v1.22.0

feat(proxy-engine): add on-demand TTS caching for voicemail and IVR prompts
v1.21.0
2026-04-12 20:45:08 +00:00 · 2026-04-12 20:45:08 +00:00 · 2026-04-11 20:04:56 +00:00 · 2026-04-11 20:04:56 +00:00 · 2026-04-11 19:20:14 +00:00 · 2026-04-11 19:20:14 +00:00
32 changed files with 1247 additions and 1143 deletions
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -1,41 +1,103 @@
 # Project Notes

-## Architecture: Hub Model (Call as Centerpiece)
+## Architecture: Hub Model in Rust (Call as Centerpiece)

-All call logic lives in `ts/call/`. The Call is the central entity with N legs.
+The call hub lives in the Rust proxy-engine (`rust/crates/proxy-engine/`). TypeScript is the **control plane only** — it configures the engine, sends high-level commands (`hangup`, `make_call`, `webrtc_offer`, etc.), and receives events (`incoming_call`, `call_answered`, `device_registered`, `webrtc_audio_rx`, …). No raw SIP/RTP ever touches TypeScript.

-### Key Files
- `ts/call/call-manager.ts` — singleton registry, factory methods, SIP routing
- `ts/call/call.ts` — the hub: owns legs, media forwarding
- `ts/call/sip-leg.ts` — SIP device/provider connection (wraps SipDialog)
- `ts/call/webrtc-leg.ts` — browser WebRTC connection (wraps werift PeerConnection)
- `ts/call/rtp-port-pool.ts` — unified RTP port pool
- `ts/sipproxy.ts` — thin bootstrap wiring everything together
- `ts/webrtcbridge.ts` — browser device registration (signaling only)
+The `Call` is still the central entity: it owns N legs and a central mixer task that provides mix-minus audio to all participants. Legs can be `SipProvider`, `SipDevice`, `WebRtc`, or `Tool` (recording/transcription observer).

-### WebRTC Browser Call Flow (Critical)
+### Key Rust files (`rust/crates/proxy-engine/src/`)

-The browser call flow has a specific signaling order that MUST be followed:
+- `call_manager.rs` — singleton registry, call factory methods, SIP routing (inbound/outbound/passthrough), B2BUA state machine, inbound route resolution
+- `call.rs` — the `Call` hub + `LegInfo` struct, owns legs and the mixer task
+- `sip_leg.rs` — full SIP dialog management for B2BUA legs (INVITE, 407 auth retry, BYE, CANCEL, early media)
+- `rtp.rs` — RTP port pool (uses `Weak<UdpSocket>` so calls auto-release ports on drop) + RTP header helpers
+- `mixer.rs` — 20 ms-tick mix-minus engine (48 kHz f32 internal, per-leg transcoding via `codec-lib`, per-leg denoising)
+- `jitter_buffer.rs` — per-leg reordering/packet-loss compensation
+- `leg_io.rs` — spawns inbound/outbound RTP I/O tasks per SIP leg
+- `webrtc_engine.rs` — browser WebRTC sessions (werift-rs based), ICE/DTLS/SRTP
+- `provider.rs` — SIP trunk registrations, public-IP detection via Via `received=`
+- `registrar.rs` — accepts REGISTER from SIP phones, tracks contacts (push-based device status)
+- `config.rs` — `AppConfig` deserialized from TS, route resolvers (`resolve_outbound_route`, `resolve_inbound_route`)
+- `main.rs` — IPC command dispatcher (`handle_command`), event emitter, top-level SIP packet router
+- `sip_transport.rs` — owning wrapper around the main SIP UDP socket
+- `voicemail.rs` / `recorder.rs` / `audio_player.rs` / `tts.rs` — media subsystems
+- `tool_leg.rs` — per-source observer audio for recording/transcription tools
+- `ipc.rs` — event-emission helper used throughout

-1. `POST /api/call` with browser deviceId → CallManager creates Call, saves pending state, notifies browser via `webrtc-incoming`
-2. Browser sends `webrtc-offer` (with its own `sessionId`) → CallManager creates a **standalone** WebRtcLeg (NOT attached to any call yet)
-3. Browser sends `webrtc-accept` (with `callId` + `sessionId`) → CallManager links the standalone WebRtcLeg to the Call, then starts the SIP provider leg
+### Key TS files (control plane)

-**The WebRtcLeg CANNOT be created at call creation time** because the browser's session ID is unknown until the `webrtc-offer` arrives.
+- `ts/sipproxy.ts` — entrypoint, wires the proxy engine bridge + web UI + WebRTC signaling
+- `ts/proxybridge.ts` — `@push.rocks/smartrust` bridge to the Rust binary, typed `TProxyCommands` map
+- `ts/config.ts` — JSON config loader (`IAppConfig`, `IProviderConfig`, etc.), sent to Rust via `configure`
+- `ts/voicebox.ts` — voicemail metadata persistence (WAV files live in `.nogit/voicemail/{boxId}/`)
+- `ts/webrtcbridge.ts` — browser WebSocket signaling, browser device registry (`deviceIdToWs`)
+- `ts/call/prompt-cache.ts` — the only remaining file under `ts/call/` (IVR prompt caching)

-### WebRTC Audio Return Channel (Critical)
+### Rust SIP protocol library

-The SIP→browser audio path works through the Call hub:
+`rust/crates/sip-proto/` is a zero-dependency SIP data library (parse/build/mutate/serialize messages, dialog management, SDP helpers, digest auth). Do not add transport or timer logic there — it's purely data-level.

-1. Provider sends RTP to SipLeg's socket
-2. SipLeg's `onRtpReceived` fires → Call hub's `forwardRtp`
-3. Call hub calls `webrtcLeg.sendRtp(data)` → which calls `forwardToBrowser()`
-4. `forwardToBrowser` transcodes (G.722→Opus) and sends via `sender.sendRtp()` (WebRTC PeerConnection)
+## Event-push architecture for device status

-**`WebRtcLeg.sendRtp()` MUST feed into `forwardToBrowser()`** (the WebRTC PeerConnection path), NOT send to a UDP address. This was a bug that caused one-way audio.
+Device status flows **via push events**, not pull-based IPC queries:

-The browser→SIP direction works independently: `ontrack.onReceiveRtp` → `forwardToSip()` → transcodes → sends directly to provider's media endpoint via UDP.
+1. Rust emits `device_registered` when a phone REGISTERs
+2. TS `sipproxy.ts` maintains a `deviceStatuses` Map, updated from the event
+3. Map snapshot goes into the WebSocket `status` broadcast
+4. Web UI (`ts_web/elements/sipproxy-devices.ts`) reads it from the push stream

-### SIP Protocol Library
+There used to be a `get_status` pull IPC for this, but it was never called from TS and has been removed. If a new dashboard ever needs a pull-based snapshot, the push Map is the right source to read from.

-`ts/sip/` is a zero-dependency SIP protocol library. Do not add transport or timer logic there — it's purely data-level (parse/build/mutate/serialize).
+## Inbound routing (wired in Commit 4 of the cleanup PR)
+
+Inbound route resolution goes through `config.resolve_inbound_route(provider_id, called_number, caller_number)` inside `create_inbound_call` (call_manager.rs). The result carries a `ring_browsers` flag that propagates to the `incoming_call` event; `ts/sipproxy.ts` gates the `webrtc-incoming` browser fan-out behind that flag.
+
+**Known limitations / TODOs** (documented in code at `create_inbound_call`):
+- Multi-target inbound fork is not yet implemented — only the first registered device from `route.device_ids` is rung.
+- `ring_browsers` is **informational only**: browsers see a toast but do not race the SIP device to answer. True first-to-answer-wins requires a multi-leg fork + per-leg CANCEL, which is not built yet.
+- `voicemail_box`, `ivr_menu_id`, `no_answer_timeout` are resolved but not yet honored downstream.
+
+## WebRTC Browser Call Flow (Critical)
+
+The browser call signaling order is strict:
+
+1. Browser initiates outbound via a TS API (e.g. `POST /api/call`) — TS creates a pending call in the Rust engine via `make_call` and notifies the browser with a `webrtc-incoming` push.
+2. Browser sends `webrtc-offer` (with its own `sessionId`) → Rust `handle_webrtc_offer` creates a **standalone** WebRTC session (NOT attached to any call yet).
+3. Browser sends `webrtc_link` (with `callId` + `sessionId`) → Rust links the standalone session to the Call and wires the WebRTC leg through the mixer.
+
+**The WebRTC leg cannot be fully attached at call-creation time** because the browser's session ID is unknown until the `webrtc-offer` arrives.
+
+### WebRTC audio return channel (Critical)
+
+The SIP→browser audio path goes through the mixer, not a direct RTP relay:
+
+1. Provider sends RTP → received on the provider leg's UDP socket (`leg_io::spawn_sip_inbound`)
+2. Packet flows through `jitter_buffer` → mixer's inbound mpsc channel
+3. Mixer decodes/resamples/denoises, computes mix-minus per leg
+4. WebRTC leg receives its mix-minus frame, encodes to Opus, and pushes via the WebRTC engine's peer connection sender
+
+Browser→SIP works symmetrically: `ontrack.onReceiveRtp` → WebRTC leg's outbound mpsc → mixer → other legs' inbound channels.
+
+## SDP/Record-Route NAT (fixed in Commit 3 of the cleanup PR)
+
+The proxy tracks a `public_ip: Option<String>` on every `LegInfo` (populated from provider-leg construction sites). When `route_passthrough_message` rewrites SDP (`c=` line) or emits a `Record-Route`, it picks `advertise_ip` based on the destination leg's kind:
+
+- `SipProvider` → `other.public_ip.unwrap_or(lan_ip)` (provider reaches us via public IP)
+- `SipDevice` / `WebRtc` / `Tool` / `Media` → `lan_ip` (everything else is LAN or proxy-internal)
+
+This fixed a real NAT-traversal bug where the proxy advertised its RFC1918 LAN IP to the provider in SDP, causing one-way or no audio for device-originated inbound traffic behind NAT.
+
+## Build & development
+
+- **Build:** `pnpm run buildRust` (never `cargo build` directly — tsrust cross-compiles for both `x86_64-unknown-linux-gnu` and `aarch64-unknown-linux-gnu`)
+- **Cross-compile setup:** the aarch64 target requires `gcc-aarch64-linux-gnu` + `libstdc++6-arm64-cross` (Debian/Ubuntu). See `rust/.cargo/config.toml` for the linker wiring. A committed symlink at `rust/.cargo/crosslibs/aarch64/libstdc++.so` → `/usr/aarch64-linux-gnu/lib/libstdc++.so.6` avoids needing the `libstdc++-13-dev-arm64-cross` package.
+- **Bundle web UI:** `pnpm run bundle` (esbuild, output: `dist_ts_web/bundle.js`)
+- **Full build:** `pnpm run build` (= `buildRust && bundle`)
+- **Start server:** `pnpm run start` (runs `tsx ts/sipproxy.ts`)
+
+## Persistent files
+
+- `.nogit/config.json` — app config (providers, devices, routes, voiceboxes, IVR menus)
+- `.nogit/voicemail/{boxId}/` — voicemail WAV files + `messages.json` index
+- `.nogit/prompts/` — cached TTS prompts for IVR menus
--- a/12
+++ b/12
@@ -2,6 +2,18 @@
 ## STAGE 1 // BUILD
 FROM code.foss.global/host.today/ht-docker-node:lts AS build

+# System build tools that the Rust dep tree needs beyond the base image:
+#   - cmake      : used by the `cmake` crate (transitive via ort_sys / a webrtc
+#                  sub-crate) to build a C/C++ library from source when a
+#                  prebuilt-binary download path doesn't apply.
+#   - pkg-config : used by audiopus_sys and other *-sys crates to locate libs
+#                  on the native target (safe no-op if they vendor their own).
+# These are normally pre-installed on dev machines but not in ht-docker-node:lts.
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      cmake \
+      pkg-config \
+ && rm -rf /var/lib/apt/lists/*
+
 # buildx sets TARGETARCH automatically for each platform it's building:
 #   linux/amd64  -> TARGETARCH=amd64
 #   linux/arm64  -> TARGETARCH=arm64
--- a/changelog.md
+++ b/changelog.md
@@ -1,5 +1,54 @@
 # Changelog

+## 2026-04-12 - 1.22.0 - feat(proxy-engine)
+add on-demand TTS caching for voicemail and IVR prompts
+
+- Route inbound calls directly to configured IVR menus and track them with a dedicated IVR call state
+- Generate voicemail greetings and IVR menu prompts inside the Rust proxy engine on demand instead of precomputing prompts in TypeScript
+- Add cacheable TTS output with sidecar metadata and enable Kokoro CMUdict support for improved prompt generation
+- Extend proxy configuration to include voiceboxes and IVR menus, and update documentation to reflect Kokoro-only prompt generation
+
+## 2026-04-11 - 1.21.0 - feat(providers)
+replace provider creation modal with a guided multi-step setup flow
+
+- Adds a stepper-based provider creation flow with provider type selection, connection, credentials, advanced settings, and review steps.
+- Applies built-in templates for Sipgate and O2/Alice from the selected provider type instead of separate add actions.
+- Adds a final review step with generated provider ID preview and duplicate ID collision handling before saving.
+
+## 2026-04-11 - 1.20.5 - fix(readme)
+improve architecture and call flow documentation with Mermaid diagrams
+
+- Replace ASCII architecture and audio pipeline diagrams with Mermaid diagrams for better readability
+- Document the WebRTC browser call setup sequence, including offer handling and session-to-call linking
+
+## 2026-04-11 - 1.20.4 - fix(deps)
+bump @design.estate/dees-catalog to ^3.71.1
+
+- Updates the @design.estate/dees-catalog dependency from ^3.70.0 to ^3.71.1 in package.json.
+
+## 2026-04-11 - 1.20.3 - fix(ts-config,proxybridge,voicebox)
+align voicebox config types and add missing proxy bridge command definitions
+
+- Reuses the canonical IVoiceboxConfig type from voicebox.ts in config.ts to eliminate duplicated type definitions and optionality mismatches.
+- Makes voicemail timing and limits optional in voicebox config so defaults can be applied consistently during initialization.
+- Adds VoiceboxManager.addMessage and updates recording handling to use it directly for persisted voicemail metadata.
+- Extends proxy bridge command typings with add_leg, remove_leg, and WebRTC signaling commands, and tightens sendCommand typing.
+
+## 2026-04-11 - 1.20.2 - fix(proxy-engine)
+fix inbound route browser ringing and provider-facing SDP advertisement while preventing RTP port exhaustion
+
+- Honor inbound routing `ringBrowsers` when emitting incoming call events so browser toast notifications can be suppressed per route.
+- Rewrite SDP and Record-Route using the destination leg's routable address, using `public_ip` for provider legs and LAN IP for device and internal legs.
+- Store provider leg public IP metadata on legs to support correct per-destination SIP message rewriting.
+- Change the RTP port pool to track sockets with `Weak<UdpSocket>` so ports are reclaimed automatically after calls end, avoiding leaked allocations and eventual 503 failures on new calls.
+- Remove unused dashboard/status, DTMF, relay, and transport helper code paths as part of engine cleanup. 
+
+## 2026-04-11 - 1.20.1 - fix(docker)
+install required native build tools for Rust dependencies in the build image
+
+- Add cmake and pkg-config to the Docker build stage so Rust native dependencies can compile successfully in the container
+- Document why these tools are needed for transitive Rust crates that build or detect native libraries
+
 ## 2026-04-11 - 1.20.0 - feat(docker)
 add multi-arch Docker build and tagged release pipeline

