Repository architecture review and documentation

[Copilot]

Comprehensive codebase review documenting architecture, tech stack, and implementation patterns for the Sift active recall learning platform.

Summary

• Architecture: Turborepo monorepo with Next.js 15 + React 19 frontend, Drizzle ORM + Neon Postgres backend
• AI Integration: Vercel AI SDK with Gemini for MCQ/flashcard generation from user content
• Auth: Better-Auth with Google OAuth + email/password, session-based authorization on all actions

Key Findings

Database Schema

• sources → sifts → questions/sift_sections → session_answers
• learning_paths for multi-module curricula with AI context continuity
• echoes for topic mastery tracking (0-100%)

Core Flows

• Content Processing: Upload/paste → parse → AI generation (3 retries w/ backoff) → questions + flashcards + takeaways
• Quiz Mode: Full-screen zen mode, keyboard shortcuts (A-D, Enter), SSE for real-time status
• Mastery: Batch echo updates on session completion, per-topic retention tracking

Packages

Package	Responsibility
@sift/db	Schema definitions, query functions
@sift/auth	Auth-wrapped actions, session validation
@sift/env	t3-env validation for server/client envs

Notable Patterns

• Server Actions for all mutations with header-based auth validation
• after() from next/server for background AI processing
• Cascade deletes from user → all dependent data

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Review	Updated (UTC)
sift-web	Ready	Preview, Comment	Jan 28, 2026 8:28am