Lovable Build — ModelMatch (Model Comparison)

PRD (paste and fill in):
Product: ModelMatch (Model Comparison + Model Swapping App)
User: Small business owners choosing models based on cost, speed, quality, privacy, and context length
Problem: Model selection is confusing; owners waste time + money and get inconsistent outcomes using the wrong model
Promise: Discovery-driven recommendations + instant model swapping with clear tradeoffs
MVP Scope:
- Discovery wizard (business function, task type, constraints)
- Recommendation page (best-fit model + rationale + alternatives)
- Swap/Compare page (side-by-side tradeoff comparison)
- Decision page (copyable recommended setup summary + save decision)
- Saved decisions list (open/review previous decisions)
Discovery Inputs:
- Business function: Marketing, Sales, Support, Ops/Admin, Finance, Hiring, Other
- Task type: Chat/Ideation, Writing, Summarize, Data Extraction, Coding, Image
- Constraints: Budget sensitivity (Low/Med/High), Speed (Low/Med/High), Quality (Low/Med/High), Privacy (Standard/High), Context length (Short/Med/Long)
- Optional: Situation (1–2 sentences)
Requirements:
- Discovery → Recommendation must be clean and fast
- Swap/Compare must be side-by-side and clickable
- Decision summary must be copyable and savable
- Everything works without login (local storage is fine)
Success Metrics:
- % users completing discovery → recommendation
- % users using swap/compare
- % users saving a decision
Out of Scope (today):
- Real-time pricing ingestion, procurement workflows, team workspaces, advanced analytics

Now turn this PRD into a Loveable build prompt. The build prompt must include:
- Screens (Home, Discovery Wizard, Recommendation, Compare/Swap, Decision, Saved Decisions)
- Data model (Model, Decision)
- Logic rules (model scoring + top 3 alternatives + swap selection)
- Acceptance criteria (what "done" means)
- Seed model library (neutral names; no vendor lock-in)

Acceptance / quality check: A user can complete the main loop in under 2 minutes — Start Discovery → submit → see recommended model → click an alternative → compare → choose → generate summary → save → reopen saved decision.