FrameWorks now publishes a versioned catalog of account events, delivers it live over a GraphQL subscription, and sends it to your HTTPS endpoints as Standard Webhooks-signed requests you can verify with existing libraries.
Where a request is answered, where a stream originates, and where a viewer receives it are three different places. Placement policy lets you state which capacity your streams may use and in what order, such as your own servers first, and enforces it where media starts.
A brief control-plane outage no longer decides who can watch. Viewers who are already watching keep watching, new viewers are admitted from signed answers the media cluster already holds, and one viewer’s approval is never reused for another.
A live stream is a ring buffer on an edge node. A channel that streams for 90 days straight is still one recording, replayable in chapters, and a plan change halfway through never shortens an archive that is already recording.
FrameWorks now runs an EU and a US region. Viewers and broadcasters are handled entirely in their own region, while analytics and billing converge in one place, so a tenant gets one bill that counts every region exactly once.
A FrameWorks edge can run on a server we have never seen. How it proves who it is without a shared secret, why it never holds our internal certificate authority, and why we run our own mesh instead of Tailscale.
Clusters can now source live streams from one another. A viewer near a cluster that never received the broadcast is served from a nearby edge after a single server-to-server pull, instead of being sent somewhere worse.
We used a multi-agent review process to audit FrameWorks across routing, ingest, billing, analytics, auth, and more. The useful part was forcing every claim to come with evidence.