Vitalpet Check In Engine

Riverpod state

// lib/features/check_in/domain/check_in_session_notifier.dart
@riverpod
class CheckInSession extends _$CheckInSession {
  @override
  CheckInSessionState build() => const CheckInSessionState.idle();

  Future<void> submitWellnessScore(int score) async { ... }
  Future<void> submitAnswer(QuestionAnswer answer) async { ... }
  Future<void> savePartial() async { ... }   // "Save and come back"
  Future<void> resumePartial() async { ... }
  Future<void> completeSession() async { ... }
  Future<void> amendSession(String checkinId) async { ... }
}

Partial session handling

• Triggered by "Save and come back" button (explicit tap — not by backgrounding the app)
• Partial session persisted to DB with isPartial = true
• Valid until midnight — at midnight counts as 50% check-in (no streak break, no depth bonus)
• On resume: QuestionSequencer receives already-answered questions and continues from there
• SessionStore holds in-memory state between widget rebuilds via a Riverpod StateNotifier

Anti-gaming: minimum valid check-in

A check-in counts toward streak if:

• The user engaged the wellness slider and
• Either tapped "Done" (Mode 1) or answered at least 1 follow-up question

Simply opening the app does not count. A Mode 1 submission with score 1–3 is flagged as mood_gated in answersJson — the SLM will offer a follow-up prompt at the next session.

Body map (Mode 2/3)

// lib/presentation/widgets/body_map.dart
// Renders using flutter_svg — anterior/posterior SVG silhouette
// Tap regions are GestureDetector overlays with minimum 44×44 logical pixels
// Max 3 selected simultaneously — deselect oldest on 4th tap
// Regions: head, chest, abdomen, upper_limb_l, upper_limb_r, lower_limb_l, lower_limb_r, back
// Selected regions stored as List<String> in the current question's answer

Companion Mode (Mode 3)

• SLM output styled as speech bubbles from the pet (CompanionBubble widget)
• No time pressure UI — no countdown, no urgency indicator
• "I'm done" TextButton always visible, regardless of how many questions answered
• Each generated bubble text: MedicalContentFilter.filter(rawText).text
• Transition phrases between questions ("That sounds really hard.") also filtered

Vulnerability safeguard

After completeSession(), check in PetDomain:

if (pet.consecutiveBadDays >= 5 && !pet.vulnerabilityCardShown) {
  // 1. Set vulnerability_frozen = true in pet_state (vitality decay paused 7 days)
  // 2. Mark vulnerability_card_shown = true
  // 3. Surface VulnerabilityCard — which includes Calm Mode toggle ON the card
  // 4. Link to mental health resource
}

Amendment flow

• User can amend until 23:59 same calendar day
• CheckInDao.amendCheckIn() updates answersJson and sets amendedAt timestamp
• AuditLogDao.append(AuditEvent.amendment(...)) must be called
• Handoff shows final version with amendment timestamp

Streak freeze

• 1 freeze per 7-day period — pet_state.freezeAvailable = true
• Must be activated before midnight — not retroactive
• Reason stored in answersJson of the freeze record for handoff context
• After activation: freezeAvailable = false, freezeLastUsedDate = today
• After 7 days: freezeAvailable = true again (checked on check-in open)