This changelog records Siftfy model versions, API changes, and product releases. Each dated entry explains what changed so you can review the impact on your integration. Use it alongside the API documentation and score guide when updating a client, checking response behavior, or adjusting your spam-review thresholds.
How should I use these release notes?
Read the dated entries below, then check the API documentation for the current request and response contract. The score guide explains how to interpret predictions; validate threshold changes against your own traffic.
content-gate·
Content Gate can now verify Web Bot Auth signatures at siteverify: on Pro, an action can exempt a cryptographically verified agent from the bot block while the spam classifier, deny list, reputation and velocity limits still decide. Find it under Action policy → Verified AI agents.
content-gate·
Content Gate: bot, deny-list, reputation and velocity decisions now count toward your monthly quota and Pro meter.
content-gate·
Content Gate's Activity page now says when verifications ran without the velocity checks: for those submissions the shared counter store was briefly unavailable, so the per-solver submission and distinct-content limits could not be evaluated — a site whose limits never block anything can tell that apart from the rule having run and found nothing.
billing·
Pro now bills requests past the 1,000,000 included each billing period at $0.0001 each, as the pricing page says. Usage is reported to Stripe every 15 minutes and appears on your next invoice; nothing changes if you stay within 1,000,000.
quota·
The Free plan's quota now works as the pricing page says: 10,000 requests per calendar month (UTC), shared by all of an account's keys and resetting on the 1st. It was enforced as 10,000 a day. This month's count starts from zero.
API responses now carry X-RateLimit-Monthly-Used and X-RateLimit-Monthly-Cap in place of X-RateLimit-Daily-Used and X-RateLimit-Daily-Cap. A 429 from either limit now includes Retry-After.
Dashboard → Usage shows this month's count against your quota.
mcp·
New: the Siftfy MCP server at https://api.siftfy.io/mcp. Create a tenant key under Dashboard → Keys, add one line to Claude Code (or a JSON block to Cursor or VS Code), and your agent can manage API keys, read usage, and set up Content Gate sites, policies and deny lists through 17 tools. Revoking, rotating, deleting and switching a site to enforce all preview first and apply only with confirm: true. Every write is logged in the dashboard. Pro plan.
New docs page: MCP server, with setup for each client, the full tool list and the error codes.
privacy·
From 2026-10-13, /v1/predict text from accounts Siftfy's operator runs itself, already kept for the paid shadow trial or live escalation, may be used to train Siftfy's spam model with TypeSafe's verdict or a reviewer's label, kept 12 months. It is forward-only: nothing collected earlier is copied. TypeSafe spam verdicts on it are copied only up to three for each legitimate one. Another customer's /v1/predict text is never used.
docs·
Content Gate docs and the dashboard's install snippet now use widget v1.4, so new sites get the automation checks from day one. v1.js keeps working at its published hash.
New in the Content Gate guide: an upgrade table for v1.js through v1.3, guidance on choosing a bot block threshold, and bot_block_threshold in the policy API example.
New page: Content Gate vs Cloudflare Turnstile, sourced from Cloudflare's developer docs. The live Content Gate demo is now linked from the docs and the CAPTCHA alternatives page.
content-gate·
Widget v1.4 and bot-rules-v2: three more yes/no checks (window geometry no real browser window has, text entered with no key pressed, headless Chrome's default screen) and a header check for a User-Agent whose Chrome version differs from the browser's own client hints. Together they catch a Playwright-driven Chromium with its automation flags removed, headless or headed. v1.4 is a drop-in replacement for v1.3; the header check covers every widget version.
content-gate·
Content Gate now scores every challenge for automation, using only SiftFy's own checks and public network data, and siteverify returns a bot object: score 0-100, coarse reasons, rules_version bot-rules-v1 (null for older challenges).
Enforce sites now block sessions with a bot score of 90 or more by default (decision block, reason bot), for every widget version. Monitor sites only see the score. Change or turn off the threshold per action in Settings or with bot_block_threshold in the policy API.
