Documentation

LakeRTC

A Databricks App that turns raw WebRTC statistics into call-quality insight — E-model MOS scores, per-second timelines, quality trends across all your calls, and AI root-cause diagnosis — running 100% inside your own Databricks workspace. Zero external egress.

InstallMap two resources, run the DDL, grant access — about ten minutes end to end. Capture a dumpExport call statistics from chrome://webrtc-internals and upload them. ScreensOverview, Sessions, Breakdown, Compare, and Session detail — timeline, streams, connections, devices, event log and AI diagnosis. Data modelThe Delta tables, the QoE scoring model, and how to build on them.

Requirements

Unity Catalog enabled in your workspace
A serverless SQL warehouse
Databricks Apps enabled
Optional, for AI diagnosis: a pay-per-token Foundation Model API endpoint

How it keeps your data yours

Storage is your Unity Catalog, compute is your SQL warehouse, and the AI diagnosis runs on your workspace's Model Serving endpoints. The app declares no external domains, makes no outbound calls, and collects no telemetry. TURN credentials are redacted and SDP bodies dropped at parse time. And what you put in, you can take out: delete any call from inside the app, or clear every table at once as an admin.

Support

Questions, early access, or a demo with your own dumps: [email protected]