redact 0.1.0

On-device, precision-first PII redaction for Dart and Flutter. Scrub emails, phones, cards, SSNs and more from text before it reaches any LLM, then restore the model's reply.

0.1.0

Initial release.

• Redactor — runs detectors, resolves overlaps (longest validated span wins), and rewrites PII with a chosen style; detect() returns spans without rewriting (for UI highlighting and audits); allowList exempts known-safe values.
• Reversible pipeline: RedactionResult.restore re-hydrates placeholder tokens in an LLM's reply — leniently, surviving common model mangling such as [EMAIL\_1] and [email_1] — in a single pass that can never corrupt restored values or pre-existing literal tokens.
• RedactionSession — conversation-scoped vault for multi-turn chats: the same value keeps its token across calls, new values keep counting up, replies referencing any turn restore correctly, and sessions serialize with toJson/fromJson.
• Precision-first built-in detectors: email, phone (E.164 + separated forms; date/ISBN shapes rejected), credit card (Luhn), US SSN and ITIN (official allocation rules), IMEI (Luhn + context), IBAN (mod-97), IPv4, IPv6 (including IPv4-mapped), MAC, JWT, PEM private keys, API keys for 15 providers, and opt-in URL. Greedy over-matches are refined back to the checksum-valid span instead of being discarded.
• Redaction styles: placeholder, label, type-aware mask, remove, plus a pluggable custom replacer.
• Extension points: Detector / PatternDetector (with validator and refine hooks), KeywordDetector deny-list, and opt-in country packs (DetectorPacks.pk — Pakistan CNIC).
• Test suite includes a false-positive trap corpus and seeded property-based round-trip fuzzing; deterministic benchmark under benchmark/.
• Zero runtime dependencies; pure Dart.