What we shipped, and when.

Replay automations against any historical profile.

You can now select any profile in the Console, pick an automation, and run it against that user's full event history as if today were a different date. Useful for debugging "why didn't this fire?" and for testing new flows against real cohorts before they go live.

• New Replay tab on every automation, with simulate mode (no real sends).
• Diffs the run against what actually happened in production.
• Replays write to a separate audit log and can be exported as JSON.

Segment counts now update live as you build.

The segment builder used to recompute counts on save. It now streams an estimated count as you add each filter, with the full count finalising in < 2s for most workspaces. Caching is per-workspace, scoped to the current event window.

• Estimated count appears in < 200ms for < 10M-event workspaces.
• Switched the underlying engine from per-query scans to a Bloom-filtered preflight.
• Existing saved segments unchanged, only the build experience is faster.

SMTP2GO joins the connected providers list.

You can now send through SMTP2GO from any GetFluxly automation or campaign. Adds to Amazon SES, Mailgun, Postmark, SendGrid, and generic SMTP. Setup takes about 90 seconds. Paste your API key, verify a domain, you're done.

• Webhook ingest for opens, clicks, bounces, and complaints.
• Region-aware (EU and US endpoints) and respects your provider's rate limits.
• Send outcomes attach back to the profile, same as every other provider.

Identity stitching: fixed a race when identify() runs before init.

If your app called window.gflux.identify() in the same tick as the SDK boot, the identify could land before the SDK had read its anonymous ID from storage, producing an orphan profile. The SDK now buffers identify calls until init has fully resolved.

• No code changes required. Affected workspaces will see orphan-rate drop to ~0.
• If you hit this previously, the Console's Identity links → Resolve tool can merge the orphans.

Live event stream in the Console.

The Events page now has a tail-mode that streams new events as they arrive over a long-lived SSE connection. Default page-load still shows the last 24 hours; press L to switch to live tail.

• Stream is filtered by your current Events filters (event name, profile, anonymous ID, time range).
• Backpressure-safe: pauses the stream automatically if rendering can't keep up.
• Keyboard: L tail, P pause, Esc exit.

HTTP Events API: retry guidance is now explicit.

The HTTP Events API docs now call out that ingest is at-least-once. Server callers should retry 5xx responses with backoff and tolerate duplicate events after unknown network outcomes.

• Backwards compatible. The event payload shape stays the same.
• Recommended pattern: use stable event names and keep backend handlers safe to run more than once.
• Server examples now show short timeouts and background delivery instead of blocking user-facing requests.

API: /v0 endpoints retired.

The /v0 namespace, which has been deprecated since November, is now removed. All endpoints live under /v1. Beta workspaces have already been migrated automatically; if you have client code calling /v0/track, swap the path to /v1/track. The request and response shapes are identical.