Skip to main content
Fleack logs every meaningful product change here — features, improvements, and fixes — dated and described in plain language. Use this page to understand what’s available in the version you’re running, track what’s changed since you last checked in, and see where the product is headed.

​
v0.1.0 — Initial public release · April 19, 2026

The first version of Fleack open to external customers. Everything in this release was built to get you from DNS switch to live A/B test with no SDK and no app changes.

​
What’s in this release

  • Edge interception — Fleack intercepts your API traffic at the edge with a sub-40 ms p95 latency budget. Fail-open by design: if anything goes wrong on Fleack’s side, your users see the original backend response.
  • Backoffice — a full management UI with multi-organisation support and role-based access control (owner, admin, member) plus a super-admin tier for the operator account.
  • AI lever discovery — Fleack analyses captured endpoint response bodies and automatically identifies which parameters are safe to test. No manual annotation required.
  • Manual lever wizard — prefer to declare your own test parameters? Browse the scalar values in any captured response body, pick the one you want, label it, set the type, and ship.
  • Test creation — create tests with up to N variants, set weighted traffic allocation per variant, and get sticky per-user assignment via a deterministic hash of the test ID and user identity.
  • Bayesian results engine — results are computed using a Beta posterior with a Jeffreys prior and a 3,000-sample Monte Carlo simulation for win probability against the control. No p-values, no arbitrary significance thresholds.
  • Multi-metric tests — choose your primary success metric: conversion rate (with a configurable time window), day-N retention, revenue (scalar delta on a profile attribute), or raw scalar value.
  • Natural-language segments — build audience segments from auto-discovered profile attributes including level, ARPU, days since install, country, and platform. No code required.
  • Anonymous-user attribution — for endpoints that don’t carry an authenticated user identity, Fleack derives an anonymous identity from the User-Agent and IP hash. Exposure counters work correctly on config endpoints and other unauthenticated paths.
  • Honest body sampling — captured response bodies always reflect what your backend actually returned, never the rewritten variant value. Even when a test is live on a path, the sample you see in the backoffice is the untouched original.
  • Public docs at docs.fleack.io.

​
Known limits at v0.1

  • No public API reference yet. Fleack’s management API is currently accessible only through the backoffice UI. We’ll publish an OpenAPI spec when there’s real demand for programmatic test creation — reach out at contact@fleack.io if that’s you.
  • English only. Documentation is in English only. Other languages may follow if demand warrants it.
  • No SLA on free and starter tiers. Pro tiers carry a 99.9% uptime SLA; Enterprise tiers carry 99.99%. Free and starter plans have no SLA commitment.

​
What’s coming next

This is a rolling preview of what we’re working on. These are directions, not commitments.
  • Per-tenant SSO for Enterprise — SAML support and Okta integration for teams that manage identity centrally.

​
How we version

Fleack follows loose SemVer for the platform as a whole.
IncrementWhen it’s used
Major (v1.0.0)A breaking change to documented behaviour that customers rely on
Minor (v0.X.0)New features with no breaking changes
Patch (v0.0.X)Bug fixes, performance improvements, security updates
Every meaningful change ships with a changelog entry. Internal refactors that don’t affect behaviour aren’t logged here — those live in the Git history.