Rezumerai

Rezumerai is an AI-assisted resume builder for job seekers who want one workspace to draft, refine, tailor, and export resumes. It combines structured resume editing with AI-powered section optimization, job-description matching, and a scope-aware assistant that can answer public, account, or admin questions based on the current user's access.

• Monorepo: Turborepo + Bun workspaces
• App: Next.js 16 App Router + React 19 + embedded Elysia API
• Data: PostgreSQL + Prisma 7
• Auth: Better Auth
• AI: Vercel AI SDK + OpenRouter + pgvector-backed assistant memory

Features

• Resume dashboard to create, search, rename, delete, and download saved resumes
• Guided resume builder for personal info, summary, experience, education, projects, and skills
• Four resume templates: classic, modern, minimal, and minimal_image
• Accent color and font-size controls with live HTML/PDF preview
• Rich-text editing for summary, experience, and project content
• Drag-and-drop reordering for experience, education, projects, and skills
• PDF export and full-page preview
• AI Resume Copilot for section optimization, job-description tailoring, and full resume review
• Streamed AI text optimization with daily credit tracking
• Built-in AI assistant with anonymous guest support, persisted thread history, and PUBLIC / USER / ADMIN scopes
• Better Auth accounts with email/password, GitHub OAuth, anonymous sessions, and admin roles
• Account settings for profile updates, password changes, AI model selection, and AI credit visibility
• Admin console for users, live AI model availability, system configuration, audit logs, analytics, and error logs
• Admin job queue dashboard for monitoring pg-boss queue metrics, dead-letter jobs, and worker configuration
• Database-backed public content for the landing page, FAQ, about, contact, privacy, and terms pages

Tech Stack

Frontend

• Next.js 16 App Router
• React 19
• TanStack Query for server state
• Zustand for UI/store state
• Tailwind CSS v4
• Tiptap rich-text editing
• react-pdf, jspdf, and html2canvas-pro for resume preview/export flows

Backend

• Elysia embedded inside apps/web
• Next.js route bridge at apps/web/src/app/api/[[...slugs]]/route.ts
• Eden Treaty client for end-to-end typed API calls
• Zod and Elysia models for request/response validation

Database

• PostgreSQL
• Prisma 7 with a modular schema in packages/database/prisma/models
• pgvector-backed assistant embedding storage

Authentication

• Better Auth with Prisma adapter
• Email/password login
• GitHub OAuth
• Anonymous guest sessions for the assistant
• Admin role management via Better Auth's admin plugin

AI Infrastructure

• Vercel AI SDK Core in the embedded Elysia API
• AI SDK UI hooks (useChat, useCompletion) on the frontend
• OpenRouter via @openrouter/ai-sdk-provider
• OpenRouter model discovery plus database-backed system configuration
• LangChain text splitters plus pgvector for assistant memory embeddings

AI Prompt Configuration

• AI_CONFIG stores runtime AI prompts, models, and limits in the SystemConfiguration table.
• ASSISTANT_SYSTEM_PROMPT is reserved for chat assistant behavior and remains separate from optimization workflows.
• Resume Copilot uses dedicated prompt keys for optimize, tailor, and review.
• /text-optimizer Text Optimizer uses TEXT_OPTIMIZER_SYSTEM_PROMPT, which is general-purpose text optimization only.
• Legacy COPILOT_SYSTEM_PROMPT and OPTIMIZE_SYSTEM_PROMPT values are normalized into the new keys at runtime and when saved through the admin System Configuration editor.

Observability And Operations

• OpenTelemetry plugin for Elysia request traces
• Pino structured JSON logging with configurable level (LOG_LEVEL) and pretty-print mode (LOG_PRETTY)
• Database-backed analytics events, audit logs, and error logs
• Optional BotID protection for mutating /api/* requests (auto-gated on VERCEL=1)

Monorepo Tooling

• Bun workspaces
• Turborepo task orchestration
• Biome for formatting and linting
• tsdown for package bundling
• Changesets for package versioning and publishing workflows

Prerequisites

• Bun 1.3.x
• PostgreSQL with pgvector support available for migrations
• GitHub OAuth credentials for local auth flows
• Docker (optional, for the current web-only container workflow)

Quick Start

1. Install dependencies

bun install

2. Configure environment variables

cp .env.example .env.local

Core variables for local development:

