Benchmark Engineering Consultant Platform

A custom-built replacement for the Base44 prototype — full control over data, auth, and the resume-parsing/keyword-search feature that a no-code builder can't do well.

Stack

• Next.js 14 (React + TypeScript) — frontend + API routes in one app
• PostgreSQL + Prisma — data model for consultants, keywords, tickets, users
• NextAuth (Auth.js) — self-hosted credentials-based auth with roles (ADMIN / RECRUITER / VIEWER, plus CLIENT reserved for later)
• Claude API — resume parsing: extracts name, title, years of experience, and tags formations / rig types / skills / certifications / software as structured keywords
• S3-compatible storage — resume file storage (AWS S3 or Cloudflare R2)
• Tailwind CSS — styling

What's built

• Auth (login, session, role-based route protection) with 5 roles: Admin, Recruiter, Viewer, Consultant (self-service), and Client (reserved for later)
• Public "Apply" page (/apply, no login) — candidates submit their info, pick a discipline (Drilling / Completions / Both), and upload a resume
• Admin application review (/admin/applications) — see submitted applications with AI-extracted keywords, Accept or Reject
• Auto-invite on acceptance — accepting an application creates their Consultant record and emails them a signup link (using the email they applied with) to set their own password
• Consultant self-service portal (/my-tickets) — once logged in, a consultant sees only their own required certifications (based on their discipline) and can add/renew their own tickets, optionally uploading proof documents
• Ticket Types admin page (/admin/ticket-types) — you control the master list of required certifications per discipline (H2S Alive, Fall Protection, etc.) — nothing is hardcoded
• Daily expiry email alerts — a scheduled job checks for tickets expiring within 60 days and emails the consultant (via Resend)
• Bulk CSV import (/admin/import) — import your existing consultants and their tickets at once. Ticket expiry dates auto-calculate from the ticket type's default validity period if left blank, and keyword columns (formations, rig types, skills, certs, software) get matched/created and tagged on each consultant automatically
• Client evaluations (/admin/evaluations) — generate a shareable link per consultant (no login required for the client), collect a 1-5 rating and comments, and see results both on the Evaluations dashboard and on the consultant's own profile
• Brand-matched visual design — navy/gold color palette, hardhat logo mark, and bold rounded headings matching your reference design, applied across the login page, the public Apply page hero, and navigation throughout the app
• Dashboard with live stats
• Consultant advanced search with toggleable keyword filters (Canadian formations, rig types, skills, certs, software) and Any/All match modes
• Consultant detail page: profile, tagged keywords, tickets, resume upload
• Resume upload → text extraction (PDF/DOCX) → Claude structured extraction → auto-tagging with confidence scores
• Admin: create internal users, assign roles
• Activity log data model (logging is wired into key actions)
• Evaluation data model (ready for an admin evaluations UI)

What's left to build (next iteration)

• Activity log viewer page (data is already being recorded)
• Consultant edit form (currently only editable via API — quick to add)
• Client-facing portal (the CLIENT role and route-permission structure already anticipate this)
• "Resend invite" button for an application whose invite email needs to go out again
• Downloadable CSV templates for the bulk import (currently the expected columns are documented on the Import page itself)
• Extending the new navy/gold visual style more deeply into the internal admin pages (currently applied to login, the public apply page, and navigation — the rest of the app already inherits the new color tokens automatically but hasn't been individually polished)

Local setup

1. Install dependencies:

   npm install
   

2. Copy .env.example to .env and fill in:

   • DATABASE_URL — a Postgres connection string. Easiest options: Neon or Supabase, both have free tiers.
   • NEXTAUTH_SECRET — generate with openssl rand -base64 32
   • ANTHROPIC_API_KEY — your Claude API key from console.anthropic.com
   • S3_* — an S3 bucket (AWS) or R2 bucket (Cloudflare, cheaper egress) for resume storage
   • RESEND_API_KEY — sign up free at resend.com, create an API key. Their test domain ([email protected]) works immediately for testing; for production you'll want to verify your own domain in their dashboard so emails come from your own address.
   • CRON_SECRET — generate with openssl rand -base64 32. When deployed on Vercel, this is automatically sent as a Bearer token to your cron endpoint — you don't need to configure anything else, Vercel reads the vercel.json in this project and calls /api/cron/ticket-expiry once a day automatically.

3. Push the schema and seed starter data:

   npx prisma generate
   npx prisma migrate dev --name init
   npx prisma db seed
   

   This creates a first admin login: [email protected] / ChangeMe123! — change this password immediately after first login (via the admin users page, or directly in the database until a "change password" UI is added).

4. Run it:

   npm run dev
   

   Visit http://localhost:3000.

Migrating your existing consultant data

Once you export your current consultants/tickets from Base44 (usually CSV or JSON), the fastest path is a one-off script using the same Prisma client — I can write this once I see the shape of your export file. It'll map your existing fields into the Consultant, Keyword, and Ticket tables and can even run each consultant's existing resume through the same Claude parsing pipeline in bulk.

Deploying

• App: push this repo to GitHub, then import it into Vercel — zero config needed beyond the env vars above.
• Database: Neon or Supabase both work well with Vercel and have generous free tiers to start.
• File storage: AWS S3 or Cloudflare R2 (R2 has no egress fees, worth it if resumes get downloaded often).

Notes on the keyword/search design

Keywords are stored as a typed taxonomy (FORMATION, RIG_TYPE, SKILL, CERTIFICATION, SOFTWARE) rather than free-text tags. This is what powers the toggle-style filter UI on the Consultants page — each keyword becomes a clickable chip, and you can require any selected keyword to match or all of them. New keywords are created automatically the first time Claude extracts them from a resume (or you can add them manually from the search page/admin), so the taxonomy grows on its own as you upload more resumes without needing a fixed master list up front.