AI-Powered Personal Compliment & Coaching Engine

Design and specify the architecture for a single-page web application named "Mirror." This application is an AI-powered personal coaching tool designed to provide emotionally intelligent and highly personalized feedback to the user.

I. Core Product Goal & Persona:
The application must function as a private, reflective journal rather than a typical application. The overall aesthetic should evoke a sense of warmth, privacy, and gentle guidance, aiming to feel like a personal diary rather than a functional utility.

II. Feature Specification:

1. Onboarding & Personalization:

   • Domain Selection: Users must select a primary domain for coaching (e.g., Career, Fitness, Creative Work, Relationships).
   • Validation Style Selection: Users must select a preferred coaching style that dictates the tone of the AI feedback:
     • Tough Love
     • Warm Encouragement
     • Analytical

2. Daily Check-in Module:

   • A simple, short form for daily submission requiring three inputs:
     • What I accomplished/did today.
     • How I felt about those activities.
     • One specific thing I am proud of from the day.

3. AI Feedback Engine (The Core):

   • The system must integrate with the [LLM API] (specifically targeting claude-sonnet-4-20250514).
   • System Prompt Instruction: The prompt sent to the LLM must instruct it to adopt the persona of a perceptive, emotionally intelligent coach. The response must strictly follow this structure:
     1. Acknowledge the user's effort and the content of their check-in.
     2. Name at least one specific strength demonstrated in the submission.
     3. Conclude with one concise, forward-looking, actionable insight.
   • Constraint: The AI must strictly avoid generic, platitude-based praise (e.g., "great job," "well done"). Feedback must be specific and insightful.

4. Data Persistence & Review:

   • Wins Archive: A dedicated section to store all past daily check-ins and the corresponding AI responses. This archive must be fully sortable by date and feature robust search functionality.
   • Streak Tracker: A simple, non-gamified counter displaying the number of consecutive daily check-ins completed.

III. User Interface (UI) & Aesthetics:

• Color Palette: Dominated by a warm, soft background color, specifically cream (#F5F0E8).
• Typography: Use clean, elegant serif typography to enhance the journal-like, reflective feeling.
• Design Philosophy: The layout must prioritize comfort, readability, and a sense of privacy. Avoid typical app visual clutter.

IV. Technical Stack & Architecture:

• Frontend: React.js.
• Data Persistence: Utilize localStorage exclusively for all user data persistence (check-ins, settings, archive).
• Backend: No external backend server is required; the application must be a fully functional Single-Page Application (SPA).
• API Integration: Direct interaction with the specified LLM API for all coaching responses.
• Notifications: Implement only one feature: a gentle, user-settable daily reminder notification.

Deliverable: Provide the detailed structure, component breakdown, and the exact system prompt required for the AI engine, alongside the React component structure necessary to achieve this vision.