== undefined to be more explicit about what should be excluded from export.

  if ('default_settings' in modelConfig && modelConfig.default_settings) {
    data.default_settings = modelConfig.default_settings;
  }

invokeai/frontend/web/src/features/modelManagerV2/subpanels/ModelPanel/ModelSettingsImportButton.tsx:80

• The error handler catches the error but doesn't add any additional error logging. Consider adding console.error or using a logger (similar to how ModelDeleteButton uses logger('models')) to aid in debugging when imports fail. This would help developers and users diagnose issues with malformed JSON files.

        } catch {
          toast({
            id: 'SETTINGS_IMPORT_INVALID',
            title: t('modelManager.settingsImportInvalidFile'),
            status: 'error',
          });
        }

invokeai/frontend/web/src/features/modelManagerV2/subpanels/ModelPanel/ModelSettingsImportButton.tsx:82

• The FileReader lacks error handling. If the file reading fails (e.g., due to permissions issues or corrupted file), no error is reported to the user. Consider adding reader.onerror and reader.onabort handlers similar to the pattern used in other parts of the codebase (see GeneratorTextareaWithFileUpload.tsx or useLoadWorkflowFromFile.tsx).

      const reader = new FileReader();
      reader.onload = (event) => {
        try {
          const json = JSON.parse(event.target?.result as string);
          applySettings(json);
        } catch {
          toast({
            id: 'SETTINGS_IMPORT_INVALID',
            title: t('modelManager.settingsImportInvalidFile'),
            status: 'error',
          });
        }
      };
      reader.readAsText(file);

invokeai/frontend/web/src/features/modelManagerV2/subpanels/ModelPanel/ModelSettingsImportButton.tsx:37

• The import function lacks validation of the imported data structure. While the API likely validates the data server-side, there's no client-side validation to check if trigger_phrases contains only strings, or if default_settings has the expected shape. Consider adding basic type validation to provide better error messages to users before attempting the API call.

  const applySettings = useCallback(
    async (data: Record<string, unknown>) => {
      const body: Record<string, unknown> = {};

      if ('default_settings' in data && data.default_settings !== null) {
        body.default_settings = data.default_settings;
      }

      if ('trigger_phrases' in data && Array.isArray(data.trigger_phrases)) {
        body.trigger_phrases = data.trigger_phrases;
      }

      if ('cpu_only' in data && data.cpu_only !== null) {
        body.cpu_only = data.cpu_only;
      }

      if (Object.keys(body).length === 0) {
        return;
      }

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[lstein]

I tested with multiple model types and it works well. However, I ran into an edge condition that I think should be addressed. If you save the config from one type of model, such as Main, and then load the .json into another type of model, for example LoRA, you get a big red cryptic error containing the pydantic validation message. It would be better to catch this exception and give the user a shorter and more informative message.