--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
  "name": "siprouter",
-  "version": "1.20.0",
+  "version": "1.22.0",
  "private": true,
  "type": "module",
  "scripts": {
@@ -13,7 +13,7 @@
    "restartBackground": "pnpm run buildRust && pnpm run bundle; test -f .server.pid && kill $(cat .server.pid) 2>/dev/null; sleep 1; rm -f sip_trace.log proxy.out && nohup tsx ts/sipproxy.ts > proxy.out 2>&1 & echo $! > .server.pid; sleep 2; cat proxy.out"
  },
  "dependencies": {
-    "@design.estate/dees-catalog": "^3.70.0",
+    "@design.estate/dees-catalog": "^3.77.0",
    "@design.estate/dees-element": "^2.2.4",
    "@push.rocks/smartrust": "^1.3.2",
    "@push.rocks/smartstate": "^2.3.0",
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -9,8 +9,8 @@ importers:
  .:
    dependencies:
      '@design.estate/dees-catalog':
-        specifier: ^3.70.0
-        version: 3.70.0(@tiptap/pm@2.27.2)
+        specifier: ^3.77.0
+        version: 3.77.0(@tiptap/pm@2.27.2)
      '@design.estate/dees-element':
        specifier: ^2.2.4
        version: 2.2.4
@@ -81,8 +81,8 @@ packages:
  '@configvault.io/interfaces@1.0.17':
    resolution: {integrity: sha512-bEcCUR2VBDJsTin8HQh8Uw/mlYl2v8A3jMIaQ+MTB9Hrqd6CZL2dL7iJdWyFl/3EIX+LDxWFR+Oq7liIq7w+1Q==}

-  '@design.estate/dees-catalog@3.70.0':
-    resolution: {integrity: sha512-bNqOxxl83FNCCV+7QoUj6oeRC0VTExWOClrLrHNMoLIU0TCtzhcmQqiuJhdWrcCwZ5RBhXHGMSFsR62d2RcWpw==}
+  '@design.estate/dees-catalog@3.77.0':
+    resolution: {integrity: sha512-2IfvH390WXCF733XcmEcUP9skqogTz9xlqQw5PUJZy0u2Hf6+hJTyQOi4mcKmhpTE/kCpaD51uw21Lr4ncW6cg==}

  '@design.estate/dees-comms@1.0.30':
    resolution: {integrity: sha512-KchMlklJfKAjQiJiR0xmofXtQ27VgZtBIxcMwPE9d+h3jJRv+lPZxzBQVOM0eyM0uS44S5vJMZ11IeV4uDXSHg==}
@@ -93,8 +93,8 @@ packages:
  '@design.estate/dees-element@2.2.4':
    resolution: {integrity: sha512-O9cA6flBMMd+pBwMQrZXwAWel9yVxgokolb+Em6gvkXxPJ0P/B5UDn4Vc2d4ts3ta55PTBm+l2dPeDVGx/bl7Q==}

-  '@design.estate/dees-wcctools@3.8.0':
-    resolution: {integrity: sha512-CC14iVKUrguzD9jIrdPBd9fZ4egVJEZMxl5y8iy0l7WLumeoYvGsoXj5INVkRPLRVLqziIdi4Je1hXqHt2NU+g==}
+  '@design.estate/dees-wcctools@3.8.4':
+    resolution: {integrity: sha512-KpFK/azK+a/Xpq33pXKcho+tdFKVHhKZM5ArvHqo9QMwTczgp5DZZgowTDUuqAofjZwnuVfCPHK/Pw9e64N46A==}

  '@emnapi/core@1.9.2':
    resolution: {integrity: sha512-UC+ZhH3XtczQYfOlu3lNEkdW/p4dsJ1r/bP7H8+rhao3TTTMO1ATq/4DdIi23XuGoFY+Cz0JmCbdVl0hz9jZcA==}
@@ -1566,8 +1566,8 @@ packages:
  humanize-ms@1.2.1:
    resolution: {integrity: sha1-xG4xWaKT9riW2ikxbYtv6Lt5u+0=}

-  ibantools@4.5.2:
-    resolution: {integrity: sha512-is+8TgZcKS/AMv/z9nW1zz0bhjhoyjpA1p0nc3A6GkW/InOdcQiUZpkufADzh/aO/LY/TOD/P3oPWncNRn5QMA==}
+  ibantools@4.5.4:
+    resolution: {integrity: sha512-6jX1gh4aH6XH+o0ey+wtkMTzkcvsEta7DakIOZSng9voZYpMw3U+gK1+tZChk3aRcPcloEt0NOzksjaRZiqXbw==}

  iconv-lite@0.4.24:
    resolution: {integrity: sha512-v3MXnZAcvnywkTUEZomIActle7RXXeedOR31wwl7VlyoXO4Qi9arvSenNQWne1TcRwhCL1HwLI21bEqdpj8/rA==}
@@ -1694,8 +1694,8 @@ packages:
    resolution: {integrity: sha512-JvNw9Y81y33E+BEYPr0U7omo+U9AySnsMsEiXgwT6yqd31VQWTLNQqmT4ou5eqPFUrTfIDFta2wKhB1hyohtAQ==}
    engines: {node: 20 || >=22}

-  lucide@0.577.0:
-    resolution: {integrity: sha512-PpC/m5eOItp/WU/GlQPFBXDOhq6HibL73KzYP37OX3LM7VmzWQF8voEj8QRWUFvy9FIKfeDQkWYoyS1D/MdWFA==}
+  lucide@1.8.0:
+    resolution: {integrity: sha512-JjV/QnadgFLj1Pyu9IKl0lknrolFEzo04B64QcYLLeRzZl/iEHpdbSrRRKbyXcv45SZNv+WGjIUCT33e7xHO6Q==}

  make-dir@3.1.0:
    resolution: {integrity: sha512-g3FeP20LNwhALb/6Cz6Dd4F2ngze0jz7tbzrD2wAV+o9FeNHe4rL+yK2md0J/fiSf1sa1ADhXqi5+oVwOM/eGw==}
@@ -2462,7 +2462,7 @@ snapshots:
      '@api.global/typedrequest-interfaces': 3.0.19
      '@api.global/typedsocket': 4.1.2(@push.rocks/smartserve@2.0.3)
      '@cloudflare/workers-types': 4.20260409.1
-      '@design.estate/dees-catalog': 3.70.0(@tiptap/pm@2.27.2)
+      '@design.estate/dees-catalog': 3.77.0(@tiptap/pm@2.27.2)
      '@design.estate/dees-comms': 1.0.30
      '@push.rocks/lik': 6.4.0
      '@push.rocks/smartdelay': 3.0.5
@@ -2529,11 +2529,11 @@ snapshots:
    dependencies:
      '@api.global/typedrequest-interfaces': 3.0.19

-  '@design.estate/dees-catalog@3.70.0(@tiptap/pm@2.27.2)':
+  '@design.estate/dees-catalog@3.77.0(@tiptap/pm@2.27.2)':
    dependencies:
      '@design.estate/dees-domtools': 2.5.4
      '@design.estate/dees-element': 2.2.4
-      '@design.estate/dees-wcctools': 3.8.0
+      '@design.estate/dees-wcctools': 3.8.4
      '@fortawesome/fontawesome-svg-core': 7.2.0
      '@fortawesome/free-brands-svg-icons': 7.2.0
      '@fortawesome/free-regular-svg-icons': 7.2.0
@@ -2551,9 +2551,9 @@ snapshots:
      '@tsclass/tsclass': 9.5.0
      echarts: 5.6.0
      highlight.js: 11.11.1
-      ibantools: 4.5.2
+      ibantools: 4.5.4
      lightweight-charts: 5.1.0
-      lucide: 0.577.0
+      lucide: 1.8.0
      monaco-editor: 0.55.1
      pdfjs-dist: 4.10.38
      xterm: 5.3.0
@@ -2610,7 +2610,7 @@ snapshots:
      - supports-color
      - vue

-  '@design.estate/dees-wcctools@3.8.0':
+  '@design.estate/dees-wcctools@3.8.4':
    dependencies:
      '@design.estate/dees-domtools': 2.5.4
      '@design.estate/dees-element': 2.2.4
@@ -4369,7 +4369,7 @@ snapshots:
    dependencies:
      ms: 2.1.3

-  ibantools@4.5.2: {}
+  ibantools@4.5.4: {}

  iconv-lite@0.4.24:
    dependencies:
@@ -4487,7 +4487,7 @@ snapshots:

  lru-cache@11.3.3: {}

-  lucide@0.577.0: {}
+  lucide@1.8.0: {}

  make-dir@3.1.0:
    dependencies:
--- a/readme.md
+++ b/readme.md
@@ -20,7 +20,7 @@ siprouter sits between your SIP trunk providers and your endpoints — hardware
 - 🎯 **Adaptive Jitter Buffer** — Per-leg jitter buffering with sequence-based reordering, adaptive depth (60–120ms), Opus PLC for lost packets, and hold/resume detection
 - 📧 **Voicemail** — Configurable voicemail boxes with TTS greetings, recording, and web playback
 - 🔢 **IVR Menus** — DTMF-navigable interactive voice response with nested menus, routing actions, and custom prompts
- 🗣️ **Neural TTS** — Kokoro-powered announcements and greetings with 25+ voice presets, backed by espeak-ng fallback
+- 🗣️ **Neural TTS** — Kokoro-powered greetings and IVR prompts with 25+ voice presets
 - 🎙️ **Call Recording** — Per-source separated WAV recording at 48kHz via tool legs
 - 🖥️ **Web Dashboard** — Real-time SPA with 9 views: live calls, browser phone, routing, voicemail, IVR, contacts, providers, and streaming logs

@@ -28,39 +28,26 @@ siprouter sits between your SIP trunk providers and your endpoints — hardware

 ## 🏗️ Architecture

-```
-┌─────────────────────────────────────┐
-│           Browser Softphone          │
-│     (WebRTC via WebSocket signaling) │
-└──────────────┬──────────────────────┘
-               │ Opus/WebRTC
-               ▼
-┌──────────────────────────────────────┐
-│            siprouter                  │
-│                                      │
-│  TypeScript Control Plane            │
-│  ┌────────────────────────────────┐  │
-│  │ Config · WebRTC Signaling      │  │
-│  │ REST API · Web Dashboard       │  │
-│  │ Voicebox Manager · TTS Cache   │  │
-│  └────────────┬───────────────────┘  │
-│          JSON-over-stdio IPC         │
-│  ┌────────────┴───────────────────┐  │
-│  │ Rust proxy-engine (data plane) │  │
-│  │                                │  │
-│  │ SIP Stack · Dialog SM · Auth   │  │
-│  │ Call Manager · N-Leg Mixer     │  │
-│  │ 48kHz f32 Bus · Jitter Buffer  │  │
-│  │ Codec Engine · RTP Port Pool   │  │
-│  │ WebRTC Engine · Kokoro TTS     │  │
-│  │ Voicemail · IVR · Recording    │  │
-│  └────┬──────────────────┬────────┘  │
-└───────┤──────────────────┤───────────┘
-        │                  │
- ┌──────┴──────┐    ┌──────┴──────┐
- │ SIP Devices │    │ SIP Trunk   │
- │ (HT801 etc) │    │ Providers   │
- └─────────────┘    └─────────────┘
+```mermaid
+flowchart TB
+    Browser["🌐 Browser Softphone<br/>(WebRTC via WebSocket signaling)"]
+    Devices["📞 SIP Devices<br/>(HT801, desk phones, ATAs)"]
+    Trunks["☎️ SIP Trunk Providers<br/>(sipgate, easybell, …)"]
+
+    subgraph Router["siprouter"]
+        direction TB
+        subgraph TS["TypeScript Control Plane"]
+            TSBits["Config · WebRTC Signaling<br/>REST API · Web Dashboard<br/>Voicebox Manager · TTS Cache"]
+        end
+        subgraph Rust["Rust proxy-engine (data plane)"]
+            RustBits["SIP Stack · Dialog SM · Auth<br/>Call Manager · N-Leg Mixer<br/>48kHz f32 Bus · Jitter Buffer<br/>Codec Engine · RTP Port Pool<br/>WebRTC Engine · Kokoro TTS<br/>Voicemail · IVR · Recording"]
+        end
+        TS <-->|"JSON-over-stdio IPC"| Rust
+    end
+
+    Browser <-->|"Opus / WebRTC"| TS
+    Rust <-->|"SIP / RTP"| Devices
+    Rust <-->|"SIP / RTP"| Trunks
 ```

 ### 🧠 Key Design Decisions
@@ -71,6 +58,37 @@ siprouter sits between your SIP trunk providers and your endpoints — hardware
 - **Per-Session Codec Isolation** — Each call leg gets its own encoder/decoder/resampler/denoiser state — no cross-call corruption.
 - **SDP Codec Negotiation** — Outbound encoding uses the codec actually negotiated in SDP answers, not just the first offered codec.

+### 📲 WebRTC Browser Call Flow
+
+Browser calls are set up in a strict three-step dance — the WebRTC leg cannot be attached at call-creation time because the browser's session ID is only known once the SDP offer arrives:
+
+```mermaid
+sequenceDiagram
+    participant B as Browser
+    participant TS as TypeScript (sipproxy.ts)
+    participant R as Rust proxy-engine
+    participant P as SIP Provider
+
+    B->>TS: POST /api/call
+    TS->>R: make_call (pending call, no WebRTC leg yet)
+    R-->>TS: call_created
+    TS-->>B: webrtc-incoming (callId)
+
+    B->>TS: webrtc-offer (sessionId, SDP)
+    TS->>R: handle_webrtc_offer
+    R-->>TS: webrtc-answer (SDP)
+    TS-->>B: webrtc-answer
+    Note over R: Standalone WebRTC session<br/>(not yet attached to call)
+
+    B->>TS: webrtc_link (callId + sessionId)
+    TS->>R: link session → call
+    R->>R: wire WebRTC leg through mixer
+    R->>P: SIP INVITE
+    P-->>R: 200 OK + SDP
+    R-->>TS: call_answered
+    Note over B,P: Bidirectional Opus ↔ codec-transcoded<br/>audio flows through the mixer
+```
+
 ---

 ## 🚀 Getting Started
@@ -80,7 +98,6 @@ siprouter sits between your SIP trunk providers and your endpoints — hardware
 - **Node.js** ≥ 20 with `tsx` globally available
 - **pnpm** for package management
 - **Rust** toolchain (for building the proxy engine)
- **espeak-ng** (optional, for TTS fallback)

 ### Install & Build

@@ -172,7 +189,7 @@ Create `.nogit/config.json`:

 ### TTS Setup (Optional)

-For neural announcements and voicemail greetings, download the Kokoro TTS model:
+For neural voicemail greetings and IVR prompts, download the Kokoro TTS model:

