STARLPrep Backend

Backend service for STARLPrep.AI, a mode-based interview coaching copilot that produces structured STARL answers and follow-up practice.

What it does

• Mode detection: Quick Polish vs Guided Build vs Follow-up Simulator
• Generates: Fit check, STARL answer, scoring rubric, follow-up questions, and improvement plan
• Guardrail: avoids fabricated metrics; recommends labeled proxies when numbers are unknown

Tech

• Node.js + Express
• LLM API (configured via env vars)
• Prompt orchestration via promptSpec.js

Endpoints

• GET /health — service status
• POST /chat — accepts { sessionId, messages } and returns { type, reply }

Local setup

1. npm install
2. Create .env with your API key
3. npm start

Roadmap (reliability + product hardening)

• Session persistence (Redis/Supabase)
• Structured outputs (schema-validated)
• Evaluation harness (quality/cost/latency/safety)

Links

• Live demo: https://starlprepai.lovable.app/
• Portfolio case study: https://pastoral-foe-73f.notion.site/Portfolio-Gianfranco-Senaja-6efc229b2a684e4b8355a91a4acf4a9c