SEO Auditor - Monorepo Architecture

A modern, high-performance Technical SEO Auditor built with Next.js, Express, WebSocket, and Python BFS Crawling Engine organized as a clean monorepo.

---

Workspace Layout

• apps/frontend: Next.js 14 Dashboard UI application.
• apps/backend: Express REST API + WebSocket Server wrapping the seo_auditor Python engine.
• packages/shared: Shared TypeScript types, schemas, DTO validations, and constants.
• docs/: Comprehensive architecture, API, WebSocket, and Environment variable documentation.
• scripts/: Monorepo development and build scripts.

---

Available Scripts

In the root directory, you can run:

npm run dev

Concurrently starts both frontend and backend development servers.

npm run dev:frontend

Starts only the Next.js frontend application.

npm run dev:backend

Starts only the Express + WebSocket backend server.

npm run build

Builds all packages and applications in proper dependency order (packages/shared -> apps/backend -> apps/frontend).

npm run lint

Runs linter checks across all packages.

npm run test

Runs automated unit and integration tests.

---

Documentation

• Folder Structure
• Architecture Overview
• REST API Specification
• WebSocket Events Specification
• Environment Variables Guide

---

Projects & history (Phase 1)

Every finished audit is stored in a central database (SQLite by default, Postgres via DATABASE_URL) under a project per domain, so runs can be listed and compared. See REST API and Environment.

• Import existing audits once: cd apps/backend && python -m app.db.import_sessions (safe to re-run).
• Scheduled audits: set a project's audit_schedule to daily/weekly, then either set SCHEDULER_ENABLED=true or have a cron service call curl -X POST -H "Authorization: Bearer $CRON_SECRET" https://<backend>/api/cron/run.
• Where crawls run: on the always-on backend (the Oracle VM in deploy/oracle). Vercel serverless functions time out long before a crawl finishes, so they must never run audits.
• Postgres in production: docker compose --profile postgres up -d and set DATABASE_URL=postgresql+psycopg://seo:<POSTGRES_PASSWORD>@postgres/seo.

How to verify (about 5 minutes)

cd apps/backend
python -m pytest -q                                   # all tests pass
# two audits of one site
python -m app.cli https://books.toscrape.com/ --max-pages 20 --max-depth 2 --skip-external-links --json | grep -o '"persisted": [a-z]*'
python -m app.cli https://books.toscrape.com/ --max-pages 20 --max-depth 2 --skip-external-links --json | grep -o '"audit_id": [0-9]*'
uvicorn app.main:app --port 5055 &
curl -s localhost:5055/api/projects                   # books.toscrape.com, audit_count 2
curl -s localhost:5055/api/projects/1/audits          # two audits, newest first
curl -s localhost:5055/api/audits/2/diff              # diff against audit 1

---

Rank tracking (Phase 2)

Track keywords per project, daily, with competitor positions and alerts. See REST API and Environment.

1. Add keywords and competitors: POST /api/projects/{id}/keywords, POST /api/projects/{id}/competitors.
2. Preview for free: POST /api/projects/{id}/rank-checks/run returns a dry-run preview. It's made-up data, never stored.
3. Go live when you choose to spend:
   • DataForSEO: SERP_PROVIDER=dataforseo, DATAFORSEO_LOGIN, DATAFORSEO_PASSWORD, then SERP_DRY_RUN=false. SERP_DAILY_CAP (default 50) is a hard daily limit.
   • Google CSE: GOOGLE_CSE_API_KEY + GOOGLE_CSE_CX, capped by GOOGLE_CSE_DAILY_QUOTA. Top 10 only; no device targeting.
4. Alerts: set SLACK_WEBHOOK_URL and/or the SMTP_* + ALERT_EMAIL_* settings.
5. Schedule: SCHEDULER_ENABLED=true, or a cron calling /api/cron/run. Scheduled checks run only with a live provider: dry-run data never becomes history.

How to verify (no money spent)

cd apps/backend
python -m pytest tests/test_rank_schedule.py -v   # daily checks write history; a drop alerts once
uvicorn app.main:app --port 5055 &
curl -s -X POST -H 'Content-Type: application/json' -d '{"domain":"example.com"}' localhost:5055/api/projects
curl -s -X POST -H 'Content-Type: application/json' -d '{"keywords":[{"phrase":"espresso"}]}' localhost:5055/api/projects/1/keywords
curl -s -X POST localhost:5055/api/projects/1/rank-checks/run                 # dry-run preview
curl -s -o /dev/null -w '%{http_code}\n' -X POST "localhost:5055/api/projects/1/rank-checks/run?dry_run=false"   # 409 until a live provider is set

---

Keywords Explorer & Content Gap (Phase 3)

• Explore a keyword: POST /api/keywords/explore {"seed": "..."}. You get ideas (Autocomplete + People Also Ask), volume, difficulty, a traffic estimate, and optional clusters. CSV via GET /api/keywords/explore.csv?seed=....
• Content gap: POST /api/projects/{id}/content-gap returns keywords your competitors rank for (top 20) that you don't (top 100), ranked by opportunity. CSV via GET /api/projects/{id}/content-gap.csv.
• Cost control: dry run until DataForSEO is live (SERP_PROVIDER=dataforseo, credentials, SERP_DRY_RUN=false). DATAFORSEO_LABS_DAILY_CAP (default 20) is a hard daily limit, and cached answers are free.

여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