 ```bash
 mkdir -p .nogit/tts
@@ -182,7 +199,7 @@ curl -L -o .nogit/tts/voices.bin \
  https://github.com/mzdk100/kokoro/releases/download/V1.0/voices.bin
 ```

-Without the model files, TTS falls back to `espeak-ng`. Without either, announcements are skipped — everything else works fine.
+Without the model files, TTS prompts (IVR menus, voicemail greetings) are skipped — everything else works fine.

 ### Run

@@ -209,7 +226,6 @@ siprouter/
 │   ├── frontend.ts                # Web dashboard HTTP/WS server + REST API
 │   ├── webrtcbridge.ts            # WebRTC signaling layer
 │   ├── registrar.ts               # Browser softphone registration
-│   ├── announcement.ts            # TTS announcement generator (espeak-ng / Kokoro)
 │   ├── voicebox.ts                # Voicemail box management
 │   └── call/
 │       └── prompt-cache.ts        # Named audio prompt WAV management
@@ -246,9 +262,17 @@ The `proxy-engine` binary handles all real-time audio processing with a **48kHz

 ### Audio Pipeline

-```
-Inbound:   Wire RTP → Jitter Buffer → Decode → Resample to 48kHz → Denoise (RNNoise) → Mix Bus
-Outbound:  Mix Bus → Mix-Minus → Resample to codec rate → Encode → Wire RTP
+```mermaid
+flowchart LR
+    subgraph Inbound["Inbound path (per leg)"]
+        direction LR
+        IN_RTP["Wire RTP"] --> IN_JB["Jitter Buffer"] --> IN_DEC["Decode"] --> IN_RS["Resample → 48 kHz"] --> IN_DN["Denoise (RNNoise)"] --> IN_BUS["Mix Bus"]
+    end
+
+    subgraph Outbound["Outbound path (per leg)"]
+        direction LR
+        OUT_BUS["Mix Bus"] --> OUT_MM["Mix-Minus"] --> OUT_RS["Resample → codec rate"] --> OUT_ENC["Encode"] --> OUT_RTP["Wire RTP"]
+    end
 ```

 - **Adaptive jitter buffer** — per-leg `BTreeMap`-based buffer keyed by RTP sequence number. Delivers exactly one frame per 20ms mixer tick in sequence order. Adaptive target depth starts at 3 frames (60ms) and adjusts between 2–6 frames based on observed network jitter. Handles hold/resume by detecting large forward sequence jumps and resetting cleanly.
@@ -262,13 +286,12 @@ Outbound:  Mix Bus → Mix-Minus → Resample to codec rate → Encode → Wire

 ## 🗣️ Neural TTS

-Announcements and voicemail greetings are synthesized using [Kokoro TTS](https://github.com/mzdk100/kokoro) — an 82M parameter neural model running via ONNX Runtime directly in the Rust process:
+Voicemail greetings and IVR prompts are synthesized using [Kokoro TTS](https://github.com/mzdk100/kokoro) — an 82M parameter neural model running via ONNX Runtime directly in the Rust process:

 - **24 kHz, 16-bit mono** output
 - **25+ voice presets** — American/British, male/female (e.g., `af_bella`, `am_adam`, `bf_emma`, `bm_george`)
 - **~800ms** synthesis time for a 3-second phrase
 - Lazy-loaded on first use — no startup cost if TTS is unused
- Falls back to `espeak-ng` if the ONNX model is not available

 ---

--- a/rust/.cargo/config.toml
+++ b/rust/.cargo/config.toml
@@ -0,0 +1,30 @@
+# Cross-compile configuration for the proxy-engine crate.
+#
+# tsrust builds for both x86_64-unknown-linux-gnu and aarch64-unknown-linux-gnu
+# from an x86_64 host. Without this config, cargo invokes the host `cc` to
+# link aarch64 objects and fails with
+#   rust-lld: error: <obj.o> is incompatible with elf64-x86-64
+#
+# Required Debian/Ubuntu packages for the aarch64 target to work:
+#   sudo apt install gcc-aarch64-linux-gnu g++-aarch64-linux-gnu \
+#                    libc6-dev-arm64-cross libstdc++6-arm64-cross
+#
+# The `libstdc++.so` dev symlink (needed by the -lstdc++ flag that the
+# kokoro-tts/ort build scripts emit) is provided by this repo at
+# ./crosslibs/aarch64/libstdc++.so, pointing at the versioned shared
+# library installed by `libstdc++6-arm64-cross`. This avoids requiring
+# the `libstdc++-13-dev-arm64-cross` package, which is not always
+# installed alongside the runtime.
+
+[target.aarch64-unknown-linux-gnu]
+linker = "aarch64-linux-gnu-gcc"
+rustflags = ["-C", "link-arg=-L.cargo/crosslibs/aarch64"]
+
+# Tell cc-rs-based build scripts (ring, zstd-sys, audiopus_sys, ort-sys) to
+# use the aarch64 cross toolchain when compiling C sources for the aarch64
+# target. Without these, they'd default to the host `cc` and produce x86_64
+# objects that the aarch64 linker then rejects.
+[env]
+CC_aarch64_unknown_linux_gnu = "aarch64-linux-gnu-gcc"
+CXX_aarch64_unknown_linux_gnu = "aarch64-linux-gnu-g++"
+AR_aarch64_unknown_linux_gnu = "aarch64-linux-gnu-ar"
--- a/rust/.cargo/crosslibs/aarch64/libstdc++.so
+++ b/rust/.cargo/crosslibs/aarch64/libstdc++.so
@@ -0,0 +1 @@
+/usr/aarch64-linux-gnu/lib/libstdc++.so.6
--- a/rust/Cargo.lock
+++ b/rust/Cargo.lock
@@ -532,6 +532,15 @@ dependencies = [
 "cc",
 ]

+[[package]]
+name = "cmudict-fast"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2c9f73004e928ed46c3e7fd7406d2b12c8674153295f08af084b49860276dc02"
+dependencies = [
+ "thiserror",
+]
+
 [[package]]
 name = "codec-lib"
 version = "0.1.0"
@@ -1730,6 +1739,7 @@ dependencies = [
 "bincode 2.0.1",
 "cc",
 "chinese-number",
+ "cmudict-fast",
 "futures",
 "jieba-rs",
 "log",
--- a/rust/crates/proxy-engine/Cargo.toml
+++ b/rust/crates/proxy-engine/Cargo.toml
@@ -19,7 +19,7 @@ regex-lite = "0.1"
 webrtc = "0.8"
 rand = "0.8"
 hound = "3.5"
-kokoro-tts = { version = "0.3", default-features = false }
+kokoro-tts = { version = "0.3", default-features = false, features = ["use-cmudict"] }
 ort = { version = "=2.0.0-rc.11", default-features = false, features = [
    "std", "download-binaries", "copy-dylibs", "ndarray",
    "tls-native-vendored"
--- a/rust/crates/proxy-engine/src/call.rs
+++ b/rust/crates/proxy-engine/src/call.rs
@@ -23,16 +23,22 @@ pub enum CallState {
    Ringing,
    Connected,
    Voicemail,
+    Ivr,
    Terminated,
 }

 impl CallState {
+    /// Wire-format string for events/dashboards. Not currently emitted —
+    /// call state changes flow as typed events (`call_answered`, etc.) —
+    /// but kept for future status-snapshot work.
+    #[allow(dead_code)]
    pub fn as_str(&self) -> &'static str {
        match self {
            Self::SettingUp => "setting-up",
            Self::Ringing => "ringing",
            Self::Connected => "connected",
            Self::Voicemail => "voicemail",
+            Self::Ivr => "ivr",
            Self::Terminated => "terminated",
        }
    }
@@ -45,6 +51,8 @@ pub enum CallDirection {
 }

 impl CallDirection {
+    /// Wire-format string. See CallState::as_str.
+    #[allow(dead_code)]
    pub fn as_str(&self) -> &'static str {
        match self {
            Self::Inbound => "inbound",
@@ -59,7 +67,12 @@ pub enum LegKind {
    SipProvider,
    SipDevice,
    WebRtc,
-    Media, // voicemail playback, IVR, recording
+    /// Voicemail playback, IVR prompt playback, recording — not yet wired up
+    /// as a distinct leg kind (those paths currently use the mixer's role
+    /// system instead). Kept behind allow so adding a real media leg later
+    /// doesn't require re-introducing the variant.
+    #[allow(dead_code)]
+    Media,
    Tool, // observer leg for recording, transcription, etc.
 }

@@ -107,11 +120,22 @@ pub struct LegInfo {
    /// For SIP legs: the SIP Call-ID for message routing.
    pub sip_call_id: Option<String>,
    /// For WebRTC legs: the session ID in WebRtcEngine.
+    ///
+    /// Populated at leg creation but not yet consumed by the hub —
+    /// WebRTC session lookup currently goes through the session registry
+    /// directly. Kept for introspection/debugging.
+    #[allow(dead_code)]
    pub webrtc_session_id: Option<String>,
    /// The RTP socket allocated for this leg.
    pub rtp_socket: Option<Arc<UdpSocket>>,
    /// The RTP port number.
    pub rtp_port: u16,
+    /// Public IP to advertise in SDP/Record-Route when THIS leg is the
+    /// destination of a rewrite. Populated only for provider legs; `None`
+    /// for LAN SIP devices, WebRTC browsers, media, and tool legs (which
+    /// are reachable via `lan_ip`). See `route_passthrough_message` for
+    /// the per-destination advertise-IP logic.
+    pub public_ip: Option<String>,
    /// The remote media endpoint (learned from SDP or address learning).
    pub remote_media: Option<SocketAddr>,
    /// SIP signaling address (provider or device).
@@ -124,14 +148,21 @@ pub struct LegInfo {

 /// A multiparty call with N legs and a central mixer.
 pub struct Call {
+    // Duplicated from the HashMap key in CallManager. Kept for future
+    // status-snapshot work.
+    #[allow(dead_code)]
    pub id: String,
    pub state: CallState,
+    // Populated at call creation but not currently consumed — dashboard
+    // pull snapshots are gone (push events only).
+    #[allow(dead_code)]
    pub direction: CallDirection,
    pub created_at: Instant,

    // Metadata.
    pub caller_number: Option<String>,
    pub callee_number: Option<String>,
+    #[allow(dead_code)]
    pub provider_id: String,

    /// Original INVITE from the device (for device-originated outbound calls).
@@ -211,42 +242,4 @@ impl Call {
            handle.abort();
        }
    }
-
-    /// Produce a JSON status snapshot for the dashboard.
-    pub fn to_status_json(&self) -> serde_json::Value {
-        let legs: Vec<serde_json::Value> = self
-            .legs
-            .values()
-            .filter(|l| l.state != LegState::Terminated)
-            .map(|l| {
-                let metadata: serde_json::Value = if l.metadata.is_empty() {
-                    serde_json::json!({})
-                } else {
-                    serde_json::Value::Object(
-                        l.metadata.iter().map(|(k, v)| (k.clone(), v.clone())).collect(),
-                    )
-                };
-                serde_json::json!({
-                    "id": l.id,
-                    "type": l.kind.as_str(),
-                    "state": l.state.as_str(),
-                    "codec": sip_proto::helpers::codec_name(l.codec_pt),
-                    "rtpPort": l.rtp_port,
-                    "remoteMedia": l.remote_media.map(|a| format!("{}:{}", a.ip(), a.port())),
-                    "metadata": metadata,
-                })
-            })
-            .collect();
-
-        serde_json::json!({
-            "id": self.id,
-            "state": self.state.as_str(),
-            "direction": self.direction.as_str(),
-            "callerNumber": self.caller_number,
-            "calleeNumber": self.callee_number,
-            "providerUsed": self.provider_id,
-            "duration": self.duration_secs(),
-            "legs": legs,
-        })
-    }
 }
--- a/rust/crates/proxy-engine/src/call_manager.rs
+++ b/rust/crates/proxy-engine/src/call_manager.rs
@@ -12,13 +12,24 @@ use crate::mixer::spawn_mixer;
 use crate::registrar::Registrar;
 use crate::rtp::RtpPortPool;
 use crate::sip_leg::{SipLeg, SipLegAction, SipLegConfig};
+use crate::tts::TtsEngine;
 use sip_proto::helpers::{build_sdp, generate_call_id, generate_tag, parse_sdp_endpoint, SdpOptions};
 use sip_proto::message::{ResponseOptions, SipMessage};
 use sip_proto::rewrite::{rewrite_sdp, rewrite_sip_uri};
 use std::collections::HashMap;
 use std::net::SocketAddr;
+use std::path::Path;
 use std::sync::Arc;
 use tokio::net::UdpSocket;
+use tokio::sync::Mutex;
+
+/// Result of creating an inbound call — carries both the call id and
+/// whether browsers should be notified (flows from the matched inbound
+/// route's `ring_browsers` flag, or the fallback default).
+pub struct InboundCallCreated {
+    pub call_id: String,
+    pub ring_browsers: bool,
+}

 /// Emit a `leg_added` event with full leg information.
 /// Free function (not a method) to avoid `&self` borrow conflicts when `self.calls` is borrowed.
@@ -94,26 +105,6 @@ impl CallManager {
        self.sip_index.contains_key(sip_call_id)
    }

-    /// Get an RTP socket for a call's provider leg (used by webrtc_link).
-    pub fn get_call_provider_rtp_socket(&self, call_id: &str) -> Option<Arc<UdpSocket>> {
-        let call = self.calls.get(call_id)?;
-        for leg in call.legs.values() {
-            if leg.kind == LegKind::SipProvider {
-                return leg.rtp_socket.clone();
-            }
-        }
-        None
-    }
-
-    /// Get all active call statuses for the dashboard.
-    pub fn get_all_statuses(&self) -> Vec<serde_json::Value> {
-        self.calls
-            .values()
-            .filter(|c| c.state != CallState::Terminated)
-            .map(|c| c.to_status_json())
-            .collect()
-    }
-
    // -----------------------------------------------------------------------
    // SIP message routing
    // -----------------------------------------------------------------------
@@ -426,8 +417,8 @@ impl CallManager {

        // Find the counterpart leg.
        let other_leg = call.legs.values().find(|l| l.id != this_leg_id && l.state != LegState::Terminated);
-        let (other_addr, other_rtp_port, other_leg_id) = match other_leg {
-            Some(l) => (l.signaling_addr, l.rtp_port, l.id.clone()),
+        let (other_addr, other_rtp_port, other_leg_id, other_kind, other_public_ip) = match other_leg {
+            Some(l) => (l.signaling_addr, l.rtp_port, l.id.clone(), l.kind, l.public_ip.clone()),
            None => return false,
        };
        let forward_to = match other_addr {
@@ -438,8 +429,14 @@ impl CallManager {
        let lan_ip = config.proxy.lan_ip.clone();
        let lan_port = config.proxy.lan_port;

