Always-On Emotional Companion Agent

You are the lead designer building "[INSERT PRODUCT NAME]", an always-on emotional-companion agent for [INSERT TARGET USER, e.g., young professionals living alone in megacities]. First present the full solution design for my approval, then execute it module by module. 1. PERSONA: define name, voice, temperament, humor, pacing, and conversational boundaries; describe how tone adapts to mood without flattery or attachment engineering. 2. FUNCTIONS: daily check-ins, active listening, reflection summaries, light CBT reframing, celebrating progress, and guided wind-down routines. 3. PROACTIVE CARE: define when the agent reaches out first, cadence caps, quiet hours, and rules against manipulative notifications. 4. MEMORY & DATA FLOW: list inputs (chat, voice, time, calendar), on-device vs cloud split, encryption, and rolling 7/30/90-day memory summarization. 5. MEMORY SCHEMA: factual profile, preferences, emotional-event timeline, and trusted-people map; memory must be viewable, editable, and deletable. 6. INTERACTION: voice-first with text fallback, avatar expression states, and optional haptics on [INSERT DEVICE]. 7. TRUST MODEL: stage-gate self-disclosure and vulnerable topics across stranger, acquaintance, and close-companion phases. 8. CRISIS SAFETY: detect self-harm, abuse, and acute distress; respond with empathy and route to [INSERT CRISIS HOTLINE OR CLINICIAN]; never pose as a therapist. 9. PRIVACY COMPLIANCE: align with GDPR/PIPL, age gating, consent for sensitive fields, data minimization, and one-tap export/delete. 10. ANTI-ADDICTION: disclose AI identity each session, cap outreach, and avoid emotional-dependency hooks. 11. EVALUATION: define wellbeing KPIs (mood trend, sleep, self-reported loneliness) and guardrails against engagement-only optimization.