Introduction

What is ObserveRTC?#

ObserveRTC is an open-source toolkit for monitoring WebRTC applications. It is a set of libraries, not a service you sign up for: you include what you need, keep your data, and build the parts that are specific to your product.

The problem it solves is narrow and real. RTCPeerConnection.getStats() gives you a flat map of monotonically increasing counters joined by string ids. Almost every question you actually want to ask — what is the bitrate right now? is this video frozen? is the CPU the bottleneck or the network? is the whole room affected or one person? — requires differencing those counters, normalising browser differences, applying hysteresis so you do not alert on one bad tick, and then correlating across participants.

That is the work these libraries do.

What you get#

Why ObserveRTC#

• Open source. No hidden magic and no secret sauce — you can read exactly why a detector fired.
• You own the data. It runs in your browser code and on your servers. Nothing is sent anywhere you did not send it.
• Modular. Take the client library alone for in-call UX. Add the observer when you need cross-participant answers. Replace any piece with your own — they agree on a schema, not on an implementation.
• Honest about its boundaries. It does not ship a database, a dashboard or an alert router, because you already have opinions about those. It gives you events and a live model.

What it is good for#

Who it is for#

• Application developers debugging real WebRTC problems in production.
• Product owners who want quality trends they can act on rather than a vendor’s opaque score.
• Operations teams who need to know quickly whether an incident is one user, one room, or the fleet.

Getting started#

npm install @observertc/client-monitor-js

import { ClientMonitor } from "@observertc/client-monitor-js";

const monitor = new ClientMonitor({
    clientId: "user-42",
    callId: "room-abc",
    collectingPeriodInMs: 2000,
    samplingPeriodInMs: 4000,
});

monitor.addSource(peerConnection);

monitor.on("issue", (issue) => console.warn(issue.type, issue.payload));
monitor.on("sample-created", (sample) => transport.send(sample));

npm install @observertc/observer-js

import { Observer } from "@observertc/observer-js";

const observer = new Observer({ closeClientIfIdleForMs: 60_000 });

observer.on("client-issue-resolved", ({ observedClient, resolvedIssue }) => {
    analytics.track("quality_episode", {
        clientId: observedClient.clientId,
        type: resolvedIssue.type,
        durationInMs: resolvedIssue.durationInMs,
    });
});

transport.on("sample", (sample) => observer.accept(sample));

Next steps#