-        // Get this leg's RTP port (for SDP rewriting — tell the other side to send RTP here).
-        let this_rtp_port = call.legs.get(this_leg_id).map(|l| l.rtp_port).unwrap_or(0);
+        // Pick the IP to advertise to the destination leg. Provider legs face
+        // the public internet and need `public_ip`; every other leg kind is
+        // on-LAN (or proxy-internal) and takes `lan_ip`. This rule is applied
+        // both to the SDP `c=` line and the Record-Route header below.
+        let advertise_ip: String = match other_kind {
+            LegKind::SipProvider => other_public_ip.unwrap_or_else(|| lan_ip.clone()),
+            _ => lan_ip.clone(),
+        };

        // Check if the other leg is a B2BUA leg (has SipLeg for proper dialog mgmt).
        let other_has_sip_leg = call.legs.get(&other_leg_id)
@@ -533,10 +530,11 @@ impl CallManager {

            // Forward other requests with SDP rewriting.
            let mut fwd = msg.clone();
-            // Rewrite SDP to point the other side to this leg's RTP port
-            // (so we receive their audio on our socket).
+            // Rewrite SDP so the destination leg sends RTP to our proxy port
+            // at an address that is routable from its vantage point
+            // (public IP for provider legs, LAN IP for everything else).
            if fwd.has_sdp_body() {
-                let (new_body, _) = rewrite_sdp(&fwd.body, &lan_ip, other_rtp_port);
+                let (new_body, _) = rewrite_sdp(&fwd.body, &advertise_ip, other_rtp_port);
                fwd.body = new_body;
                fwd.update_content_length();
            }
@@ -548,7 +546,8 @@ impl CallManager {
                }
            }
            if fwd.is_dialog_establishing() {
-                fwd.prepend_header("Record-Route", &format!("<sip:{lan_ip}:{lan_port};lr>"));
+                // Record-Route must also be routable from the destination leg.
+                fwd.prepend_header("Record-Route", &format!("<sip:{advertise_ip}:{lan_port};lr>"));
            }
            let _ = socket.send_to(&fwd.serialize(), forward_to).await;
            return true;
@@ -560,15 +559,10 @@ impl CallManager {
            let cseq_method = msg.cseq_method().unwrap_or("").to_uppercase();

            let mut fwd = msg.clone();
-            // Rewrite SDP so the forward-to side sends RTP to the correct leg port.
+            // Rewrite SDP so the forward-to side sends RTP to the correct
+            // leg port at a routable address (see `advertise_ip` above).
            if fwd.has_sdp_body() {
-                let rewrite_ip = if this_kind == LegKind::SipDevice {
-                    // Response from device → send to provider: use LAN/public IP.
-                    &lan_ip
-                } else {
-                    &lan_ip
-                };
-                let (new_body, _) = rewrite_sdp(&fwd.body, rewrite_ip, other_rtp_port);
+                let (new_body, _) = rewrite_sdp(&fwd.body, &advertise_ip, other_rtp_port);
                fwd.body = new_body;
                fwd.update_content_length();
            }
@@ -690,7 +684,8 @@ impl CallManager {
        rtp_pool: &mut RtpPortPool,
        socket: &UdpSocket,
        public_ip: Option<&str>,
-    ) -> Option<String> {
+        tts_engine: Arc<Mutex<TtsEngine>>,
+    ) -> Option<InboundCallCreated> {
        let call_id = self.next_call_id();
        let lan_ip = &config.proxy.lan_ip;
        let lan_port = config.proxy.lan_port;
@@ -707,17 +702,65 @@ impl CallManager {
            .unwrap_or("")
            .to_string();

-        // Resolve target device.
-        let device_addr = match self.resolve_first_device(config, registrar) {
+        // Resolve via the configured inbound routing table. This honors
+        // user-defined routes from the UI (numberPattern, callerPattern,
+        // sourceProvider, targets, ringBrowsers). If no route matches, the
+        // fallback returns an empty `device_ids` and `ring_browsers: true`,
+        // which preserves pre-routing behavior via the `resolve_first_device`
+        // fallback below.
+        //
+        // TODO: Multi-target inbound fork is not yet implemented.
+        //   - `route.device_ids` beyond the first registered target are ignored.
+        //   - `ring_browsers` is informational only — browsers see a toast but
+        //     do not race the SIP device. First-to-answer-wins requires a
+        //     multi-leg fork + per-leg CANCEL, which is not built yet.
+        let route = config.resolve_inbound_route(provider_id, &called_number, &caller_number);
+        let ring_browsers = route.ring_browsers;
+
+        // IVR routing: if the route targets an IVR menu, go there directly.
+        if let Some(ref ivr_menu_id) = route.ivr_menu_id {
+            if let Some(ivr) = &config.ivr {
+                if ivr.enabled {
+                    if let Some(menu) = ivr.menus.iter().find(|m| m.id == *ivr_menu_id) {
+                        let call_id = self
+                            .route_to_ivr(
+                                &call_id, invite, from_addr, &caller_number,
+                                provider_id, provider_config, config, rtp_pool, socket,
+                                public_ip, menu, &tts_engine,
+                            )
+                            .await?;
+                        return Some(InboundCallCreated { call_id, ring_browsers });
+                    }
+                }
+            }
+        }
+
+        // Pick the first registered device from the matched targets, or fall
+        // back to any-registered-device if the route has no resolved targets.
+        let device_addr = route
+            .device_ids
+            .iter()
+            .find_map(|id| registrar.get_device_contact(id))
+            .or_else(|| self.resolve_first_device(config, registrar));
+
+        let device_addr = match device_addr {
            Some(addr) => addr,
            None => {
                // No device registered → voicemail.
-                return self
+                // Resolve greeting WAV on-demand (may trigger TTS generation).
+                let greeting_wav = resolve_greeting_wav(
+                    config,
+                    route.voicemail_box.as_deref(),
+                    &tts_engine,
+                ).await;
+                let call_id = self
                    .route_to_voicemail(
                        &call_id, invite, from_addr, &caller_number,
                        provider_id, provider_config, config, rtp_pool, socket, public_ip,
+                        greeting_wav,
                    )
-                    .await;
+                    .await?;
+                return Some(InboundCallCreated { call_id, ring_browsers });
            }
        };

@@ -781,6 +824,7 @@ impl CallManager {
                webrtc_session_id: None,
                rtp_socket: Some(provider_rtp.socket.clone()),
                rtp_port: provider_rtp.port,
+                public_ip: public_ip.map(|s| s.to_string()),
                remote_media: provider_media,
                signaling_addr: Some(from_addr),
                metadata: HashMap::new(),
@@ -801,6 +845,7 @@ impl CallManager {
                webrtc_session_id: None,
                rtp_socket: Some(device_rtp.socket.clone()),
                rtp_port: device_rtp.port,
+                public_ip: None,
                remote_media: None, // Learned from device's 200 OK.
                signaling_addr: Some(device_addr),
                metadata: HashMap::new(),
@@ -844,7 +889,7 @@ impl CallManager {
            }
        }

-        Some(call_id)
+        Some(InboundCallCreated { call_id, ring_browsers })
    }

    /// Initiate an outbound B2BUA call from the dashboard.
@@ -920,6 +965,7 @@ impl CallManager {
                webrtc_session_id: None,
                rtp_socket: Some(rtp_alloc.socket.clone()),
                rtp_port: rtp_alloc.port,
+                public_ip: public_ip.map(|s| s.to_string()),
                remote_media: None,
                signaling_addr: Some(provider_dest),
                metadata: HashMap::new(),
@@ -1030,6 +1076,7 @@ impl CallManager {
                sip_leg: None,
                sip_call_id: Some(device_sip_call_id.clone()),
                webrtc_session_id: None,
+                public_ip: None,
                rtp_socket: Some(device_rtp.socket.clone()),
                rtp_port: device_rtp.port,
                remote_media: device_media,
@@ -1076,6 +1123,7 @@ impl CallManager {
                webrtc_session_id: None,
                rtp_socket: Some(provider_rtp.socket.clone()),
                rtp_port: provider_rtp.port,
+                public_ip: public_ip.map(|s| s.to_string()),
                remote_media: None,
                signaling_addr: Some(provider_dest),
                metadata: HashMap::new(),
@@ -1114,7 +1162,7 @@ impl CallManager {
        public_ip: Option<&str>,
        registered_aor: &str,
    ) -> Option<String> {
-        let call = self.calls.get(call_id)?;
+        self.calls.get(call_id)?; // existence check; the call is re-fetched via get_mut below
        let lan_ip = &config.proxy.lan_ip;
        let lan_port = config.proxy.lan_port;

@@ -1151,6 +1199,7 @@ impl CallManager {
            webrtc_session_id: None,
            rtp_socket: Some(rtp_alloc.socket.clone()),
            rtp_port: rtp_alloc.port,
+            public_ip: public_ip.map(|s| s.to_string()),
            remote_media: None,
            signaling_addr: Some(provider_dest),
            metadata: HashMap::new(),
@@ -1182,7 +1231,7 @@ impl CallManager {
        socket: &UdpSocket,
    ) -> Option<String> {
        let device_addr = registrar.get_device_contact(device_id)?;
-        let call = self.calls.get(call_id)?;
+        self.calls.get(call_id)?; // existence check; the call is re-fetched via get_mut below
        let lan_ip = &config.proxy.lan_ip;
        let lan_port = config.proxy.lan_port;

@@ -1221,6 +1270,7 @@ impl CallManager {
            webrtc_session_id: None,
            rtp_socket: Some(rtp_alloc.socket.clone()),
            rtp_port: rtp_alloc.port,
+            public_ip: None,
            remote_media: None,
            signaling_addr: Some(device_addr),
            metadata: HashMap::new(),
@@ -1514,6 +1564,7 @@ impl CallManager {
        rtp_pool: &mut RtpPortPool,
        socket: &UdpSocket,
        public_ip: Option<&str>,
+        greeting_wav: Option<String>,
    ) -> Option<String> {
        let lan_ip = &config.proxy.lan_ip;
        let pub_ip = public_ip.unwrap_or(lan_ip.as_str());
@@ -1581,6 +1632,7 @@ impl CallManager {
                webrtc_session_id: None,
                rtp_socket: Some(rtp_alloc.socket.clone()),
                rtp_port: rtp_alloc.port,
+                public_ip: public_ip.map(|s| s.to_string()),
                remote_media: Some(provider_media),
                signaling_addr: Some(from_addr),
                metadata: HashMap::new(),
@@ -1607,8 +1659,6 @@ impl CallManager {
            .as_millis();
        let recording_dir = "nogit/voicemail/default".to_string();
        let recording_path = format!("{recording_dir}/msg-{timestamp}.wav");
-        let greeting_wav = find_greeting_wav();
-
        let out_tx = self.out_tx.clone();
        let call_id_owned = call_id.to_string();
        let caller_owned = caller_number.to_string();
@@ -1625,6 +1675,211 @@ impl CallManager {
        Some(call_id.to_string())
    }

+    // -----------------------------------------------------------------------
+    // IVR routing
+    // -----------------------------------------------------------------------
+
+    #[allow(clippy::too_many_arguments)]
+    async fn route_to_ivr(
+        &mut self,
+        call_id: &str,
+        invite: &SipMessage,
+        from_addr: SocketAddr,
+        caller_number: &str,
+        provider_id: &str,
+        provider_config: &ProviderConfig,
+        config: &AppConfig,
+        rtp_pool: &mut RtpPortPool,
+        socket: &UdpSocket,
+        public_ip: Option<&str>,
+        menu: &crate::config::IvrMenuConfig,
+        tts_engine: &Arc<Mutex<TtsEngine>>,
+    ) -> Option<String> {
+        let lan_ip = &config.proxy.lan_ip;
+
+        let rtp_alloc = match rtp_pool.allocate().await {
+            Some(a) => a,
+            None => {
+                let resp = SipMessage::create_response(503, "Service Unavailable", invite, None);
+                let _ = socket.send_to(&resp.serialize(), from_addr).await;
+                return None;
+            }
+        };
+
+        let codec_pt = provider_config.codecs.first().copied().unwrap_or(9);
+        let pub_ip = public_ip.unwrap_or(lan_ip.as_str());
+
+        let sdp = sip_proto::helpers::build_sdp(&sip_proto::helpers::SdpOptions {
+            ip: pub_ip,
+            port: rtp_alloc.port,
+            payload_types: &provider_config.codecs,
+            ..Default::default()
+        });
+
+        let response = SipMessage::create_response(
+            200, "OK", invite,
+            Some(sip_proto::message::ResponseOptions {
+                to_tag: Some(sip_proto::helpers::generate_tag()),
+                contact: Some(format!("<sip:{}:{}>", lan_ip, config.proxy.lan_port)),
+                body: Some(sdp),
+                content_type: Some("application/sdp".to_string()),
+                ..Default::default()
+            }),
+        );
+        let _ = socket.send_to(&response.serialize(), from_addr).await;
+
+        let provider_media = if invite.has_sdp_body() {
+            parse_sdp_endpoint(&invite.body)
+                .and_then(|ep| format!("{}:{}", ep.address, ep.port).parse().ok())
+        } else {
+            Some(from_addr)
+        };
+        let provider_media = provider_media.unwrap_or(from_addr);
+
+        // Create call with IVR state.
+        let (mixer_cmd_tx, mixer_task) = spawn_mixer(call_id.to_string(), self.out_tx.clone());
+        let mut call = Call::new(
+            call_id.to_string(),
+            CallDirection::Inbound,
+            provider_id.to_string(),
+            mixer_cmd_tx.clone(),
+            mixer_task,
+        );
+        call.state = CallState::Ivr;
+        call.caller_number = Some(caller_number.to_string());
+
+        let provider_leg_id = format!("{call_id}-prov");
+        call.legs.insert(
+            provider_leg_id.clone(),
+            LegInfo {
+                id: provider_leg_id.clone(),
+                kind: LegKind::SipProvider,
+                state: LegState::Connected,
+                codec_pt,
+                sip_leg: None,
+                sip_call_id: Some(invite.call_id().to_string()),
+                webrtc_session_id: None,
+                rtp_socket: Some(rtp_alloc.socket.clone()),
+                rtp_port: rtp_alloc.port,
+                public_ip: public_ip.map(|s| s.to_string()),
+                remote_media: Some(provider_media),
+                signaling_addr: Some(from_addr),
+                metadata: HashMap::new(),
+            },
+        );
+
+        self.sip_index.insert(
+            invite.call_id().to_string(),
+            (call_id.to_string(), provider_leg_id.clone()),
+        );
+        self.calls.insert(call_id.to_string(), call);
+
+        // Emit leg_added for the provider leg.
+        if let Some(call) = self.calls.get(call_id) {
+            for leg in call.legs.values() {
+                emit_leg_added_event(&self.out_tx, call_id, leg);
+            }
+        }
+
+        // Generate IVR prompt on-demand via TTS (cached).
+        let voice = menu.prompt_voice.as_deref().unwrap_or("af_bella");
+        let prompt_output = format!(".nogit/tts/ivr-menu-{}.wav", menu.id);
+        let prompt_params = serde_json::json!({
+            "model": ".nogit/tts/kokoro-v1.0.onnx",
+            "voices": ".nogit/tts/voices.bin",
+            "voice": voice,
+            "text": &menu.prompt_text,
+            "output": &prompt_output,
+            "cacheable": true,
+        });
+
+        let prompt_wav = {
+            let mut tts = tts_engine.lock().await;
+            match tts.generate(&prompt_params).await {
+                Ok(_) => Some(prompt_output),
+                Err(e) => {
+                    eprintln!("[ivr] TTS generation failed: {e}");
+                    None
+                }
+            }
+        };
+
+        // Load prompt and run interaction via the mixer.
+        let out_tx = self.out_tx.clone();
+        let call_id_owned = call_id.to_string();
+        let expected_digits: Vec<char> = menu
+            .entries
+            .iter()
+            .filter_map(|e| e.digit.chars().next())
+            .collect();
+        let timeout_ms = menu.timeout_sec.unwrap_or(5) * 1000;
+
+        tokio::spawn(async move {
+            // Load prompt PCM frames if available.
+            let prompt_frames = prompt_wav.as_ref().and_then(|wav| {
+                crate::audio_player::load_prompt_pcm_frames(wav).ok()
+            });
+
+            if let Some(frames) = prompt_frames {
+                let (result_tx, result_rx) = tokio::sync::oneshot::channel();
+                let _ = mixer_cmd_tx
+                    .send(crate::mixer::MixerCommand::StartInteraction {
+                        leg_id: provider_leg_id.clone(),
+                        prompt_pcm_frames: frames,
+                        expected_digits: expected_digits.clone(),
+                        timeout_ms,
+                        result_tx,
+                    })
+                    .await;
+
+                // Wait for digit or timeout.
+                let safety = tokio::time::Duration::from_millis(timeout_ms as u64 + 30000);
+                let result = match tokio::time::timeout(safety, result_rx).await {
+                    Ok(Ok(r)) => r,
+                    Ok(Err(_)) => crate::mixer::InteractionResult::Cancelled,
+                    Err(_) => crate::mixer::InteractionResult::Timeout,
+                };
+
+                match &result {
+                    crate::mixer::InteractionResult::Digit(d) => {
+                        eprintln!("[ivr] caller pressed '{d}' on call {call_id_owned}");
+                        emit_event(
+                            &out_tx,
+                            "ivr_digit",
+                            serde_json::json!({
+                                "call_id": call_id_owned,
+                                "digit": d.to_string(),
+                            }),
+                        );
+                    }
+                    crate::mixer::InteractionResult::Timeout => {
+                        eprintln!("[ivr] timeout on call {call_id_owned}");
+                        emit_event(
+                            &out_tx,
+                            "ivr_timeout",
+                            serde_json::json!({ "call_id": call_id_owned }),
+                        );
+                    }
+                    crate::mixer::InteractionResult::Cancelled => {
+                        eprintln!("[ivr] cancelled on call {call_id_owned}");
+                    }
+                }
+            } else {
+                eprintln!("[ivr] no prompt available for call {call_id_owned}, ending");
+                emit_event(
+                    &out_tx,
+                    "ivr_error",
+                    serde_json::json!({
+                        "call_id": call_id_owned,
+                        "error": "no prompt available",
+                    }),
+                );
+            }
+        });
+
+        Some(call_id.to_string())
+    }
+
    // -----------------------------------------------------------------------
    // Internal helpers
    // -----------------------------------------------------------------------
@@ -1639,13 +1894,56 @@ impl CallManager {
    }
 }

