From b54130d991da4ef9426ce486d34bd4c8ffd3fd76 Mon Sep 17 00:00:00 2001
From: Nexus Dev <nexus@local>
Date: Sat, 4 Apr 2026 03:23:44 +0000
Subject: [PATCH] docs(39): auto-generated context (discuss skipped)

---
 .../phases/39-voice-polish/39-CONTEXT.md      | 62 +++++++++++++++++++
 1 file changed, 62 insertions(+)
 create mode 100644 .planning/phases/39-voice-polish/39-CONTEXT.md

diff --git a/.planning/phases/39-voice-polish/39-CONTEXT.md b/.planning/phases/39-voice-polish/39-CONTEXT.md
new file mode 100644
index 00000000..fcd8e2f9
--- /dev/null
+++ b/.planning/phases/39-voice-polish/39-CONTEXT.md
@@ -0,0 +1,62 @@
+# Phase 39: Voice Polish - Context
+
+**Gathered:** 2026-04-04
+**Status:** Ready for planning
+**Mode:** Auto-generated (discuss skipped via workflow.skip_discuss)
+
+<domain>
+## Phase Boundary
+
+Voice responses begin playing before synthesis is complete (sentence-buffered), a single response can be synthesized in multiple languages simultaneously, and new installs can detect STT/TTS hardware capability during onboarding and enable voice in one step.
+
+Requirements: VPIPE-07, VPIPE-08, ONBRD-01, ONBRD-02
+
+</domain>
+
+<decisions>
+## Implementation Decisions
+
+### Claude's Discretion
+All implementation choices are at Claude's discretion — discuss phase was skipped per user setting.
+
+Key research findings to incorporate:
+- Sentence-buffered TTS: split response on sentence boundaries (.!?), synthesize first sentence immediately, start playback while subsequent sentences synthesize
+- Multi-language TTS: Piper supports multiple language models; user requests same text as audio in multiple languages (e.g. English + Danish) without a second agent call
+- Onboarding hardware detection: extend existing hardware probe to check for Whisper/Piper binary availability and hardware capability
+- VoiceStep already exists from v1.5 (Phase 34) — enhance it with hardware probe results rather than creating a new step
+- Use existing systeminformation probe pattern from Phase 30
+- Sentence splitting: simple regex on .!? followed by whitespace; no NLP library needed
+
+</decisions>
+
+<code_context>
+## Existing Code Insights
+
+### Reusable Assets
+- `server/src/services/voice-pipeline.ts` — VoicePipelineService (synthesize already does sentence chunking)
+- `ui/src/components/ChatVoicePlayer.tsx` — audio playback (needs streaming support)
+- `ui/src/components/onboarding/VoiceStep.tsx` — existing voice enable/skip step
+- `server/src/routes/voice.ts` — POST /api/synthesize
+- Hardware detection from Phase 30 (systeminformation probe)
+
+### Integration Points
+- `server/src/routes/voice.ts` — new endpoint for multi-language synthesis
+- `ui/src/components/ChatVoicePlayer.tsx` — sentence-buffered playback
+- `ui/src/components/onboarding/VoiceStep.tsx` — hardware capability display
+- `server/src/services/nexus-settings.ts` — piperBinaryPath, whisperBinaryPath
+
+</code_context>
+
+<specifics>
+## Specific Ideas
+
+No specific requirements — discuss phase skipped.
+
+</specifics>
+
+<deferred>
+## Deferred Ideas
+
+None — discuss phase skipped.
+
+</deferred>