Skip to content

jev-pii-checkerFind PII in text with Jev

Detect personally identifiable information using TypeSafe's Jev model, regex patterns, and word segmentation

jev-pii-checker

WARNING

The scanned text is sent to TypeSafe's Jev API. Detection happens on their servers, so this tool is not a local-only guard: everything you scan leaves your machine. Use it only on data you are allowed to hand to a third party, such as text you were going to send to an LLM anyway, and check TypeSafe's data-retention terms first. For a guard that never sends anything, see sensitive-canary. Details on the Security page.

Why jev-pii-checker?

Regex alone misses context and generates noise. Jev alone costs per API call on every character. jev-pii-checker layered them: gate screens first, then regex finds candidates, then Jev judges only what matters.

ApproachCostAccuracySpan
Regex onlyLow🟡 High noise
Jev on all text💰💰 High✅ High
jev-pii-checker (gate → regex → Jev)💰 Medium✅ High

Getting Started

Set up your API key and scan text in seconds:

bash
export TYPESAFE_API_KEY="your-key"
echo "佐藤健一郎 090-1234-5678" | jev-pii-checker --json

See Getting Started for detailed setup.

Features

  • Three-layer detection: Gate presence screening, regex extraction, Jev judgment of candidates
  • JSON and human output: Machine-readable or compact human-friendly reports
  • Severity filtering: Exit with code 2 when findings exceed a threshold
  • Masking by default: Show ja…n not japan@example.com; disable with --show-values
  • Configurable thresholds: Gate probability, span probability, chunk size, concurrency

Install

bash
bun install -g @coo-quack/jev-pii-checker
jev-pii-checker --version

No server required. Requires Node.js 20+ or Bun.

See Install for source builds and alternatives.

Released under the MIT License.