-fn find_greeting_wav() -> Option<String> {
-    let candidates = [
+/// Resolve the greeting WAV for a voicemail box.
+///
+/// Priority:
+/// 1. Pre-recorded WAV from voicebox config (`greetingWavPath`)
+/// 2. On-demand TTS generation from greeting text (cached via `cacheable: true`)
+/// 3. Legacy hardcoded paths (`.nogit/voicemail/default/greeting.wav`, etc.)
+/// 4. None — voicemail session plays beep only
+async fn resolve_greeting_wav(
+    config: &AppConfig,
+    voicebox_id: Option<&str>,
+    tts_engine: &Arc<Mutex<TtsEngine>>,
+) -> Option<String> {
+    // 1. Look up voicebox config.
+    let vb = voicebox_id
+        .and_then(|id| config.voiceboxes.iter().find(|v| v.id == id && v.enabled));
+
+    if let Some(vb) = vb {
+        // 2. Pre-recorded WAV takes priority.
+        if let Some(ref wav) = vb.greeting_wav_path {
+            if Path::new(wav).exists() {
+                return Some(wav.clone());
+            }
+        }
+        // 3. TTS on-demand with caching.
+        let text = vb.greeting_text.as_deref().unwrap_or(
+            "The person you are trying to reach is not available. Please leave a message after the tone.",
+        );
+        let voice = vb.greeting_voice.as_deref().unwrap_or("af_bella");
+        let output = format!(".nogit/tts/voicemail-greeting-{}.wav", vb.id);
+
+        let params = serde_json::json!({
+            "model": ".nogit/tts/kokoro-v1.0.onnx",
+            "voices": ".nogit/tts/voices.bin",
+            "voice": voice,
+            "text": text,
+            "output": &output,
+            "cacheable": true,
+        });
+        let mut tts = tts_engine.lock().await;
+        if tts.generate(&params).await.is_ok() {
+            return Some(output);
+        }
+    }
+
+    // 4. Fallback: legacy hardcoded paths.
+    for path in &[
        ".nogit/voicemail/default/greeting.wav",
        ".nogit/voicemail/greeting.wav",
-    ];
-    for path in &candidates {
-        if std::path::Path::new(path).exists() {
+    ] {
+        if Path::new(path).exists() {
            return Some(path.to_string());
        }
    }
--- a/rust/crates/proxy-engine/src/config.rs
+++ b/rust/crates/proxy-engine/src/config.rs
@@ -30,6 +30,11 @@ impl Endpoint {
 }

 /// Provider quirks for codec/protocol workarounds.
+//
+// Deserialized from provider config for TS parity. Early-media silence
+// injection and related workarounds are not yet ported to the Rust engine,
+// so every field is populated by serde but not yet consumed.
+#[allow(dead_code)]
 #[derive(Debug, Clone, Deserialize)]
 pub struct Quirks {
    #[serde(rename = "earlyMediaSilence")]
@@ -44,6 +49,9 @@ pub struct Quirks {
 #[derive(Debug, Clone, Deserialize)]
 pub struct ProviderConfig {
    pub id: String,
+    // UI label — populated by serde for parity with the TS config, not
+    // consumed at runtime.
+    #[allow(dead_code)]
    #[serde(rename = "displayName")]
    pub display_name: String,
    pub domain: String,
@@ -54,6 +62,8 @@ pub struct ProviderConfig {
    #[serde(rename = "registerIntervalSec")]
    pub register_interval_sec: u32,
    pub codecs: Vec<u8>,
+    // Workaround knobs populated by serde but not yet acted upon — see Quirks.
+    #[allow(dead_code)]
    pub quirks: Quirks,
 }

@@ -84,6 +94,10 @@ pub struct RouteMatch {

 /// Route action.
 #[derive(Debug, Clone, Deserialize)]
+// Several fields (voicemail_box, ivr_menu_id, no_answer_timeout) are read
+// by resolve_inbound_route but not yet honored downstream — see the
+// multi-target TODO in CallManager::create_inbound_call.
+#[allow(dead_code)]
 pub struct RouteAction {
    pub targets: Option<Vec<String>>,
    #[serde(rename = "ringBrowsers")]
@@ -106,7 +120,11 @@ pub struct RouteAction {
 /// A routing rule.
 #[derive(Debug, Clone, Deserialize)]
 pub struct Route {
+    // `id` and `name` are UI identifiers, populated by serde but not
+    // consumed by the resolvers.
+    #[allow(dead_code)]
    pub id: String,
+    #[allow(dead_code)]
    pub name: String,
    pub priority: i32,
    pub enabled: bool,
@@ -141,6 +159,10 @@ pub struct AppConfig {
    pub providers: Vec<ProviderConfig>,
    pub devices: Vec<DeviceConfig>,
    pub routing: RoutingConfig,
+    #[serde(default)]
+    pub voiceboxes: Vec<VoiceboxConfig>,
+    #[serde(default)]
+    pub ivr: Option<IvrConfig>,
 }

 #[derive(Debug, Clone, Deserialize)]
@@ -148,6 +170,59 @@ pub struct RoutingConfig {
    pub routes: Vec<Route>,
 }

+// ---------------------------------------------------------------------------
+// Voicebox config
+// ---------------------------------------------------------------------------
+
+#[allow(dead_code)]
+#[derive(Debug, Clone, Deserialize)]
+pub struct VoiceboxConfig {
+    pub id: String,
+    #[serde(default)]
+    pub enabled: bool,
+    #[serde(rename = "greetingText")]
+    pub greeting_text: Option<String>,
+    #[serde(rename = "greetingVoice")]
+    pub greeting_voice: Option<String>,
+    #[serde(rename = "greetingWavPath")]
+    pub greeting_wav_path: Option<String>,
+    #[serde(rename = "maxRecordingSec")]
+    pub max_recording_sec: Option<u32>,
+}
+
+// ---------------------------------------------------------------------------
+// IVR config
+// ---------------------------------------------------------------------------
+
+#[allow(dead_code)]
+#[derive(Debug, Clone, Deserialize)]
+pub struct IvrConfig {
+    pub enabled: bool,
+    pub menus: Vec<IvrMenuConfig>,
+    #[serde(rename = "entryMenuId")]
+    pub entry_menu_id: String,
+}
+
+#[derive(Debug, Clone, Deserialize)]
+pub struct IvrMenuConfig {
+    pub id: String,
+    #[serde(rename = "promptText")]
+    pub prompt_text: String,
+    #[serde(rename = "promptVoice")]
+    pub prompt_voice: Option<String>,
+    pub entries: Vec<IvrMenuEntry>,
+    #[serde(rename = "timeoutSec")]
+    pub timeout_sec: Option<u32>,
+}
+
+#[allow(dead_code)]
+#[derive(Debug, Clone, Deserialize)]
+pub struct IvrMenuEntry {
+    pub digit: String,
+    pub action: String,
+    pub target: Option<String>,
+}
+
 // ---------------------------------------------------------------------------
 // Pattern matching (ported from ts/config.ts)
 // ---------------------------------------------------------------------------
@@ -192,10 +267,18 @@ pub fn matches_pattern(pattern: Option<&str>, value: &str) -> bool {
 /// Result of resolving an outbound route.
 pub struct OutboundRouteResult {
    pub provider: ProviderConfig,
+    // TODO: prefix rewriting is unfinished — this is computed but the
+    // caller ignores it and uses the raw dialed number.
+    #[allow(dead_code)]
    pub transformed_number: String,
 }

 /// Result of resolving an inbound route.
+//
+// `device_ids` and `ring_browsers` are consumed by create_inbound_call.
+// The remaining fields (voicemail_box, ivr_menu_id, no_answer_timeout)
+// are resolved but not yet acted upon — see the multi-target TODO.
+#[allow(dead_code)]
 pub struct InboundRouteResult {
    pub device_ids: Vec<String>,
    pub ring_browsers: bool,
--- a/rust/crates/proxy-engine/src/dtmf.rs
+++ b/rust/crates/proxy-engine/src/dtmf.rs
@@ -1,200 +0,0 @@
-//! DTMF detection — parses RFC 2833 telephone-event RTP packets.
-//!
-//! Deduplicates repeated packets (same digit sent multiple times with
-//! increasing duration) and fires once per detected digit.
-//!
-//! Ported from ts/call/dtmf-detector.ts.
-
-use crate::ipc::{emit_event, OutTx};
-
-/// RFC 2833 event ID → character mapping.
-const EVENT_CHARS: &[char] = &[
-    '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '*', '#', 'A', 'B', 'C', 'D',
-];
-
-/// Safety timeout: report digit if no End packet arrives within this many ms.
-const SAFETY_TIMEOUT_MS: u64 = 200;
-
-/// DTMF detector for a single RTP stream.
-pub struct DtmfDetector {
-    /// Negotiated telephone-event payload type (default 101).
-    telephone_event_pt: u8,
-    /// Clock rate for duration calculation (default 8000 Hz).
-    clock_rate: u32,
-    /// Call ID for event emission.
-    call_id: String,
-
-    // Deduplication state.
-    current_event_id: Option<u8>,
-    current_event_ts: Option<u32>,
-    current_event_reported: bool,
-    current_event_duration: u16,
-
-    out_tx: OutTx,
-}
-
-impl DtmfDetector {
-    pub fn new(call_id: String, out_tx: OutTx) -> Self {
-        Self {
-            telephone_event_pt: 101,
-            clock_rate: 8000,
-            call_id,
-            current_event_id: None,
-            current_event_ts: None,
-            current_event_reported: false,
-            current_event_duration: 0,
-            out_tx,
-        }
-    }
-
-    /// Feed an RTP packet. Checks PT; ignores non-DTMF packets.
-    /// Returns Some(digit_char) if a digit was detected.
-    pub fn process_rtp(&mut self, data: &[u8]) -> Option<char> {
-        if data.len() < 16 {
-            return None; // 12-byte header + 4-byte telephone-event minimum
-        }
-
-        let pt = data[1] & 0x7F;
-        if pt != self.telephone_event_pt {
-            return None;
-        }
-
-        let marker = (data[1] & 0x80) != 0;
-        let rtp_timestamp = u32::from_be_bytes([data[4], data[5], data[6], data[7]]);
-
-        // Parse telephone-event payload.
-        let event_id = data[12];
-        let end_bit = (data[13] & 0x80) != 0;
-        let duration = u16::from_be_bytes([data[14], data[15]]);
-
-        if event_id as usize >= EVENT_CHARS.len() {
-            return None;
-        }
-
-        // Detect new event.
-        let is_new = marker
-            || self.current_event_id != Some(event_id)
-            || self.current_event_ts != Some(rtp_timestamp);
-
-        if is_new {
-            // Report pending unreported event.
-            let pending = self.report_pending();
-
-            self.current_event_id = Some(event_id);
-            self.current_event_ts = Some(rtp_timestamp);
-            self.current_event_reported = false;
-            self.current_event_duration = duration;
-
-            if pending.is_some() {
-                return pending;
-            }
-        }
-
-        if duration > self.current_event_duration {
-            self.current_event_duration = duration;
-        }
-
-        // Report on End bit (first time only).
-        if end_bit && !self.current_event_reported {
-            self.current_event_reported = true;
-            let digit = EVENT_CHARS[event_id as usize];
-            let duration_ms = (self.current_event_duration as f64 / self.clock_rate as f64) * 1000.0;
-
-            emit_event(
-                &self.out_tx,
-                "dtmf_digit",
-                serde_json::json!({
-                    "call_id": self.call_id,
-                    "digit": digit.to_string(),
-                    "duration_ms": duration_ms.round() as u32,
-                    "source": "rfc2833",
-                }),
-            );
-
-            return Some(digit);
-        }
-
-        None
-    }
-
-    /// Report a pending unreported event.
-    fn report_pending(&mut self) -> Option<char> {
-        if let Some(event_id) = self.current_event_id {
-            if !self.current_event_reported && (event_id as usize) < EVENT_CHARS.len() {
-                self.current_event_reported = true;
-                let digit = EVENT_CHARS[event_id as usize];
-                let duration_ms =
-                    (self.current_event_duration as f64 / self.clock_rate as f64) * 1000.0;
-
-                emit_event(
-                    &self.out_tx,
-                    "dtmf_digit",
-                    serde_json::json!({
-                        "call_id": self.call_id,
-                        "digit": digit.to_string(),
-                        "duration_ms": duration_ms.round() as u32,
-                        "source": "rfc2833",
-                    }),
-                );
-
-                return Some(digit);
-            }
-        }
-        None
-    }
-
-    /// Process a SIP INFO message body for DTMF.
-    pub fn process_sip_info(&mut self, content_type: &str, body: &str) -> Option<char> {
-        let ct = content_type.to_ascii_lowercase();
-
-        if ct.contains("application/dtmf-relay") {
-            // Format: "Signal= 5\r\nDuration= 160\r\n"
-            let signal = body
-                .lines()
-                .find(|l| l.to_ascii_lowercase().starts_with("signal"))
-                .and_then(|l| l.split('=').nth(1))
-                .map(|s| s.trim().to_string())?;
-
-            if signal.len() != 1 {
-                return None;
-            }
-            let digit = signal.chars().next()?.to_ascii_uppercase();
-            if !"0123456789*#ABCD".contains(digit) {
-                return None;
-            }
-
-            emit_event(
-                &self.out_tx,
-                "dtmf_digit",
-                serde_json::json!({
-                    "call_id": self.call_id,
-                    "digit": digit.to_string(),
-                    "source": "sip-info",
-                }),
-            );
-
-            return Some(digit);
-        }
-
-        if ct.contains("application/dtmf") {
-            let digit = body.trim().chars().next()?.to_ascii_uppercase();
-            if !"0123456789*#ABCD".contains(digit) {
-                return None;
-            }
-
-            emit_event(
-                &self.out_tx,
-                "dtmf_digit",
-                serde_json::json!({
-                    "call_id": self.call_id,
-                    "digit": digit.to_string(),
-                    "source": "sip-info",
-                }),
-            );
-
-            return Some(digit);
-        }
-
-        None
-    }
-}
--- a/rust/crates/proxy-engine/src/main.rs
+++ b/rust/crates/proxy-engine/src/main.rs
@@ -10,7 +10,6 @@ mod audio_player;
 mod call;
 mod call_manager;
 mod config;
-mod dtmf;
 mod ipc;
 mod jitter_buffer;
 mod leg_io;
@@ -51,11 +50,12 @@ struct ProxyEngine {
    registrar: Registrar,
    call_mgr: CallManager,
    rtp_pool: Option<RtpPortPool>,
+    tts_engine: Arc<Mutex<tts::TtsEngine>>,
    out_tx: OutTx,
 }

