Wizard save and edit integration config flow
epic-external-system-integration-configuration-admin-ui-task-014 — Implement wizard completion saving: on final confirmation, POST or PATCH the complete integration configuration to the Integration Config Service, handle success by navigating to the dashboard with a success toast, and handle API errors inline on the confirmation screen. Support re-opening the wizard in edit mode by pre-populating all steps from existing config.
Acceptance Criteria
Technical Requirements
Execution Context
Tier 5 - 253 tasks
Can start after Tier 4 completes
Handles integration between different epics or system components. Requires coordination across multiple development streams.
Implementation Notes
Store the full wizard state in a single WizardStateCubit that is provided at the wizard route level and disposed on completion or cancellation. Use a discriminated union (sealed class) for wizard mode: CreateMode() and EditMode(integrationId). In EditMode, fetch existing config on wizard init and map each config field to the corresponding step's form state — use a WizardConfigMapper class to keep this mapping testable and separate from UI. For credential fields in edit mode, send a sentinel value (e.g., '__UNCHANGED__') from the client; the backend interprets this as 'keep existing value' to avoid re-transmitting secrets.
Validate all steps client-side before dispatching the save call to reduce unnecessary round-trips. On 422, parse the error response's field path to identify which step owns the errored field and surface the error on that step using the BLoC error state.
Testing Requirements
Write unit tests for WizardSaveCubit covering: POST success → navigates to dashboard, PATCH success → navigates to dashboard, 422 validation error → field errors mapped to correct steps, 5xx error → inline banner shown, double-tap prevention via button disablement. Write widget tests for the confirmation step verifying correct CTA label ('Create' vs 'Update') and loading state appearance. Write integration tests for round-trip create → fetch → edit → update flow using staging Supabase. Test credential masking: verify raw credential values are not present in widget tree during edit mode.
Test concurrent save prevention: rapid double-tap must result in exactly one API call.
The multi-step Integration Setup Wizard must render different credential fields, field mapping targets, and validation rules depending on the selected integration type. If the type-specific branching logic is implemented as conditional widget trees rather than driven by the Integration Type Registry, the wizard becomes unmaintainable and adding new integration types requires UI code changes.
Mitigation & Contingency
Mitigation: Design the wizard to be metadata-driven from the Integration Type Registry from day one. Credential form fields, required field validation, and mapping target lists are all fetched from the registry, not hardcoded in widgets. Implement one integration type end-to-end first (Xledger) to validate the pattern before building the others.
Contingency: If the metadata-driven approach proves too complex for the initial delivery, implement Xledger and Dynamics as hardcoded wizard variants and create a registry-driven refactor as a follow-up technical debt ticket with a fixed deadline.
The Excluded Features Configuration Panel must wire directly into the feature flag system to suppress HLF app features. If the feature flag system does not yet expose a writable admin interface, this panel cannot save its configuration, blocking the HLF-specific acceptance criteria.
Mitigation & Contingency
Mitigation: Verify that the Organization-scoped Feature Flags feature (a declared dependency) exposes a Dart API for programmatic flag writes before starting this panel. Coordinate with the feature flags team to ensure the write API is available. If needed, schedule this panel as the last item in the epic.
Contingency: If the feature flag write API is unavailable at implementation time, store excluded features in the integration's JSONB settings column and wire them into a local feature flag provider that merges database state with the standard flag system at app startup.
The Field Mapping Editor's usability for non-technical org admins is high-risk. If the visual mapping interface is confusing, admins will configure incorrect mappings that cause silent data corruption in accounting exports — a serious financial risk discovered only at month-end reconciliation.
Mitigation & Contingency
Mitigation: Conduct usability testing with at least one admin user from Blindeforbundet on the field mapping editor prototype before full implementation. Provide descriptive labels and sample data values for all fields. Add a 'test mapping' preview that shows a transformed sample record before saving.
Contingency: If usability testing reveals the visual editor is too complex, implement a simplified list-based mapping editor (select app field → select external field, one row at a time) as a fallback, deferring the drag-and-drop visual editor to a future iteration.
The Credential Management Form's masked fields and connection-test flow may conflict with screen reader requirements — VoiceOver and JAWS must be able to navigate the form, understand which fields are already configured, and receive feedback on connection test results without exposing credential values in accessible text.
Mitigation & Contingency
Mitigation: Design accessible semantics labels for masked fields (e.g., 'API key: configured, last 4 characters: abcd') from the start. Use Flutter's Semantics widget to provide screen-reader-specific text that differs from visual display. Test with VoiceOver on iOS and TalkBack on Android during development, not only at QA.
Contingency: If accessibility conflicts with security requirements for the credential form, implement a separate 'accessibility mode' flow where credential configuration is done through a separate confirmation step that provides more explicit semantic feedback without risk of value exposure.