docs: real-time audio bytes integration (#3373)

Add new section on Real-Time Audio Bytes integration, including guide, key points, example request, and use cases

Author: vats98754

docs/doc/developer/apps/Integrations.mdx

@@ -40,6 +40,36 @@ These apps process conversation transcripts as they occur, enabling real-time an
 - Real-time web searches or fact-checking
 - Emotional state analysis and supportive responses
 
+### 3. 🔊 Real-Time Audio Bytes (audio_bytes)
+
+#### Guide:
+
+Omi can stream raw audio bytes from a DevKit directly to an external endpoint. This integration uses the trigger type `audio_bytes` and is useful for applications that need raw PCM audio (VAD, custom ASR, feature extraction, or long-term storage).
+
+Key points:
+
+- Trigger name: `audio_bytes`
+- Request: POST
+- Query parameters appended by the backend: `sample_rate` (e.g. 16000) and `uid`
+- Content-Type: `application/octet-stream` (raw PCM16 bytes)
+
+Example request:
+
+`POST /your-endpoint?sample_rate=16000&uid=user123` with the request body containing raw PCM16 (16-bit little-endian) audio bytes.
+
+Notes and recommendations:
+
+- The bytes are raw PCM16 (16-bit little-endian), 2 bytes per sample. To produce a playable WAV file you must prepend a WAV header and concatenate the received chunks.
+- You control delivery frequency in the Omi app Developer Settings (the webhook field accepts `url,seconds` — the second value configures how many seconds of audio are sent per request).
+- For an example implementation and deployment-ready code, see the Real-Time Audio Streaming guide: `doc/developer/apps/AudioStreaming.mdx`.
+
+#### Example Use Cases
+
+- Voice activity detection (VAD) and endpointing
+- Feed custom ASR or speech models for research or improved accuracy
+- Extract audio features (spectrograms, embeddings) in real-time
+- Store raw audio chunks to cloud storage for later processing
+
 ## Creating an Integration App
 
 ### Step 1: Define Your App 🎯