 impl ProxyEngine {
-    fn new(out_tx: OutTx) -> Self {
+    fn new(out_tx: OutTx, tts_engine: Arc<Mutex<tts::TtsEngine>>) -> Self {
        Self {
            config: None,
            transport: None,
@@ -63,6 +63,7 @@ impl ProxyEngine {
            registrar: Registrar::new(out_tx.clone()),
            call_mgr: CallManager::new(out_tx.clone()),
            rtp_pool: None,
+            tts_engine,
            out_tx,
        }
    }
@@ -89,15 +90,16 @@ async fn main() {
    // Emit ready event.
    emit_event(&out_tx, "ready", serde_json::json!({}));

-    // Shared engine state (SIP side).
-    let engine = Arc::new(Mutex::new(ProxyEngine::new(out_tx.clone())));
+    // TTS engine — separate internal lock, lazy-loads model on first use.
+    let tts_engine = Arc::new(Mutex::new(tts::TtsEngine::new()));
+
+    // Shared engine state (SIP side). TTS engine is stored inside so the
+    // SIP packet handler path can reach it for on-demand voicemail/IVR generation.
+    let engine = Arc::new(Mutex::new(ProxyEngine::new(out_tx.clone(), tts_engine)));

    // WebRTC engine — separate lock to avoid deadlock with SIP handlers.
    let webrtc = Arc::new(Mutex::new(WebRtcEngine::new(out_tx.clone())));

-    // TTS engine — separate lock, lazy-loads model on first use.
-    let tts_engine = Arc::new(Mutex::new(tts::TtsEngine::new()));
-
    // Read commands from stdin.
    let stdin = tokio::io::stdin();
    let reader = BufReader::new(stdin);
@@ -118,12 +120,11 @@ async fn main() {

        let engine = engine.clone();
        let webrtc = webrtc.clone();
-        let tts_engine = tts_engine.clone();
        let out_tx = out_tx.clone();

        // Handle commands — some are async, so we spawn.
        tokio::spawn(async move {
-            handle_command(engine, webrtc, tts_engine, &out_tx, cmd).await;
+            handle_command(engine, webrtc, &out_tx, cmd).await;
        });
    }
 }
@@ -131,7 +132,6 @@ async fn main() {
 async fn handle_command(
    engine: Arc<Mutex<ProxyEngine>>,
    webrtc: Arc<Mutex<WebRtcEngine>>,
-    tts_engine: Arc<Mutex<tts::TtsEngine>>,
    out_tx: &OutTx,
    cmd: Command,
 ) {
@@ -140,7 +140,6 @@ async fn handle_command(
        "configure" => handle_configure(engine, out_tx, &cmd).await,
        "hangup" => handle_hangup(engine, out_tx, &cmd).await,
        "make_call" => handle_make_call(engine, out_tx, &cmd).await,
-        "get_status" => handle_get_status(engine, out_tx, &cmd).await,
        "add_leg" => handle_add_leg(engine, out_tx, &cmd).await,
        "remove_leg" => handle_remove_leg(engine, out_tx, &cmd).await,
        // WebRTC commands — lock webrtc only (no engine contention).
@@ -157,8 +156,8 @@ async fn handle_command(
        "add_tool_leg" => handle_add_tool_leg(engine, out_tx, &cmd).await,
        "remove_tool_leg" => handle_remove_tool_leg(engine, out_tx, &cmd).await,
        "set_leg_metadata" => handle_set_leg_metadata(engine, out_tx, &cmd).await,
-        // TTS command — lock tts_engine only (no SIP/WebRTC contention).
-        "generate_tts" => handle_generate_tts(tts_engine, out_tx, &cmd).await,
+        // TTS command — gets tts_engine from inside ProxyEngine.
+        "generate_tts" => handle_generate_tts(engine, out_tx, &cmd).await,
        _ => respond_err(out_tx, &cmd.id, &format!("unknown command: {}", cmd.method)),
    }
 }
@@ -327,10 +326,12 @@ async fn handle_sip_packet(
            ref registrar,
            ref mut call_mgr,
            ref mut rtp_pool,
+            ref tts_engine,
            ..
        } = *eng;
+        let tts_clone = tts_engine.clone();
        let rtp_pool = rtp_pool.as_mut().unwrap();
-        let call_id = call_mgr
+        let inbound = call_mgr
            .create_inbound_call(
                &msg,
                from_addr,
@@ -341,10 +342,11 @@ async fn handle_sip_packet(
                rtp_pool,
                socket,
                public_ip.as_deref(),
+                tts_clone,
            )
            .await;

-        if let Some(call_id) = call_id {
+        if let Some(inbound) = inbound {
            // Emit event so TypeScript knows about the call (for dashboard, IVR routing, etc).
            let from_header = msg.get_header("From").unwrap_or("");
            let from_uri = SipMessage::extract_uri(from_header).unwrap_or("Unknown");
@@ -357,10 +359,11 @@ async fn handle_sip_packet(
                &eng.out_tx,
                "incoming_call",
                serde_json::json!({
-                    "call_id": call_id,
+                    "call_id": inbound.call_id,
                    "from_uri": from_uri,
                    "to_number": called_number,
                    "provider_id": provider_id,
+                    "ring_browsers": inbound.ring_browsers,
                }),
            );
        }
@@ -383,7 +386,7 @@ async fn handle_sip_packet(
        let route_result = config_ref.resolve_outbound_route(
            &dialed_number,
            device_id.as_deref(),
-            &|pid: &str| {
+            &|_pid: &str| {
                // Can't call async here — use a sync check.
                // For now, assume all configured providers are available.
                true
@@ -454,13 +457,6 @@ async fn handle_sip_packet(
    );
 }

-/// Handle `get_status` — return active call statuses from Rust.
-async fn handle_get_status(engine: Arc<Mutex<ProxyEngine>>, out_tx: &OutTx, cmd: &Command) {
-    let eng = engine.lock().await;
-    let calls = eng.call_mgr.get_all_statuses();
-    respond_ok(out_tx, &cmd.id, serde_json::json!({ "calls": calls }));
-}
-
 /// Handle `make_call` — initiate an outbound call to a number via a provider.
 async fn handle_make_call(engine: Arc<Mutex<ProxyEngine>>, out_tx: &OutTx, cmd: &Command) {
    let number = match cmd.params.get("number").and_then(|v| v.as_str()) {
@@ -665,6 +661,7 @@ async fn handle_webrtc_link(
                        webrtc_session_id: Some(session_id.clone()),
                        rtp_socket: None,
                        rtp_port: 0,
+                        public_ip: None,
                        remote_media: None,
                        signaling_addr: None,
                        metadata: std::collections::HashMap::new(),
@@ -1116,6 +1113,7 @@ async fn handle_add_tool_leg(
                webrtc_session_id: None,
                rtp_socket: None,
                rtp_port: 0,
+                public_ip: None,
                remote_media: None,
                signaling_addr: None,
                metadata,
@@ -1237,10 +1235,11 @@ async fn handle_set_leg_metadata(

 /// Handle `generate_tts` — synthesize text to a WAV file using Kokoro TTS.
 async fn handle_generate_tts(
-    tts_engine: Arc<Mutex<tts::TtsEngine>>,
+    engine: Arc<Mutex<ProxyEngine>>,
    out_tx: &OutTx,
    cmd: &Command,
 ) {
+    let tts_engine = engine.lock().await.tts_engine.clone();
    let mut tts = tts_engine.lock().await;
    match tts.generate(&cmd.params).await {
        Ok(result) => respond_ok(out_tx, &cmd.id, result),
--- a/rust/crates/proxy-engine/src/mixer.rs
+++ b/rust/crates/proxy-engine/src/mixer.rs
@@ -39,6 +39,10 @@ pub struct RtpPacket {
    /// RTP sequence number for reordering.
    pub seq: u16,
    /// RTP timestamp from the original packet header.
+    ///
+    /// Set on inbound RTP but not yet consumed downstream — reserved for
+    /// future jitter/sync work in the mixer.
+    #[allow(dead_code)]
    pub timestamp: u32,
 }

@@ -136,8 +140,6 @@ pub enum MixerCommand {
        timeout_ms: u32,
        result_tx: oneshot::Sender<InteractionResult>,
    },
-    /// Cancel an in-progress interaction (e.g., leg being removed).
-    CancelInteraction { leg_id: String },

    /// Add a tool leg that receives per-source unmerged audio.
    AddToolLeg {
@@ -295,16 +297,6 @@ async fn mixer_loop(
                        let _ = result_tx.send(InteractionResult::Cancelled);
                    }
                }
-                Ok(MixerCommand::CancelInteraction { leg_id }) => {
-                    if let Some(slot) = legs.get_mut(&leg_id) {
-                        if let LegRole::Isolated(ref mut state) = slot.role {
-                            if let Some(tx) = state.result_tx.take() {
-                                let _ = tx.send(InteractionResult::Cancelled);
-                            }
-                        }
-                        slot.role = LegRole::Participant;
-                    }
-                }
                Ok(MixerCommand::AddToolLeg {
                    leg_id,
                    tool_type,
--- a/rust/crates/proxy-engine/src/provider.rs
+++ b/rust/crates/proxy-engine/src/provider.rs
@@ -331,17 +331,6 @@ impl ProviderManager {
        }
        None
    }
-
-    /// Check if a provider is currently registered.
-    pub async fn is_registered(&self, provider_id: &str) -> bool {
-        for ps_arc in &self.providers {
-            let ps = ps_arc.lock().await;
-            if ps.config.id == provider_id {
-                return ps.is_registered;
-            }
-        }
-        false
-    }
 }

 /// Registration loop for a single provider.
--- a/rust/crates/proxy-engine/src/recorder.rs
+++ b/rust/crates/proxy-engine/src/recorder.rs
@@ -178,5 +178,8 @@ impl Recorder {
 pub struct RecordingResult {
    pub file_path: String,
    pub duration_ms: u64,
+    // Running-sample total kept for parity with the TS recorder; not yet
+    // surfaced through any event or dashboard field.
+    #[allow(dead_code)]
    pub total_samples: u64,
 }
--- a/rust/crates/proxy-engine/src/registrar.rs
+++ b/rust/crates/proxy-engine/src/registrar.rs
@@ -19,11 +19,19 @@ const MAX_EXPIRES: u32 = 300;
 #[derive(Debug, Clone)]
 pub struct RegisteredDevice {
    pub device_id: String,
+    // These fields are populated at REGISTER time for logging/debugging but are
+    // not read back — device identity flows via the `device_registered` push
+    // event, not via struct queries. Kept behind allow(dead_code) because
+    // removing them would churn handle_register for no runtime benefit.
+    #[allow(dead_code)]
    pub display_name: String,
+    #[allow(dead_code)]
    pub extension: String,
    pub contact_addr: SocketAddr,
+    #[allow(dead_code)]
    pub registered_at: Instant,
    pub expires_at: Instant,
+    #[allow(dead_code)]
    pub aor: String,
 }

@@ -134,11 +142,6 @@ impl Registrar {
        Some(entry.contact_addr)
    }

-    /// Check if a source address belongs to a known device.
-    pub fn is_known_device_address(&self, addr: &str) -> bool {
-        self.devices.iter().any(|d| d.expected_address == addr)
-    }
-
    /// Find a registered device by its source IP address.
    pub fn find_by_address(&self, addr: &SocketAddr) -> Option<&RegisteredDevice> {
        let ip = addr.ip().to_string();
@@ -146,26 +149,4 @@ impl Registrar {
            e.contact_addr.ip().to_string() == ip && Instant::now() <= e.expires_at
        })
    }
-
-    /// Get all device statuses for the dashboard.
-    pub fn get_all_statuses(&self) -> Vec<serde_json::Value> {
-        let now = Instant::now();
-        let mut result = Vec::new();
-
-        for dc in &self.devices {
-            let reg = self.registered.get(&dc.id);
-            let connected = reg.map(|r| now <= r.expires_at).unwrap_or(false);
-            result.push(serde_json::json!({
-                "id": dc.id,
-                "displayName": dc.display_name,
-                "address": reg.filter(|_| connected).map(|r| r.contact_addr.ip().to_string()),
-                "port": reg.filter(|_| connected).map(|r| r.contact_addr.port()),
-                "aor": reg.map(|r| r.aor.as_str()).unwrap_or(""),
-                "connected": connected,
-                "isBrowser": false,
-            }));
-        }
-
-        result
-    }
 }
--- a/rust/crates/proxy-engine/src/rtp.rs
+++ b/rust/crates/proxy-engine/src/rtp.rs
@@ -1,17 +1,19 @@
-//! RTP port pool and media forwarding.
+//! RTP port pool for media sockets.
 //!
-//! Manages a pool of even-numbered UDP ports for RTP media.
-//! Each port gets a bound tokio UdpSocket. Supports:
-//! - Direct forwarding (SIP-to-SIP, no transcoding)
-//! - Transcoding forwarding (via codec-lib, e.g. G.722 ↔ Opus)
-//! - Silence generation
-//! - NAT priming
+//! Manages a pool of even-numbered UDP ports for RTP media. `allocate()`
+//! hands back an `Arc<UdpSocket>` to the caller (stored on the owning
+//! `LegInfo`), while the pool itself keeps only a `Weak<UdpSocket>`. When
+//! the call terminates and `LegInfo` is dropped, the strong refcount
+//! reaches zero, the socket is closed, and `allocate()` prunes the dead
+//! weak ref the next time it scans that slot — so the port automatically
+//! becomes available for reuse without any explicit `release()` plumbing.
 //!
-//! Ported from ts/call/rtp-port-pool.ts + sip-leg.ts RTP handling.
+//! This fixes the previous leak where the pool held `Arc<UdpSocket>` and
+//! `release()` was never called, eventually exhausting the port range and
+//! causing "503 Service Unavailable" on new calls.

