> ## Documentation Index > Fetch the complete documentation index at: https://developer.suki.ai/llms.txt > Use this file to discover all available pages before exploring further. # Implementation > Implementation questions about the Suki Headless Web SDK The SDK does not replay the audio chunk stream after a network recovery. While replaying chunks might seem like a logical way to fill gaps, it introduces significant complexity to your integration. For example, it would create the following issues: * **Duplication**: It would send duplicate audio ranges (the live stream plus the replayed history). * **Complexity**: You would need to build complex logic to track chunk IDs, deduplicate data, and enforce idempotency. * **Performance**: Replaying history creates a large set of data to process immediately upon reconnection. You may then ask, **"How does the SDK manage data?"** The SDK is intentionally designed to hide this complexity from you. So that you can focus on building your own user interface and features. * **Real-time chunks**: Use the `onAudioChunkAvailable` callback only for live user interface features, such as drawing a waveform. * **Offline robustness**: For data persistence, rely on the SDK's built-in offline manager. The SDK buffers encrypted audio in `IndexedDB` and automatically uploads it to the Suki backend server when the connection returns. We recommend you rely on the **final session artifact** (e.g., the generated note or WAV file) rather than attempting to stitch together a complete recording from the real-time chunk stream.