Metier

Find your calling. Screen out the noise. Metier is the recruiter legitimacy check for career professionals. Paste any recruiter email, LinkedIn DM, or outreach message and get a clear read on whether it's worth your time — plus a tracker for every job you're pursuing.

Pronounced "met-yay" — from métier*, your calling, your craft, your vocation.*

⚡ Runs free, no API key required. Metier ships with a deterministic rules engine and an updatable scam-technique database that vet every message locally — no AI bill, fully self-hostable (even on a Raspberry Pi). Add your own API key to layer on an AI-written deep read; without one, the rules engine is the verdict.

• 🎯 Clear verdict — Legit, Dubious, or Scam — with the reasoning behind it
• ⚙️ Deterministic technical forensics — domain age, DNS/MX/SPF/DMARC, link & URL checks, lookalike domains, disposable inboxes, ATS-slug and email-header (spoof) analysis
• 🧠 Updatable scam-technique database — fee/PII/urgency/off-channel/ghost-job patterns in a simple JSON file; add a new technique, no redeploy
• 🔍 Identity & impersonation checks — corporate-domain mismatch, free-inbox executives, template/placeholder fakes
• 🔗 Known-contact correlation — flags (and hard-floors) senders you previously marked as scams
• 🌐 Optional AI layer + live web research — when an API key is present (Claude / OpenAI / Grok)
• 📋 Job tracker — every vetted contact and self-found role in one pipeline, with status tracking

---

See it in action

Screenshots use sample data.

🔍 The Analyzer — paste a message, get a verdict

Paste a recruiter email or DM and Métier returns a clear verdict (Legit / Dubious / Scam) with the red flags it found, what to do, tailored questions to send back, company intelligence, and live web-research findings.

📋 Job Tracker — your whole pipeline in one place

Every vetted contact and self-found role, with a legitimacy verdict, status, comp, and source at a glance.

🚩 Common Red Flags  ·  ✏️ Questions to Ask

Know the patterns Métier checks for, and the questions any real recruiter answers in seconds.

  

---

Quick start

git clone https://github.com/your-username/metier.git
cd metier
pip install -r requirements.txt
python app.py

That's it — open http://localhost:5000 and start vetting. No API key needed; the rules engine runs immediately and for free.

Optional: add an AI layer

Want the AI-written deep read and live web research on top of the rules engine? Add a key:

cp .env.example .env
# Open .env and add your key (any one provider)

When a key is present, the LLM layers on top of the deterministic baseline. When it isn't — or if the call fails — the rules engine is the verdict. You never need to pay for inference.

---

How the verdict works

Every analysis runs the deterministic layer first (no API, no cost):

1. Technical forensics — domain age (RDAP), DNS/MX/SPF/DMARC, link & URL intelligence, lookalike-domain detection, disposable-inbox checks, ATS-slug verification, and raw email-header (spoof) analysis.
2. Scam-technique matching — the message is scored against the pattern database (see below).
3. Known-contact correlation — a strong match to a sender you previously flagged as a scam hard-floors the verdict to Scam.

These produce a 0–100 score → Legit / Dubious / Scam with the matched evidence as red flags. If an API key is configured, the LLM then enriches that baseline (it's told the rule score and may add nuance, but can't quietly score it safer without reason).

Updating the scam-technique database

New scam trends are just data. Edit scam_patterns.json — each entry is one technique:

{
  "category": "fee", "match_type": "keyword", "pattern": "onboarding fee",
  "weight": 42, "severity": "critical", "polarity": "risk",
  "title": "Upfront fee requested",
  "explanation": "Legitimate recruiters are paid by the employer; a candidate never pays."
}

match_type is keyword or regex; polarity is risk (raises the score) or trust (lowers it). Add an entry and restart — the file is synced into the database on startup, and live rows you add directly to the DB are preserved.

---

Choosing a model provider (optional AI layer)

If you add a key, Métier uses Claude by default, but supports OpenAI and Grok (xAI) too. Pick one in .env:

# Claude (default)
LLM_PROVIDER=anthropic
ANTHROPIC_API_KEY=...

# OpenAI
LLM_PROVIDER=openai
OPENAI_API_KEY=...

# Grok / xAI
LLM_PROVIDER=grok
XAI_API_KEY=...

Only the key for your chosen provider is required. OpenAI and Grok need the openai package (already in requirements.txt). The prompts are tuned for Claude, so other models may score a little differently. The footer under each result shows which provider produced the read.

Open http://localhost:5000

---

Running as a service (Linux / systemd)

To keep Metier running in the background on any systemd-based Linux host, a unit file is included. Edit the paths in metier.service to match where you cloned it, then:

sudo cp metier.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable metier
sudo systemctl start metier

You can also deploy it anywhere that runs Python — a VPS, a home server, or a platform like Railway, Render, or Fly.io.

---

Project structure

metier/
├── app.py                  # Flask server, rules engine, technical checks + optional LLM
├── scam_patterns.json      # Updatable scam-technique database (synced into SQLite on startup)
├── templates/
│   ├── index.html          # Analyzer
│   ├── history.html        # Job Tracker
│   ├── questions.html      # Questions to Ask
│   └── redflags.html       # Common Red Flags
├── assets/                 # Logos & app icons (SVG)
│   ├── metier-logo.svg
│   ├── metier-logo-dark.svg
│   ├── metier-app-icon.svg
│   └── metier-icon-small.svg
├── data/                   # SQLite database (git-ignored)
├── requirements.txt
├── .env.example
└── metier.service          # systemd unit (optional, for Linux hosts)

---

License

MIT