 use std::collections::HashMap;
-use std::net::SocketAddr;
-use std::sync::Arc;
+use std::sync::{Arc, Weak};
 use tokio::net::UdpSocket;

 /// A single RTP port allocation.
@@ -24,7 +26,7 @@ pub struct RtpAllocation {
 pub struct RtpPortPool {
    min: u16,
    max: u16,
-    allocated: HashMap<u16, Arc<UdpSocket>>,
+    allocated: HashMap<u16, Weak<UdpSocket>>,
 }

 impl RtpPortPool {
@@ -41,11 +43,19 @@ impl RtpPortPool {
    pub async fn allocate(&mut self) -> Option<RtpAllocation> {
        let mut port = self.min;
        while port < self.max {
+            // Prune a dead weak ref at this slot: if the last strong Arc
+            // (held by the owning LegInfo) was dropped when the call ended,
+            // the socket is already closed and the slot is free again.
+            if let Some(weak) = self.allocated.get(&port) {
+                if weak.strong_count() == 0 {
+                    self.allocated.remove(&port);
+                }
+            }
            if !self.allocated.contains_key(&port) {
                match UdpSocket::bind(format!("0.0.0.0:{port}")).await {
                    Ok(sock) => {
                        let sock = Arc::new(sock);
-                        self.allocated.insert(port, sock.clone());
+                        self.allocated.insert(port, Arc::downgrade(&sock));
                        return Some(RtpAllocation { port, socket: sock });
                    }
                    Err(_) => {
@@ -57,83 +67,6 @@ impl RtpPortPool {
        }
        None // Pool exhausted.
    }
-
-    /// Release a port back to the pool.
-    pub fn release(&mut self, port: u16) {
-        self.allocated.remove(&port);
-        // Socket is dropped when the last Arc reference goes away.
-    }
-
-    pub fn size(&self) -> usize {
-        self.allocated.len()
-    }
-
-    pub fn capacity(&self) -> usize {
-        ((self.max - self.min) / 2) as usize
-    }
-}
-
-/// An active RTP relay between two endpoints.
-/// Receives on `local_socket` and forwards to `remote_addr`.
-pub struct RtpRelay {
-    pub local_port: u16,
-    pub local_socket: Arc<UdpSocket>,
-    pub remote_addr: Option<SocketAddr>,
-    /// If set, transcode packets using this codec session before forwarding.
-    pub transcode: Option<TranscodeConfig>,
-    /// Packets received counter.
-    pub pkt_received: u64,
-    /// Packets sent counter.
-    pub pkt_sent: u64,
-}
-
-pub struct TranscodeConfig {
-    pub from_pt: u8,
-    pub to_pt: u8,
-    pub session_id: String,
-}
-
-impl RtpRelay {
-    pub fn new(port: u16, socket: Arc<UdpSocket>) -> Self {
-        Self {
-            local_port: port,
-            local_socket: socket,
-            remote_addr: None,
-            transcode: None,
-            pkt_received: 0,
-            pkt_sent: 0,
-        }
-    }
-
-    pub fn set_remote(&mut self, addr: SocketAddr) {
-        self.remote_addr = Some(addr);
-    }
-}
-
-/// Send a 1-byte NAT priming packet to open a pinhole.
-pub async fn prime_nat(socket: &UdpSocket, remote: SocketAddr) {
-    let _ = socket.send_to(&[0u8], remote).await;
-}
-
-/// Build an RTP silence frame for PCMU (payload type 0).
-pub fn silence_frame_pcmu() -> Vec<u8> {
-    // 12-byte RTP header + 160 bytes of µ-law silence (0xFF)
-    let mut frame = vec![0u8; 172];
-    frame[0] = 0x80; // V=2
-    frame[1] = 0; // PT=0 (PCMU)
-    // seq, timestamp, ssrc left as 0 — caller should set these
-    frame[12..].fill(0xFF); // µ-law silence
-    frame
-}
-
-/// Build an RTP silence frame for G.722 (payload type 9).
-pub fn silence_frame_g722() -> Vec<u8> {
-    // 12-byte RTP header + 160 bytes of G.722 silence
-    let mut frame = vec![0u8; 172];
-    frame[0] = 0x80; // V=2
-    frame[1] = 9; // PT=9 (G.722)
-    // G.722 silence: all zeros is valid silence
-    frame
 }

 /// Build an RTP header with the given parameters.
--- a/rust/crates/proxy-engine/src/sip_leg.rs
+++ b/rust/crates/proxy-engine/src/sip_leg.rs
@@ -16,7 +16,6 @@ use sip_proto::helpers::{
 };
 use sip_proto::message::{RequestOptions, SipMessage};
 use std::net::SocketAddr;
-use std::sync::Arc;
 use tokio::net::UdpSocket;

 /// State of a SIP leg.
@@ -40,6 +39,9 @@ pub struct SipLegConfig {
    /// SIP target endpoint (provider outbound proxy or device address).
    pub sip_target: SocketAddr,
    /// Provider credentials (for 407 auth).
+    // username is carried for parity with the provider config but digest auth
+    // rebuilds the username from the registered AOR, so this slot is never read.
+    #[allow(dead_code)]
    pub username: Option<String>,
    pub password: Option<String>,
    pub registered_aor: Option<String>,
@@ -51,6 +53,10 @@ pub struct SipLegConfig {

 /// A SIP leg with full dialog management.
 pub struct SipLeg {
+    // Leg identity is tracked via the enclosing LegInfo's key in the call's
+    // leg map; SipLeg itself never reads this field back. Kept to preserve
+    // the (id, config) constructor shape used by the call manager.
+    #[allow(dead_code)]
    pub id: String,
    pub state: LegState,
    pub config: SipLegConfig,
@@ -411,11 +417,6 @@ impl SipLeg {
        dialog.terminate();
        Some(msg.serialize())
    }
-
-    /// Get the SIP Call-ID for routing.
-    pub fn sip_call_id(&self) -> Option<&str> {
-        self.dialog.as_ref().map(|d| d.call_id.as_str())
-    }
 }

 /// Actions produced by the SipLeg message handler.
--- a/rust/crates/proxy-engine/src/sip_transport.rs
+++ b/rust/crates/proxy-engine/src/sip_transport.rs
@@ -27,22 +27,6 @@ impl SipTransport {
        self.socket.clone()
    }

-    /// Send a raw SIP message to a destination.
-    pub async fn send_to(&self, data: &[u8], dest: SocketAddr) -> Result<usize, String> {
-        self.socket
-            .send_to(data, dest)
-            .await
-            .map_err(|e| format!("send to {dest}: {e}"))
-    }
-
-    /// Send a raw SIP message to an address:port pair.
-    pub async fn send_to_addr(&self, data: &[u8], addr: &str, port: u16) -> Result<usize, String> {
-        let dest: SocketAddr = format!("{addr}:{port}")
-            .parse()
-            .map_err(|e| format!("bad address {addr}:{port}: {e}"))?;
-        self.send_to(data, dest).await
-    }
-
    /// Spawn the UDP receive loop. Calls the handler for every received packet.
    pub fn spawn_receiver<F>(
        &self,
--- a/rust/crates/proxy-engine/src/tts.rs
+++ b/rust/crates/proxy-engine/src/tts.rs
@@ -1,8 +1,13 @@
 //! Text-to-speech engine — synthesizes text to WAV files using Kokoro neural TTS.
 //!
 //! The model is loaded lazily on first use. If the model/voices files are not
-//! present, the generate command returns an error and the TS side falls back
-//! to espeak-ng.
+//! present, the generate command returns an error and the caller skips the prompt.
+//!
+//! Caching is handled internally via a `.meta` sidecar file next to each WAV.
+//! When `cacheable` is true, the engine checks whether the existing WAV was
+//! generated from the same text+voice; if so it returns immediately (cache hit).
+//! Callers never need to check for cached files — that is entirely this module's
+//! responsibility.

 use kokoro_tts::{KokoroTts, Voice};
 use std::path::Path;
@@ -32,6 +37,8 @@ impl TtsEngine {
    ///   - `voice`: voice name (e.g. "af_bella")
    ///   - `text`: text to synthesize
    ///   - `output`: output WAV file path
+    ///   - `cacheable`: if true, skip synthesis when the output WAV already
+    ///     matches the same text+voice (checked via a `.meta` sidecar file)
    pub async fn generate(&mut self, params: &serde_json::Value) -> Result<serde_json::Value, String> {
        let model_path = params.get("model").and_then(|v| v.as_str())
            .ok_or("missing 'model' param")?;
@@ -43,11 +50,19 @@ impl TtsEngine {
            .ok_or("missing 'text' param")?;
        let output_path = params.get("output").and_then(|v| v.as_str())
            .ok_or("missing 'output' param")?;
+        let cacheable = params.get("cacheable").and_then(|v| v.as_bool())
+            .unwrap_or(false);

        if text.is_empty() {
            return Err("empty text".into());
        }

+        // Cache check: if cacheable and the sidecar matches, return immediately.
+        if cacheable && self.is_cache_hit(output_path, text, voice_name) {
+            eprintln!("[tts] cache hit: {output_path}");
+            return Ok(serde_json::json!({ "output": output_path }));
+        }
+
        // Check that model/voices files exist.
        if !Path::new(model_path).exists() {
            return Err(format!("model not found: {model_path}"));
@@ -56,6 +71,11 @@ impl TtsEngine {
            return Err(format!("voices not found: {voices_path}"));
        }

+        // Ensure parent directory exists.
+        if let Some(parent) = Path::new(output_path).parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+
        // Lazy-load or reload if paths changed.
        if self.tts.is_none()
            || self.loaded_model_path != model_path
@@ -95,9 +115,41 @@ impl TtsEngine {
        }
        writer.finalize().map_err(|e| format!("WAV finalize: {e}"))?;

+        // Write sidecar for future cache checks.
+        if cacheable {
+            self.write_cache_meta(output_path, text, voice_name);
+        }
+
        eprintln!("[tts] wrote {output_path}");
        Ok(serde_json::json!({ "output": output_path }))
    }
+
+    // -----------------------------------------------------------------------
+    // Cache helpers
+    // -----------------------------------------------------------------------
+
+    /// Check if the WAV + sidecar on disk match the given text+voice.
+    fn is_cache_hit(&self, output_path: &str, text: &str, voice: &str) -> bool {
+        let meta_path = format!("{output_path}.meta");
+        if !Path::new(output_path).exists() || !Path::new(&meta_path).exists() {
+            return false;
+        }
+        match std::fs::read_to_string(&meta_path) {
+            Ok(contents) => contents == Self::cache_key(text, voice),
+            Err(_) => false,
+        }
+    }
+
+    /// Write the sidecar `.meta` file next to the WAV.
+    fn write_cache_meta(&self, output_path: &str, text: &str, voice: &str) {
+        let meta_path = format!("{output_path}.meta");
+        let _ = std::fs::write(&meta_path, Self::cache_key(text, voice));
+    }
+
+    /// Build the cache key from text + voice.
+    fn cache_key(text: &str, voice: &str) -> String {
+        format!("{}\0{}", text, voice)
+    }
 }

