dw-kit 1.8.0-rc.1 → 1.8.0-rc.2

package/.dw/config/connectors.local.yml

@@ -6,3 +6,33 @@ telegram:
   bot_token: 8693679403:AAG9FrgUd5Ig9eDTAWnA9RqhbMnShRi3Si0
   allowed_user_ids:
     - 6603235862
+
+# Voice channel (Phase 4 substrate of G-rgoal-realtime-orch).
+# `dw voice` boots a localhost HTTP server with a browser page using Web
+# Speech API for ASR + TTS. Optional orchestrator hybrid: when the regex
+# parser misses, hand the transcript to the configured agent (Claude Code /
+# Codex / Gemini) with a voice-aware system prompt.
+voice:
+  # Default UI language for SpeechRecognition + TTS. User can override on the
+  # page via dropdown. Common values: en-US · vi-VN · ja-JP · ko-KR · zh-CN.
+  # Browser support varies — Chrome / Edge / Safari are best.
+  lang: en-US
+  # Extra languages to surface in the dropdown (beyond `lang` + en-US default).
+  extra_langs: [vi-VN]
+  # Server-side TTS fallback when no native browser/OS voice matches the
+  # selected language. F-23 (G-dogfood-v1.7): a fresh Windows install has
+  # no Vietnamese SAPI voice; we proxy MP3 from Google Translate's public
+  # TTS endpoint. Modes:
+  #   auto    — only when no native voice matches (default; balanced)
+  #   always  — always use server-side (best for cross-machine consistency)
+  #   none    — never use server-side (privacy: spoken text stays local)
+  # Privacy note: in `auto`/`always`, the spoken text leaves the machine to
+  # translate.google.com. Disable with `none` if that is unacceptable.
+  fallback_tts: auto
+  orchestrator:
+    enabled: true          # opt-in. To enable WITHOUT committing it, add the
+                             # same `voice:` section to `.dw/config/connectors.local.yml`
+                             # (gitignored) — the local file overrides this template
+                             # via deep-merge (same pattern as bot_token + allow-list).
+    agent: claude            # must exist in .dw/config/agents.yml; CLI must be on PATH
+    timeout_ms: 30000        # max wait per turn (Claude is typically 2-8s)

package/CLAUDE.md

@@ -3,7 +3,7 @@
 Workflow toolkit codebase. Rules live in `.claude/rules/` (auto-loaded).
 
 **v2.0 direction:** Context-First SDLC Governance Layer (5 pillars — see `.dw/core/PILLARS.md`)
-**Current:** **v1.8.0-rc.1** (2026-05-25) — 4/5 phase substrate for [G-rgoal-realtime-orch](.dw/goals/G-rgoal-realtime-orch/goal.md) voice-meeting Root Goal: persistent CLI-agent session runtime (`dw session *`), Telegram chat bridge (`dw connector telegram setup`) with 1-command interactive wizard, multi-workspace registry (`dw workspace *`), and browser voice MVP (`dw voice`) with hybrid orchestrator fallback (Claude/Codex/Gemini), full bilingual UX (en + vi), voice-not-installed fallback via Google Translate TTS proxy. **v1.7.0 stable** on `main` (2026-05-24). 284/284 smoke tests. Active ADRs: ADR-0001 (Pragmatic Lean), ADR-0005/0006 (Supply-Chain Guard; sunset review 2026-08-12), ADR-0008 (Task Docs v3, v1.5), ADR-0009 (Agent OS, v1.6), ADR-0010 (Goals Layer, v1.7). ADR-0011 (Session Runtime + Voice Orchestrator, v1.8) pending codification.
+**Current:** **v1.8.0-rc.1** (2026-05-25) — 4/5 phase substrate for [G-rgoal-realtime-orch](.dw/goals/G-rgoal-realtime-orch/goal.md) voice-meeting Root Goal: persistent CLI-agent session runtime (`dw session *`), Telegram chat bridge (`dw connector telegram setup`) with 1-command interactive wizard, multi-workspace registry (`dw workspace *`), and browser voice MVP (`dw voice`) with hybrid orchestrator fallback (Claude/Codex/Gemini), full bilingual UX (en + vi), voice-not-installed fallback via Google Translate TTS proxy. **v1.7.0 stable** on `main` (2026-05-24). 284/284 smoke tests. Active ADRs: ADR-0001 (Pragmatic Lean), ADR-0005/0006 (Supply-Chain Guard; sunset review 2026-08-12), ADR-0008 (Task Docs v3, v1.5), ADR-0009 (Agent OS, v1.6), ADR-0010 (Goals Layer, v1.7), ADR-0011 (Session Runtime + Voice Orchestrator, v1.8 — Accepted).
 
 ---
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dw-kit",
-  "version": "1.8.0-rc.1",
+  "version": "1.8.0-rc.2",
   "description": "AI development workflow toolkit — structured, quality-assured, team-ready. From requirements to dashboard.",
   "type": "module",
   "bin": {

package/src/cli.mjs

@@ -352,10 +352,12 @@ export function run(argv) {
   // ─ voice: browser-based voice MVP (G-rgoal phase 4 — KR-A destination) ─
   program
     .command('voice')
-    .description('Browser-based voice MVP — speak dw commands, hear results. Local server + Web Speech API. Spike scope: localhost only.')
-    .option('-p, --port <port>', 'Local server port (default 4500; auto-finds next free)', parseInt)
+    .description('Browser-based voice MVP — speak dw commands, hear results. Web Speech API + optional HTTPS for mobile/LAN reach.')
+    .option('-p, --port <port>', 'Server port (default 4500; auto-finds next free)', parseInt)
     .option('--default-agent <name>', 'Default agent for "start ..." without explicit agent name')
     .option('--no-open', 'Skip auto-opening browser tab (also implied by DW_NO_OPEN=1, CI=true, or non-TTY stdout)')
+    .option('--https', 'Serve over HTTPS using self-signed cert (auto-generated via mkcert or openssl). Required for mic access from non-localhost browsers (mobile / LAN / Tailscale).')
+    .option('--bind <addr>', 'Bind address (default 127.0.0.1). Use 0.0.0.0 to expose on all interfaces (LAN/Tailscale). MUST combine with --https for remote mic access.', '127.0.0.1')
     .action(async (opts) => {
       const { voiceCommand } = await import('./commands/voice.mjs');
       await voiceCommand(opts);