The bot joins a Google Meet call through a Playwright-controlled Chromium browser and hooks its WebRTC peer connections to access every incoming media track. Video tracks are identified by matching each participant tile's DOM SSRC attribute, while audio tracks are identified by correlating Meet's speaking indicators with audio-level spikes and passing every candidate through a 4-strategy bonding filter before locking it to a participant. Each participant then gets a dedicated canvas + MediaRecorder pipeline that records audio and video together, streaming chunks over a local WebSocket to Node.js, which writes them as segments on disk. Mid-call re-negotiations are caught by two-signal detection and handled with a dual-recorder hand-off, and on stop a 3-phase traffic-light brake finalizes every file before FFmpeg normalizes and stitches the segments into one synced .mp4 per participant with a timeline of their join offsets.