meetcap-capture

The shared, backend-agnostic recording engine behind the meetcap family — the capture state machine (idle/recording/paused events, start-epoch abort protocol, pause-aware duration accounting, ordered crash-safe chunk persistence, MediaRecorder wiring) with the three environment-specific concerns injected as a CaptureBackend:

1. stream acquisition — Electron loopback system audio vs browser getDisplayMedia/caller-provided streams
2. persistence — disk over IPC vs IndexedDB vs memory-only
3. permission snapshots — macOS TCC vs the Permissions API

Application code usually consumes one of the packaged backends instead:

| Package | Environment | |---|---| | meetcap-renderer | Electron renderer (loopback system audio, disk persistence, meeting detection) | | meetcap-web | Any browser (display/tab/camera/stream capture, IndexedDB persistence) | | meetcap-client | Hybrid web/Electron codebases — runtime negotiation between the two |

Import from here when building a custom backend (e.g. a desktop-agent transport):

import { createCaptureRecorder, type CaptureBackend } from 'meetcap-capture'

const recorder = createCaptureRecorder(myBackend, { timesliceMs: 1000 })
recorder.on('chunk', ({ blob }) => upload(blob))
await recorder.start(meeting)

Also exports the shared errors (PermissionDeniedError, StartTimeoutError, BridgeUnavailableError) and pure utilities (pickMimeType, pickVideoMimeType, buildFilename, computeDuration, deniedMedia, withTimeout) that every backend package re-exports.