feat(pii): custom user-supplied regex patterns for redaction

[TheodoreSpeaks]

Summary

• Add user-supplied custom regex patterns to PII redaction — { name, regex, replacement }, matched and redacted everywhere PII runs: block input, block outputs, logs, and the guardrails block/tool
• Per-stage inline storage on PiiStagePolicy (no DB migration — data_retention_settings is a schemaless JSON column); bounds enforced at the contract boundary (name ≤100, regex 1–512, replacement ≤100, ≤20 patterns/stage)
• Presidio service builds per-request ad-hoc recognizers + replace-operators; a bad regex fails fast as 400; the spaCy-skip fast path stays intact (custom entities are non-NER). Custom-only requests never widen to detect-all
• ReDoS backstop: REGEX_TIMEOUT_SECONDS=2 (Presidio parses it with int()) + client-side safe-regex2 validation in the shared Custom Patterns editor
• UI: Custom Patterns section in Data retention (per stage) + a Custom Patterns table on the Guardrails block (Validation Type = PII)

Type of Change

• New feature

Testing

• Built the PII image and validated live: custom-only redaction, custom + built-in together, block-mode detection, 400 on invalid regex, nlp=skip fast path intact
• 43 TS/TSX unit tests (schema/contract/resolver/normalize/editor), type-check, check:api-validation:strict, check:migrations, py_compile all green
• Note: Presidio isn't installable in CI, so the server paths are validated via the local image build

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		Jul 17, 2026 5:29pm

[cursor]

PR Summary

Medium Risk
Changes PII/data-retention behavior across execution and logging with user-supplied regex; mitigations include boundary validation, Presidio compile/timeout, and existing fail-closed scrub/throw paths on mask failures.

Overview
Adds user-defined regex patterns (name, regex, replacement) to PII redaction across workflow input, block outputs, logs, guardrails validate/mask-batch, and the Guardrails block.

The Presidio service accepts optional patterns on analyze/anonymize/redact routes: ad-hoc recognizers with score 1.0, per-pattern replace operators (tokens wrapped in angle brackets), invalid regex → 400, and entity resolution so guardrails “empty = detect all” still runs custom patterns while data-retention “empty entity list + patterns” means custom-only.

App layer: customPatternSchema on API contracts (syntax + safe-regex2), sanitizeCustomPatterns, retention resolution/UI (CustomPatternsEditor), executor and log redaction paths pass customPatterns through to maskPIIBatch; guardrails tool maps table rows and coerces null entity lists. Docker: REGEX_TIMEOUT_SECONDS=2 on the PII image.

^{Reviewed by Cursor Bugbot for commit b6ed40d. Bugbot is set up for automated code reviews on this repo. Configure here.}

[greptile-apps]

Greptile Summary

This PR adds user-defined regex patterns to PII redaction across the application. The main changes are:

• Adds bounded custom-pattern contracts and shared regex safety validation.
• Applies custom patterns to inputs, block outputs, logs, and guardrails.
• Builds per-request Presidio recognizers and replacement operators.
• Adds custom-pattern editors for data retention and guardrails.
• Preserves custom-only policies and regex-only processing paths.

Confidence Score: 5/5

This looks safe to merge.

• The updated contract rejects malformed and known unsafe regexes before persistence or execution.
• Custom patterns are carried through the main redaction paths without a confirmed blocking failure.
• No distinct production issue remains within the follow-up scope.

Important Files Changed

Filename	Overview
apps/sim/lib/api/contracts/primitives.ts	Adds bounded custom-pattern schemas with syntax and unsafe-backtracking validation.
apps/sim/lib/guardrails/validate_regex.ts	Provides shared regex validation for API contracts and the editor.
apps/pii/server.py	Adds per-request custom recognizers, entity resolution, and replacement operators.
apps/sim/lib/guardrails/validate_pii.ts	Carries custom patterns through PII analysis and masking requests.
apps/sim/lib/billing/retention.ts	Resolves custom patterns into effective per-stage redaction policies.
apps/sim/components/pii/custom-patterns-editor.tsx	Adds the shared editor for creating and validating custom patterns.

_{Reviews (5): Last reviewed commit: "fix(pii): keep detect-all when a custom ..." | Re-trigger Greptile}

[TheodoreSpeaks]

@greptile review

[TheodoreSpeaks]

@greptile review

[TheodoreSpeaks]

@greptile review

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

There are 4 total unresolved issues (including 3 from previous reviews).

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 5e86c7d. Configure here.}

[TheodoreSpeaks]

@greptile review