docs: wake word detection section, updated roadmap

- architecture.mdx: added Wake Word Detection section
- overview.mdx: added Wake step to how-it-works, updated roadmap blurb
- roadmap.mdx: wake word moved to In Progress

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Author: 0xharkirat

content/docs/hark/architecture.mdx

@@ -57,6 +57,12 @@ Main (Dart)    -> HarkMainApi (Kotlin)     -> Overlay relay    -> HarkOverlayFlu
 4. The overlay renders chat bubbles with app icons, a keyboard/mic toggle, and auto-activates the mic on open.
 5. On dismiss, the overlay engine is kept warm for the next invocation.
 
+## Wake word detection
+
+Hark supports hands-free activation via the wake phrase "Hey Hark". Wake word detection runs on-device using [openWakeWord](https://github.com/dscripka/openWakeWord) (Apache 2.0) with ONNX Runtime for inference. A custom-trained model (201KB) listens continuously for the trigger phrase and auto-starts the microphone when detected. The wake word engine is integrated through Pigeon bindings in `hark_platform`, consistent with the rest of the native bridge.
+
+AudioRecord is a shared resource on Android, so the wake word listener and STT cannot run simultaneously. When the wake phrase is detected, the wake word engine stops its audio stream before STT begins listening. Once the voice command completes, the wake word engine restarts. This mutual exclusion is handled automatically in the platform plugin. The wake word engine persists even when the app is backgrounded, so detection continues across app switches. Background service support (Phase 2) will allow detection when the app is fully closed.
+
 ## Two-stage resolution
 
 ### Stage 1 - Intent selection (EmbeddingGemma 308M)

content/docs/hark/overview.mdx

@@ -20,6 +20,7 @@ Hark is the open-source assistant built on [OACP](/docs/oacp/what-is-oacp). It d
 
 ## How it works at a glance
 
+0. **Wake** - on-device wake word detection listens for "Hey Hark" and activates the assistant hands-free.
 1. **Listen** - on-device speech-to-text captures your voice command.
 2. **Discover** - scans installed apps for OACP capability manifests via Android `ContentProvider`.
 3. **Resolve** - two-stage on-device AI pipeline matches your command to the right app action and extracts parameters.
@@ -92,4 +93,4 @@ flutter run
 
 ## Roadmap
 
-The short version: the protocol and Kotlin SDK ship today, and the lightweight assistant overlay is shipped. Hark is focused next on self-hosted inference, better speech input, wake word activation, and chat persistence. See the [roadmap](/docs/roadmap) for the tracked priorities.
+The short version: the protocol and Kotlin SDK ship today, and the lightweight assistant overlay is shipped. Wake word detection ("Hey Hark") is in progress with in-app listening working now and a background service coming next. Hark is also focused on self-hosted inference, better speech input, and chat persistence. See the [roadmap](/docs/roadmap) for the tracked priorities.

content/docs/roadmap.mdx

@@ -35,14 +35,13 @@ Other SDKs are not documented as installable products until there is code to ins
 
 ### In progress
 
+- **Wake word** - "Hey Hark" on-device detection using openWakeWord with ONNX Runtime. Phase 1 (in-app listening, auto-mic activation) is working. Phase 2 (background service for detection when the app is fully closed) is next.
 - **Self-hosted inference** - connect Hark to Ollama or LM Studio on a local network for unlimited context and `OACP.md` consumption
 - **BYOK cloud fallback** - bring-your-own-key for OpenAI, Gemini, Anthropic when no local model is available
 - **Better STT** - evaluate whisper.cpp and sherpa-onnx for fully on-device speech recognition
 - **Disambiguation UI** - show top candidates as chips when scores are close and learn from user feedback
 
 ### Planned
-
-- **Wake word** - "Hey Hark" activation without touching the phone
 - **Personalization** - assistant learns user preferences and per-app vocabulary over time
 - **Gemma 4 single-model pipeline** - once `flutter_gemma` supports it, collapse the two-stage pipeline into a single call
 - **iOS** - exploring feasibility