Decidr

The Problem

Hard decisions rarely fail because of a lack of information — they fail because that information never gets structured into a clear comparison. Spreadsheets and chat threads bury the trade-offs; Decidr is a personal SaaS exploring whether an AI-native tool can make the structuring step feel obvious.

What I Built

Decidr turns a messy dilemma into a weighted option × criteria matrix with explainable scoring — not another chat wrapper.

• Paste a dilemma or walk a multi-step wizard (problem → options → criteria → weights).
• AI extracts options and criteria into structured JSON via OpenRouter (model-agnostic).
• Server-side ranking with risk, confidence, and a maturity score; results show ranked tables, charts, heatmaps, and an expandable reasoning trace.
• What-if weight simulation recomputes locally without another AI call; side-by-side decision compare; public share links and PDF via print.
• Email auth, Supabase data, and Stripe Pro billing on a Next.js App Router stack.

Live demo: decidr-henna.vercel.app.

Major Features

• Decision wizard — plain-text or step-by-step intake that lands in a scored matrix.
• Explainable AI scoring — ranked options with reasoning you can expand, not a black-box “pick this.”
• What-if simulation — drag weights and see rankings update without re-calling the model.
• Compare & share — side-by-side decisions and public /d/[token] links.
• Billing-ready — Stripe Pro wired for the SaaS path.

Key Challenges

Using OpenRouter instead of a single model provider means prompts and output parsing have to stay model-agnostic — structured JSON extraction has to work across vendors, not tuned to one model’s quirks. Keeping the matrix as the product surface (not a chat log) forced clear UX boundaries between AI assistance and user-owned weights.

Results

Decidr is in active development as a personal SaaS: core decision flow, charts, share links, and Stripe are in place on a public Vercel demo.

Stack Decisions

OpenRouter avoids locking decision intelligence to one vendor’s pricing and availability. Supabase and Stripe cover auth, data, and billing without a custom backend, so product time stayed on the matrix UX and scoring pipeline rather than infrastructure.