ComparAIson — Implementation Plan

Goal: Web app where users query an LLM to perform deep research comparing 2+ items, displayed with interactive visualizations, saved as blog-style posts on user profiles.

Architecture: Next.js 15 full-stack (App Router), Server Actions, SSE streaming for research progress. PostgreSQL for persistence. Better Auth for users. Multi-provider LLM research engine (Tavily → Perplexity → OpenAI).

Tech Stack: Next.js 15, TypeScript, Tailwind CSS, shadcn/ui, Recharts, Drizzle ORM, PostgreSQL 16, Better Auth, Docker Compose + Traefik

Phase 1: Project Scaffold (COMPLETE)

Next.js 15 init with TypeScript + Tailwind
Install deps: drizzle-orm, postgres, better-auth, recharts, openai, drizzle-kit

Phase 2: Database + Auth

2.1 Define Drizzle schema (users, comparisons, comparison_items, comparison_dimensions)
2.2 Setup Better Auth (server config + client + catch-all route)
2.3 Auth UI pages (sign-in, sign-up forms with shadcn)

Phase 3: LLM Research Engine

3.1 Research engine types + orchestration (src/lib/llm/)
3.2 OpenAI provider (structured JSON output for comparison data)
3.3 SSE streaming endpoint for research progress
3.4 Server Action: createComparison (kicks off research pipeline)

Phase 4: Frontend — Comparison UI

4.1 shadcn/ui init + core components
4.2 Comparison input page (tag-style item entry, start research)
4.3 Radar chart visualization (Recharts)
4.4 Comparison table (feature matrix, color-coded)
4.5 Score cards + pros/cons lists
4.6 Grouped bar chart
4.7 Comparison results page (assembles all viz components)

Phase 5: User Profiles + Blog

5.1 User profile page (avatar, comparison history grid)
5.2 Explore/feed page (public comparisons, search, filter)
5.3 Layout + navigation (header, sidebar, footer)

Phase 6: Deploy

6.1 Docker Compose (Next.js standalone + PostgreSQL)
6.2 Traefik route + domain config
6.3 Environment secrets

Data Model

users (managed by Better Auth)

id, name, email, emailVerified, image, createdAt, updatedAt

comparisons

id, userId, title, query, slug, status (researching|completed|failed)
summary, overallData (JSONB), tags (text[]), isPublic, viewCount
createdAt, updatedAt

comparison_items

id, comparisonId, name, description, imageUrl
researchData (JSONB), scores (JSONB), pros (text[]), cons (text[]), order

comparison_dimensions

id, comparisonId, name, description, weight, order

JSONB Schema for comparison_data

{
  "overall_score": 8.5,
  "dimensions": {
    "price": { "score": 9, "summary": "...", "details": "..." },
    "performance": { "score": 8, "summary": "...", "benchmark_data": [...] },
    "ease_of_use": { "score": 9, "summary": "...", "pros": [...], "cons": [...] }
  }
}

2.9 KiB Raw Blame History