Detect → Classify → Fix.
Watch the whole loop in eleven seconds.
Pick a streaming incident most ops teams dread — CDN outage, encoder failure, DRM cold start, manifest corruption, regional congestion, or device-specific playback bug — and click Inject failure. Three phase panels light up in sequence as the timeline plays: Detect joins per-channel signals into one correlated timeline, Classify scores the root cause, and Fix runs and verifies the remediation. Each phase unlocks a real next action when it reveals.
Inject failure · Detect → Classify → Fix
The picker on the left controls all three panels. Click Inject failure to start the timeline; each phase unlocks its per-phase CTA the moment it reveals. The Detect and Classify CTAs are pure deep-links — Fix ends the demo by creating a Stripe-hosted checkout session through the installed stripe-billing proxy.
An edge PoP drops mid-rush; rerouting kicks in before viewers notice.
Detect → Classify → Fix panels unlock as the timeline plays.
Detect
Per-channel signals fire within the first 200 ms; reliability agents join them into one correlated timeline.
- +0.0sIncident introduced
—
- +1.2sAnomaly detected
—
- +2.4sQoE + CDN correlated
—
Classify
The hypothesis tree ranks the failure mode with a confidence score before the remediation clock starts.
- +3.6sRoot cause classified
—
Fix
The agent picks the cheapest fix that resolves the ranked root cause, runs it, and verifies QoE has recovered.
- +4.6sRemediation initiated
—
- +6.8sQoE recovered
—
- +9.0sPostmortem generated
—
What each phase actually runs.
The three-phase frame is the same loop Streamwake ships — not three point-products bolted together. Every incident you pick plays the same beat so you can compare apples to apples.
Per-channel signals fire within 200 ms
Edge RTT, segment-fetch latency, rebuffer ratio, CSP errors, and decoder clock-drift are joined into one correlated timeline before the next segment publishes.
Failure mode named with confidence
The hypothesis tree ranks the root cause with a confidence score and rules out the lower hypotheses — usually CDN, encoder, manifest, or device-class.
Cheapest remediation runs and verifies
Routing, cutover, rewarming, ladder-shifting, or manifest rewrite — chosen against the ranked root cause. Re-injected signals confirm QoE has actually recovered.
Pick the failure mode your stack tolerates.
Every class runs a different first-failure probe, a different fallback ladder, and a different remediation playbook — but Detect → Classify → Fix stays constant so you can see exactly where the loop wins or loses for each.
CDN outage
Edge PoP drops during peak — routing agent shifts to healthy PoPs.
Encoder failure
GOP drift collapses the ABR ladder — backup encoder cuts over in 90 s.
DRM failure
License-server cold start times out — DRM agent warms extra shards.
Manifest corruption
Discontinuity tag misaligned — manifest agent self-heals on the fly.
Regional congestion
Last-mile ISP brown-out — ABR agent shifts affected cohort to -2 rung.
Device-specific playback
Codec pack pin breaks Android TV — codec-aware agent rewrites the pack.
Could your current stack
do this automatically?
One channel, $49 — the same recovery your team just watched runs against a live cohort. If we don't beat your current MTTR within 30 days, the next month is free.
Want this on a live signal you control?
The demo above is scripted against synthetic telemetry; the production pipeline runs against your real ingest in under a week. Tell us the channel and the failure mode you're most worried about — we'll come back with a concrete plan.