Create 7 slide for GuideLight, an AI-powered citizen services navigator. A citizen describes their goal in plain language (e.g., "I need an income certificate for college admission"), and the app guides them through a conversational flow: identify the correct government service → check eligibility → upload documents for AI verification → flag common rejection risks → (stretch goal) auto-fill the application form. Tech stack: Frontend: Next.js 14+ (App Router), TypeScript, Tailwind CSS Backend: Next.js API routes (no separate backend server — keep this simple, one deployable app) LLM: Google Gemini API (model: gemini-2.5-flash or gemini-3-flash — use whichever is confirmed available; agent should verify current model name against Google's docs before hardcoding) No database. No authentication. No user accounts. Every session is ephemeral and identified only by a session token. Feature 1: Service Identification + Staged Eligibility/Checklist Flow Feature 2: Document Readiness Verification Feature 3: Rejection-Pattern Intelligence(Verification Rule: Construct a session state that should trigger at least one rule; confirm the correct rule fires, is phrased clearly, AND the percentile score reflects the triggered rule's severity weight correctly). Feature 4:Helping chatbot for Glossary terms such as Tehsildar, Village Officer, Non-Creamy Layer, the Nativity/Residence/Domicile distinction (these three get confused with each other even on the real portal, worth nailing down), Akshaya Centre, attested vs. notarized vs. self-attested, etc.