Widget v1.3 adds yes/no browser automation checks to the challenge request: answers only, used once, nothing stored or identifying. Content Gate is bot-resistant, not bot-proof.
privacy·
/v1/feedback ledger rows still located by the old unsalted hash are deleted from 2026-11-07. A correction reported again before then moves to the keyed hash and keeps its count; one that is not loses its count.
admin·
The admin Labeling tab is gone: Siftfy's model now trains on TypeSafe verdicts and site corrections, not operator labels. The admin page opens straight to the overview, and old ?tab=labeling links land there.
privacy·
From 2026-10-08, TypeSafe verdicts on Content Gate submissions from Siftfy's own sites may be used to train Siftfy's spam model. Only the spam or legitimate verdict is stored with the submission text, kept 12 months; verdicts given earlier are included while the 30-day evaluation record still exists. This replaces the statement of 2026-09-27 that TypeSafe output was not used for training.
From 2026-10-08, spam and not-spam corrections reported to /v1/feedback by Siftfy's own sites (Folio and EchoThread) may be used to train Siftfy's spam model, with the reported text kept 12 months. Other customers' corrections are still stored only as a fingerprint and are not used for training.
The /v1/feedback ledger row is now located by a keyed hash (HMAC) under a server-side secret instead of an unsalted hash, so knowing a tenant ID is no longer enough to confirm a guess at a reported correction. Existing report counts carry over.
privacy·
From 2026-09-27, Content Gate submissions from Siftfy's own sites that our reviewer has labeled may be used to train Siftfy's spam, AI-authorship and automation models, and are kept 12 months for that purpose. Submissions labeled from 2026-09-22 are also used. TypeSafe output is not used for training. Another customer's content still needs its written authorization.
docs·
Spelled out /v1/predict's text validation: whitespace-only or zero-width-only bodies are rejected with 422, and the 20,000-character limit is checked after Unicode normalization, so a shorter raw string can still be rejected.
admin-tabs·
Labeling now has its own admin tab, with a direct link and preserved review progress when switching to Overview.
admin-recovery·
Fixed skipped submissions during paginated review, concurrent-label recovery, and removed-text handling. Submission rows now load independently, cancel obsolete requests, and scroll into view after navigation.
admin·
Admin evaluation now shows 25 submissions per page by default, offers 10, 25, 50, or 100 rows, and saves each submission label immediately with row-level retry and conflict recovery.
reliability·
Checkout retries reuse an open session, and delayed billing events reconcile current subscription access.
Corrected analytics visit counts and performance attribution when consent is given after navigation.
Made analytics preferences scrollable on short screens and long billing errors readable on mobile.
dashboard·
Billing now offers Pro checkout when a legacy test customer is unavailable in live mode, preserving current access while setup is completed.
Marketing analytics now wait for your consent, with Accept, Reject, and revisitable cookie preferences.
Shortened page descriptions and updated integration links and release-note guidance.
docs·
Documented what a spam score means once it reaches max_confidence: the value is censored at the ceiling, so it no longer ranks and no longer supports arithmetic. Compare against the ceiling rather than subtracting from the score.
Stated the population the score is calibrated for — short user-generated and form text — and that reaching the ceiling is the common case on text the model cannot separate, not a corner case.
content-gate·
Content Gate verifications now return max_confidence, the ceiling the spam score was clamped to, so a threshold can be compared against the model's own bound instead of against a constant the score can never reach.
security·
Improved mobile page layouts, keyboard scrolling in code examples, and recovery from missing pages.
Content Gate rejects incomplete model responses; Pro checkout access waits for payment confirmation.
Excluded sign-in and dashboard pages from analytics and removed query strings from page-view reporting.
Hardened inference connections, updated security dependencies, and required complete database and shared-rate-limit integration coverage in CI.
cal3·
Replaced the saturated OTIS checkpoint with a revision-pinned BERT-tiny classifier plus bounded promotional and link signals.
The cal3 1,956-comment evaluation measured 89.2% F1, 96.6% precision, and 82.8% recall; model and lexical evidence remain capped below the auto-block band.
site·
Accessibility fixes from a production QA pass: blog topic-filter chips are now valid toggle buttons (dropped an invalid ARIA role/attribute pairing), the code-block CLI-flag syntax color now meets WCAG AA contrast, and the primary and documentation navigation landmarks are now uniquely labelled.
Reworded the /about page title so it no longer reads as a colon-prefixed fragment.
site·
Added /use-cases/fake-signups: prevention-strategy landing page targeting "prevent fake user registrations", with a three-layer framework and a webhook-style code example (closes #103).
v1·
Public launch of the /v1/predict endpoint and the self-serve dashboard.
Free tier: 10,000 requests / month. Pro tier: $9 + $0.0001/req past 1M included.
Calibrated softmax probabilities, single round-trip, sub-10ms p99 on CPU.