Variable	Required	Purpose
DATABASE_URL	Yes	Runtime PostgreSQL connection for the app, Prisma client, and assistant memory storage
DIRECT_URL	Yes	Direct PostgreSQL connection for Prisma CLI commands such as db push, migrate, and generate
NEXT_PUBLIC_SITE_URL	Yes	Public app URL used by metadata, OpenRouter headers, and the Eden API client
BETTER_AUTH_URL	Yes	Better Auth server base URL
BETTER_AUTH_SECRET	Yes	Better Auth signing and encryption secret
BETTER_AUTH_GITHUB_CLIENT_ID	Yes	GitHub OAuth client ID
BETTER_AUTH_GITHUB_CLIENT_SECRET	Yes	GitHub OAuth client secret
OPENROUTER_API_KEY	Yes	OpenRouter API key for Copilot, assistant, and optimize-text flows

Optional variables used by the current codebase or reserved in the env template:

Variable	Purpose
BETTER_AUTH_GOOGLE_CLIENT_ID	Optional placeholder in the env schema and template; the current auth config exposes GitHub only
BETTER_AUTH_GOOGLE_CLIENT_SECRET	Optional placeholder in the env schema and template; the current auth config exposes GitHub only
CRON_SECRET	Reserved in .env.example and Turbo env inputs for scheduled/background-job protection
DB_SEED_USER_EMAIL	Extra email to include when running bun run db:seed
LOG_LEVEL	Structured logging level (debug, info, warn, error, fatal). Defaults to "debug" in development, "info" in production
LOG_PRETTY	Enable pretty-printed logs in non-production environments. Set to "true" to force pretty printing
OTEL_EXPORTER_OTLP_ENDPOINT	OTLP collector endpoint for OpenTelemetry trace export
OTEL_EXPORTER_OTLP_HEADERS	Comma-separated auth headers for the OTLP exporter
SENTRY_DSN	Optional env validated in apps/web/src/env.ts; no Sentry bootstrap is wired in the current app
ANALYTICS_ID	Optional env validated in apps/web/src/env.ts; no external analytics provider is wired in the current app
AI_MEMORY_REINDEX_LIMIT	Optional limit consumed by bun run assistant:reindex-memory
JOB_QUEUE_ENABLED	Enable the pg-boss background job queue. Defaults to "true"
JOB_QUEUE_WORKER_COUNT	Number of worker processes. Defaults to "1"
CORS_ALLOWED_ORIGINS	Extra comma-separated origins allowed to call the embedded API

3. Initialize database

bun run db:setup

Optional local reset/demo data:

bun run db:seed:system
bun run db:seed

4. Start development

bun run dev

Notes:

• Root predev builds @rezumerai/database before turbo dev.
• apps/web runs scripts/download-pdf-worker.ts on predev and prebuild.
• Deploy-oriented bun run db:migrate now bootstraps any missing required system configuration and public content rows after Prisma migrations.
• bun run db:seed:system is a development reset tool that rewrites AI/system configuration and public content back to the current defaults.
• bun run db:seed seeds dummy resumes for existing users; it does not create auth users.

Development Workflow

Run only the web app:

bun run --filter=web dev

Common quality checks:

bun run check
bun run check:types
bun run test
bun run build

Single pre-PR command:

bun run code:verify

Useful targeted commands:

bun run db:studio
bun run assistant:reindex-memory

Notes:

• Use Bun only; the repo is configured as a Bun workspace monorepo.
• bun run check syncs nested Biome config versions before running workspace checks.
• Workspace check scripts use Biome in write mode, so they can apply formatting fixes.

Root Scripts

Build and Run

Script	Description
bun run dev	Start monorepo dev tasks with turbo dev
bun run start	Start workspace start tasks
bun run build	Build all apps and packages
bun run build:production	Build all workspaces with NODE_ENV=production
bun run build:packages	Build only workspace packages under packages/*

Code Quality and Tests

Script	Description
bun run check	Sync Biome versions, then run workspace check scripts
bun run biome	Run Biome from the repo root in write mode
bun run update:biome-configs	Update nested Biome config versions to match the root tool version
bun run check:types	Run workspace type checks
bun run code:check	Run check and check:types
bun run test	Run workspace tests
bun run test:watch	Run workspace tests in watch mode
bun run test:coverage	Run workspace tests with coverage
bun run code:verify	Run lint/format, type checks, tests, and build

Database

Script	Description
bun run db:setup	Run db:push and db:generate in packages/database
bun run db:generate	Generate Prisma client and Prismabox artifacts
bun run db:push	Push the Prisma schema to the configured database
bun run db:migrate	Deploy Prisma migrations and bootstrap missing required system configuration rows
bun run db:migrate:dev	Create and apply a development migration
bun run db:migrate:status	Show Prisma migration status
bun run db:bootstrap:system	Generate Prisma client artifacts, then create any missing required system configuration and public content rows without overwriting existing values
bun run db:pull	Pull schema changes from the database
bun run db:seed:system	Reset AI/system configuration and public content rows back to the current defaults
bun run db:seed	Seed dummy resume data for existing users
bun run assistant:reindex-memory	Run the assistant memory reindex script
bun run db:psql	Open psql against DATABASE_URL
bun run db:inspect-vectors	Inspect recent assistant embedding rows
bun run db:studio	Open Prisma Studio on port 5556
bun run db:reset	Force-reset the database with Prisma migrate reset

