v0.3.0 — Identity context developer preview

Janitor v0.3 developer preview adds a server-side identity directory alongside browser continuity and Jev risk scoring. Register person or agent principals, associate verified email/external/public-key references, and assess scoped, expiring, revocable delegation against server-authenticated context. D1 and Postgres store hashed identity keys and support cascading erasure. The application owns authentication and access policy.

The public site now leads with Jev and identity context for the agentic era. Read the working APIs and trust boundaries at https://janitor.holycoders.io/docs/agentic-identity/. This remains a prelaunch preview; APIs may evolve.

Validation: 98 unit/integration tests, six browser tests, typecheck, lint, Next.js production build, Cloudflare dry-run build, both database migrations, and isolated package installation passed. External Jev calls are mocked; no calibrated takeover or human/agent detection accuracy is claimed.

The bundle contains seven installable ESM/TypeScript tarballs and a consumer package.json with local pnpm overrides. Extract into an empty folder and run pnpm install. These package names are not published on npm.

Janitor v0.2.0 — Behavior, account linking, and browser benchmarks

Janitor v0.2.0 adds optional behavior summaries, verified account labels across devices, and a reproducible real-browser benchmark.

• Opt into mouse travel, active movement time, direction changes, pauses, wheel summaries, and aggregate interaction timing with behavior: "extended". Counts remain the default. No absolute coordinates, raw event trails, actual keys or form contents are collected.
• Configure subjectLinking: { secret, namespace } and pass a server-verified authenticatedSubject to handle(). Requests authenticated to the same account receive the same opaque HMAC subjectId across devices, while retaining separate browser IDs. Anonymous requests do not receive a subject. No account graph or extra tables are stored.
• pnpm benchmark:browser collects 30 observations using real Chromium, Firefox and WebKit and runs 78 deterministic identity trials. The report includes 24/24 isolated restorations and 6/6 false matches for unseen profiles identical to the only enrolled profile. This is a controlled single-host experiment, not population accuracy or a human/bot benchmark.
• 79 unit/integration tests and six browser integration/UI tests pass locally, alongside strict type checks, lint and example production builds.

Documentation: https://janitor.holycoders.io/docs/extensions/
Benchmark methodology and results: https://janitor.holycoders.io/docs/benchmarks/

Download janitor-0.2.0.tar.gz, extract into an empty folder and run pnpm install. It contains seven compiled packages and a consumer manifest with local pnpm overrides. Packages remain unpublished on npm. Deploy updated server adapters before enabling extended behavior in the browser; v0.1 adapters reject the new fields.

Browser signals cannot prove person identity or malicious intent. Deliberately hidden values remain missing. AI inference was not used in the controlled benchmark; risk calibration remains unverified. No database migration is needed for this release.

Janitor v0.1.0

Janitor is a small TypeScript library for durable first-party browser identity, with optional Jev-assisted matching and risk scoring.

• Cookie continuity plus bounded observation-history matching.
• Safe browser collection and aggregate event counts.
• Cloudflare D1 and Postgres storage, with Node, Vercel, and Cloudflare adapters.
• Direct Jev and Workers AI evaluators with typed outputs and deterministic fallback.
• Three runnable example applications, privacy documentation, and practical tests.
• A public documentation site and synthetic playground at https://janitor.holycoders.io.

Install the prebuilt packages

Download janitor-0.1.0.tar.gz, extract it into an empty directory, then run pnpm install. The bundle includes seven ESM/declaration package archives and a consumer manifest with the pnpm overrides needed to resolve the unpublished sibling packages locally. For an existing application, merge its dependencies and overrides and adjust the archive paths. Node 22.12+ and pnpm 9.12.0 are recommended.

These packages are not published on npm. Confidence thresholds and AI judgments are experimental and require application-specific calibration; identity is not authentication. External AI calls are mocked in the test suite. The library never blocks users or displays a CAPTCHA.

MIT licensed. See the README for the public API, examples, migrations, and validation details.