Release and Packages

Script	Description
bun run changeset	Create a Changesets release entry
bun run version:packages	Build packages, then apply Changesets version bumps
bun run publish:packages	Publish packages with Changesets

Workers

Script	Description
bun run worker	Start the background job worker with default jobs
bun run worker:all	Start worker with all job types: embeddings, analytics, audit log, error-log retention

Docker and Maintenance

Script	Description
bun run docker:build	Build and start Docker Compose services using .env.local
bun run docker:build:standalone	Build the standalone web Docker image from apps/web/Dockerfile
bun run docker:up	Start Docker Compose services
bun run docker:down	Stop Docker Compose services
bun run start:redis	Attempt to start a redis compose service if you add one
bun run ai:superpowers	Refresh the repo-local superpowers skill bundle and Claude symlink
bun run install:modules	Install dependencies and rebuild workspace packages
bun run update:modules:1	Run the repository dependency update helper script
bun run update:modules:2	Upgrade dependencies with Bun's latest update commands
bun run clean	Remove build, cache, coverage, and install artifacts
bun run clean:install	Clean, reinstall dependencies, and rebuild workspace packages
bun run outdated	Show outdated dependencies
bun run security:audit	Run Bun's production dependency audit
bun run security:check	Run audit-ci with the repository config
bun run prepare	Install Husky Git hooks

Architecture Overview

Next.js App Router + API Bridge

• UI routes live under apps/web/src/app.
• apps/web/src/app/api/[[...slugs]]/route.ts forwards /api/* requests into the embedded Elysia app.
• apps/web/src/app/api/auth/[...all]/route.ts is the dedicated Better Auth handler.
• apps/web/src/proxy.ts enforces route protection, admin gating, redirects, and security/CSP headers for non-API requests.

Elysia API Composition

• apps/web/src/elysia-api/app.ts is the API entrypoint and single composition root.
• Security and infrastructure are registered first: OpenTelemetry, trace logging, observability, Helmet, CORS, rate limiting in production, CSRF, Prisma, request logging, and centralized error handling.
• Dev-only Swagger and OpenAPI docs are enabled in development.
• GET /api/health provides a basic health check.
• A daily cron job cleans up expired error logs.

API Modules

• user: authenticated user endpoints
• profile: user profile read and update (name, email, image)
• resume: resume CRUD and search
• ai: assistant chat/messages, model settings, Copilot actions, and streamed text optimization
• admin: users, AI models, system configuration, audit logs, analytics, and error log management
• jobs: pg-boss job queue management and background worker handlers

Module Structure

• index.ts defines routes and transport concerns
• model.ts defines Elysia models and route contracts
• service.ts owns business logic and orchestration
• repository.ts is used where Prisma access is non-trivial
• AI-specific integrations live under controller/, memory/, prompts/, providers/, and tools/

Shared Packages

• @rezumerai/database: Prisma client, schema modules, migrations, and seeds
• @rezumerai/types: shared Zod schemas and TypeScript contracts for AI, resume, user, admin, error, and content domains
• @rezumerai/utils: shared date, string, React, UUID, and style helpers
• @rezumerai/ui: shared UI primitives, auth form components, and skeletons
• @rezumerai/tsconfig: shared TypeScript configuration presets

Conventions

• Use Bun for package management, scripts, and tests.
• Use Biome for formatting and linting; the root config lives in biome.json.
• Prefer shared schemas from @rezumerai/types and validate boundaries with Zod/Elysia models.
• Keep shared logic in @rezumerai/database, @rezumerai/types, @rezumerai/utils, or @rezumerai/ui instead of duplicating app-local code.
• Use ROUTES from apps/web/src/constants/routing.ts instead of hardcoded route strings.
• Keep Elysia feature modules organized around index.ts, model.ts, service.ts, and repository.ts where needed.
• Keep tests colocated in __tests__ when that pattern already exists.
• Use bun test; React/component tests run with React Testing Library and Happy DOM in the packages/apps that declare Bun test preload config.

Docker Notes

• apps/web/Dockerfile defines two build targets: web (Next.js app) and worker (background job processor).
• docker-compose.yml defines both the web and worker services; both run by default with bun run docker:up.
• The PostgreSQL service is present only as commented reference configuration.
• The Docker helper scripts load secrets from .env.local.
• The root Dockerfile is a placeholder; use apps/web/Dockerfile for the actual application image.
• start:redis exists as a helper script, but the default compose file does not define a redis service; the job queue uses pg-boss over PostgreSQL.