 /// Map voice name string to Kokoro Voice enum variant.
--- a/ts/00_commitinfo_data.ts
+++ b/ts/00_commitinfo_data.ts
@@ -3,6 +3,6 @@
 */
 export const commitinfo = {
  name: 'siprouter',
-  version: '1.20.0',
+  version: '1.22.0',
  description: 'undefined'
 }
--- a/ts/announcement.ts
+++ b/ts/announcement.ts
@@ -1,137 +0,0 @@
-/**
- * TTS announcement module — generates announcement WAV files at startup.
- *
- * Engine priority: espeak-ng (formant TTS, fast) → Kokoro neural TTS via
- * proxy-engine → disabled.
- *
- * The generated WAV is left on disk for Rust's audio_player / start_interaction
- * to play during calls. No encoding or RTP playback happens in TypeScript.
- */
-
-import { execSync } from 'node:child_process';
-import fs from 'node:fs';
-import path from 'node:path';
-import { sendProxyCommand, isProxyReady } from './proxybridge.ts';
-
-// ---------------------------------------------------------------------------
-// State
-// ---------------------------------------------------------------------------
-
-const TTS_DIR = path.join(process.cwd(), '.nogit', 'tts');
-const ANNOUNCEMENT_TEXT = "Hello. I'm connecting your call now.";
-const CACHE_WAV = path.join(TTS_DIR, 'announcement.wav');
-
-// Kokoro fallback constants.
-const KOKORO_MODEL = 'kokoro-v1.0.onnx';
-const KOKORO_VOICES = 'voices.bin';
-const KOKORO_VOICE = 'af_bella';
-
-// ---------------------------------------------------------------------------
-// TTS generators
-// ---------------------------------------------------------------------------
-
-/** Check if espeak-ng is available on the system. */
-function isEspeakAvailable(): boolean {
-  try {
-    execSync('which espeak-ng', { stdio: 'pipe' });
-    return true;
-  } catch {
-    return false;
-  }
-}
-
-/** Generate announcement WAV via espeak-ng (primary engine). */
-function generateViaEspeak(wavPath: string, text: string, log: (msg: string) => void): boolean {
-  log('[tts] generating announcement audio via espeak-ng...');
-  try {
-    execSync(
-      `espeak-ng -v en-us -s 150 -w "${wavPath}" "${text}"`,
-      { timeout: 10000, stdio: 'pipe' },
-    );
-    log('[tts] espeak-ng WAV generated');
-    return true;
-  } catch (e: any) {
-    log(`[tts] espeak-ng failed: ${e.message}`);
-    return false;
-  }
-}
-
-/** Generate announcement WAV via Kokoro TTS (fallback, runs inside proxy-engine). */
-async function generateViaKokoro(wavPath: string, text: string, log: (msg: string) => void): Promise<boolean> {
-  const modelPath = path.join(TTS_DIR, KOKORO_MODEL);
-  const voicesPath = path.join(TTS_DIR, KOKORO_VOICES);
-
-  if (!fs.existsSync(modelPath) || !fs.existsSync(voicesPath)) {
-    log('[tts] Kokoro model/voices not found — Kokoro fallback unavailable');
-    return false;
-  }
-
-  if (!isProxyReady()) {
-    log('[tts] proxy-engine not ready — Kokoro fallback unavailable');
-    return false;
-  }
-
-  log('[tts] generating announcement audio via Kokoro TTS (fallback)...');
-  try {
-    await sendProxyCommand('generate_tts', {
-      model: modelPath,
-      voices: voicesPath,
-      voice: KOKORO_VOICE,
-      text,
-      output: wavPath,
-    });
-    log('[tts] Kokoro WAV generated (via proxy-engine)');
-    return true;
-  } catch (e: any) {
-    log(`[tts] Kokoro failed: ${e.message}`);
-    return false;
-  }
-}
-
-// ---------------------------------------------------------------------------
-// Initialization
-// ---------------------------------------------------------------------------
-
-/**
- * Pre-generate the announcement WAV file.
- * Must be called after the proxy engine is initialized.
- *
- * Engine priority: espeak-ng → Kokoro → disabled.
- */
-export async function initAnnouncement(log: (msg: string) => void): Promise<boolean> {
-  fs.mkdirSync(TTS_DIR, { recursive: true });
-
-  try {
-    if (!fs.existsSync(CACHE_WAV)) {
-      let generated = false;
-
-      // Try espeak-ng first.
-      if (isEspeakAvailable()) {
-        generated = generateViaEspeak(CACHE_WAV, ANNOUNCEMENT_TEXT, log);
-      } else {
-        log('[tts] espeak-ng not installed — trying Kokoro fallback');
-      }
-
-      // Fall back to Kokoro (via proxy-engine).
-      if (!generated) {
-        generated = await generateViaKokoro(CACHE_WAV, ANNOUNCEMENT_TEXT, log);
-      }
-
-      if (!generated) {
-        log('[tts] no TTS engine available — announcements disabled');
-        return false;
-      }
-    }
-
-    log('[tts] announcement WAV ready');
-    return true;
-  } catch (e: any) {
-    log(`[tts] init error: ${e.message}`);
-    return false;
-  }
-}
-
-/** Get the path to the cached announcement WAV, or null if not generated. */
-export function getAnnouncementWavPath(): string | null {
-  return fs.existsSync(CACHE_WAV) ? CACHE_WAV : null;
-}
--- a/ts/call/prompt-cache.ts
+++ b/ts/call/prompt-cache.ts
@@ -1,275 +0,0 @@
-/**
- * PromptCache — manages named audio prompt WAV files for IVR and voicemail.
- *
- * Generates WAV files via espeak-ng (primary) or Kokoro TTS through the
- * proxy-engine (fallback). Also supports loading pre-existing WAV files
- * and programmatic tone generation.
- *
- * All audio playback happens in Rust (audio_player / start_interaction).
- * This module only manages WAV files on disk.
- */
-
-import { execSync } from 'node:child_process';
-import fs from 'node:fs';
-import path from 'node:path';
-import { Buffer } from 'node:buffer';
-import { sendProxyCommand, isProxyReady } from '../proxybridge.ts';
-
-// ---------------------------------------------------------------------------
-// Types
-// ---------------------------------------------------------------------------
-
-/** A cached prompt — just a WAV file path and metadata. */
-export interface ICachedPrompt {
-  /** Unique prompt identifier. */
-  id: string;
-  /** Path to the WAV file on disk. */
-  wavPath: string;
-  /** Total duration in milliseconds (approximate, from WAV header). */
-  durationMs: number;
-}
-
-// ---------------------------------------------------------------------------
-// TTS helpers
-// ---------------------------------------------------------------------------
-
-const TTS_DIR = path.join(process.cwd(), '.nogit', 'tts');
-
-/** Check if espeak-ng is available. */
-function isEspeakAvailable(): boolean {
-  try {
-    execSync('which espeak-ng', { stdio: 'pipe' });
-    return true;
-  } catch {
-    return false;
-  }
-}
-
-/** Generate WAV via espeak-ng. */
-function generateViaEspeak(wavPath: string, text: string): boolean {
-  try {
-    execSync(
-      `espeak-ng -v en-us -s 150 -w "${wavPath}" "${text}"`,
-      { timeout: 10000, stdio: 'pipe' },
-    );
-    return true;
-  } catch {
-    return false;
-  }
-}
-
-/** Generate WAV via Kokoro TTS (runs inside proxy-engine). */
-async function generateViaKokoro(wavPath: string, text: string, voice: string): Promise<boolean> {
-  const modelPath = path.join(TTS_DIR, 'kokoro-v1.0.onnx');
-  const voicesPath = path.join(TTS_DIR, 'voices.bin');
-  if (!fs.existsSync(modelPath) || !fs.existsSync(voicesPath)) return false;
-  if (!isProxyReady()) return false;
-
-  try {
-    await sendProxyCommand('generate_tts', {
-      model: modelPath,
-      voices: voicesPath,
-      voice,
-      text,
-      output: wavPath,
-    });
-    return true;
-  } catch {
-    return false;
-  }
-}
-
-/** Read a WAV file's duration from its header. */
-function getWavDurationMs(wavPath: string): number {
-  try {
-    const wav = fs.readFileSync(wavPath);
-    if (wav.length < 44) return 0;
-    if (wav.toString('ascii', 0, 4) !== 'RIFF') return 0;
-
-    let sampleRate = 16000;
-    let dataSize = 0;
-    let bitsPerSample = 16;
-    let channels = 1;
-    let offset = 12;
-
-    while (offset < wav.length - 8) {
-      const chunkId = wav.toString('ascii', offset, offset + 4);
-      const chunkSize = wav.readUInt32LE(offset + 4);
-      if (chunkId === 'fmt ') {
-        channels = wav.readUInt16LE(offset + 10);
-        sampleRate = wav.readUInt32LE(offset + 12);
-        bitsPerSample = wav.readUInt16LE(offset + 22);
-      }
-      if (chunkId === 'data') {
-        dataSize = chunkSize;
-      }
-      offset += 8 + chunkSize;
-      if (offset % 2 !== 0) offset++;
-    }
-
-    const bytesPerSample = (bitsPerSample / 8) * channels;
-    const totalSamples = bytesPerSample > 0 ? dataSize / bytesPerSample : 0;
-    return sampleRate > 0 ? Math.round((totalSamples / sampleRate) * 1000) : 0;
-  } catch {
-    return 0;
-  }
-}
-
-// ---------------------------------------------------------------------------
-// PromptCache
-// ---------------------------------------------------------------------------
-
-export class PromptCache {
-  private prompts = new Map<string, ICachedPrompt>();
-  private log: (msg: string) => void;
-  private espeakAvailable: boolean | null = null;
-
-  constructor(log: (msg: string) => void) {
-    this.log = log;
-  }
-
-  // -------------------------------------------------------------------------
-  // Public API
-  // -------------------------------------------------------------------------
-
-  /** Get a cached prompt by ID. */
-  get(id: string): ICachedPrompt | null {
-    return this.prompts.get(id) ?? null;
-  }
-
-  /** Check if a prompt is cached. */
-  has(id: string): boolean {
-    return this.prompts.has(id);
-  }
-
-  /** List all cached prompt IDs. */
-  listIds(): string[] {
-    return [...this.prompts.keys()];
-  }
-
-  /**
-   * Generate a TTS prompt WAV and cache its path.
-   * Uses espeak-ng (primary) or Kokoro (fallback).
-   */
-  async generatePrompt(id: string, text: string, voice = 'af_bella'): Promise<ICachedPrompt | null> {
-    fs.mkdirSync(TTS_DIR, { recursive: true });
-    const wavPath = path.join(TTS_DIR, `prompt-${id}.wav`);
-
-    // Check espeak availability once.
-    if (this.espeakAvailable === null) {
-      this.espeakAvailable = isEspeakAvailable();
-    }
-
-    // Generate WAV if not already on disk.
-    if (!fs.existsSync(wavPath)) {
-      let generated = false;
-      if (this.espeakAvailable) {
-        generated = generateViaEspeak(wavPath, text);
-      }
-      if (!generated) {
-        generated = await generateViaKokoro(wavPath, text, voice);
-      }
-      if (!generated) {
-        this.log(`[prompt-cache] failed to generate TTS for "${id}"`);
-        return null;
-      }
-      this.log(`[prompt-cache] generated WAV for "${id}"`);
-    }
-
-    return this.registerWav(id, wavPath);
-  }
-
-  /**
-   * Load a pre-existing WAV file as a prompt.
-   */
-  async loadWavPrompt(id: string, wavPath: string): Promise<ICachedPrompt | null> {
-    if (!fs.existsSync(wavPath)) {
-      this.log(`[prompt-cache] WAV not found: ${wavPath}`);
-      return null;
-    }
-    return this.registerWav(id, wavPath);
-  }
-
-  /**
-   * Generate a beep tone WAV and cache it.
-   */
-  async generateBeep(
-    id: string,
-    freqHz = 1000,
-    durationMs = 500,
-    amplitude = 8000,
-  ): Promise<ICachedPrompt | null> {
-    fs.mkdirSync(TTS_DIR, { recursive: true });
-    const wavPath = path.join(TTS_DIR, `prompt-${id}.wav`);
-
-    if (!fs.existsSync(wavPath)) {
-      // Generate 16kHz 16-bit mono sine wave WAV.
-      const sampleRate = 16000;
-      const totalSamples = Math.floor((sampleRate * durationMs) / 1000);
-      const pcm = Buffer.alloc(totalSamples * 2);
-
-      for (let i = 0; i < totalSamples; i++) {
-        const t = i / sampleRate;
-        const fadeLen = Math.floor(sampleRate * 0.01); // 10ms fade
-        let envelope = 1.0;
-        if (i < fadeLen) envelope = i / fadeLen;
-        else if (i > totalSamples - fadeLen) envelope = (totalSamples - i) / fadeLen;
-
-        const sample = Math.round(Math.sin(2 * Math.PI * freqHz * t) * amplitude * envelope);
-        pcm.writeInt16LE(Math.max(-32768, Math.min(32767, sample)), i * 2);
-      }
-
-      // Write WAV file.
-      const headerSize = 44;
-      const dataSize = pcm.length;
-      const wav = Buffer.alloc(headerSize + dataSize);
-
-      // RIFF header
-      wav.write('RIFF', 0);
-      wav.writeUInt32LE(36 + dataSize, 4);
-      wav.write('WAVE', 8);
-
-      // fmt chunk
-      wav.write('fmt ', 12);
-      wav.writeUInt32LE(16, 16);        // chunk size
-      wav.writeUInt16LE(1, 20);         // PCM format
-      wav.writeUInt16LE(1, 22);         // mono
-      wav.writeUInt32LE(sampleRate, 24);
-      wav.writeUInt32LE(sampleRate * 2, 28); // byte rate
-      wav.writeUInt16LE(2, 32);         // block align
-      wav.writeUInt16LE(16, 34);        // bits per sample
-
-      // data chunk
-      wav.write('data', 36);
-      wav.writeUInt32LE(dataSize, 40);
-      pcm.copy(wav, 44);
-
-      fs.writeFileSync(wavPath, wav);
-      this.log(`[prompt-cache] beep WAV generated for "${id}"`);
-    }
-
-    return this.registerWav(id, wavPath);
-  }
-
-  /** Remove a prompt from the cache. */
-  remove(id: string): void {
-    this.prompts.delete(id);
-  }
-
-  /** Clear all cached prompts. */
-  clear(): void {
-    this.prompts.clear();
-  }
-
-  // -------------------------------------------------------------------------
-  // Internal
-  // -------------------------------------------------------------------------
-
-  private registerWav(id: string, wavPath: string): ICachedPrompt {
-    const durationMs = getWavDurationMs(wavPath);
-    const prompt: ICachedPrompt = { id, wavPath, durationMs };
-    this.prompts.set(id, prompt);
-    this.log(`[prompt-cache] cached "${id}": ${wavPath} (${(durationMs / 1000).toFixed(1)}s)`);
-    return prompt;
-  }
-}
--- a/ts/config.ts
+++ b/ts/config.ts
@@ -8,6 +8,7 @@

 import fs from 'node:fs';
 import path from 'node:path';
+import type { IVoiceboxConfig } from './voicebox.js';

 // ---------------------------------------------------------------------------
 // Shared types (previously in ts/sip/types.ts, now inlined)
@@ -160,24 +161,13 @@ export interface IContact {
 // Voicebox configuration
 // ---------------------------------------------------------------------------

-export interface IVoiceboxConfig {
-  /** Unique ID — typically matches device ID or extension. */
-  id: string;
-  /** Whether this voicebox is active. */
-  enabled: boolean;
-  /** Custom TTS greeting text. */
-  greetingText?: string;
-  /** TTS voice ID (default 'af_bella'). */
-  greetingVoice?: string;
-  /** Path to uploaded WAV greeting (overrides TTS). */
-  greetingWavPath?: string;
-  /** Seconds to wait before routing to voicemail (default 25). */
-  noAnswerTimeoutSec?: number;
-  /** Maximum recording duration in seconds (default 120). */
-  maxRecordingSec?: number;
-  /** Maximum stored messages per box (default 50). */
-  maxMessages?: number;
-}
+// Canonical definition lives in voicebox.ts (imported at the top of this
+// file) — re-exported here so consumers can import everything from a
+// single config module without pulling in the voicebox implementation.
+// This used to be a duplicated interface and caused
+// "number | undefined is not assignable to number" type errors when
+// passing config.voiceboxes into VoiceboxManager.init().
+export type { IVoiceboxConfig };

 // ---------------------------------------------------------------------------
 // IVR configuration
--- a/ts/proxybridge.ts
+++ b/ts/proxybridge.ts
@@ -41,6 +41,14 @@ type TProxyCommands = {
    params: { call_id: string };
    result: { file_path: string; duration_ms: number };
  };
+  add_leg: {
+    params: { call_id: string; number: string; provider_id?: string };
+    result: { leg_id: string };
+  };
+  remove_leg: {
+    params: { call_id: string; leg_id: string };
+    result: Record<string, never>;
+  };
  add_device_leg: {
    params: { call_id: string; device_id: string };
    result: { leg_id: string };
@@ -80,9 +88,37 @@ type TProxyCommands = {
    result: Record<string, never>;
  };
  generate_tts: {
-    params: { model: string; voices: string; voice: string; text: string; output: string };
+    params: { model: string; voices: string; voice: string; text: string; output: string; cacheable?: boolean };
    result: { output: string };
  };
+  // WebRTC signaling — bridged from the browser via the TS control plane.
+  webrtc_offer: {
+    params: { session_id: string; sdp: string };
+    result: { sdp: string };
+  };
+  webrtc_ice: {
+    params: {
+      session_id: string;
+      candidate: string;
+      sdp_mid?: string;
+      sdp_mline_index?: number;
+    };
+    result: Record<string, never>;
+  };
+  webrtc_link: {
+    params: {
+      session_id: string;
+      call_id: string;
+      provider_media_addr: string;
+      provider_media_port: number;
+      sip_pt?: number;
+    };
+    result: Record<string, never>;
+  };
+  webrtc_close: {
+    params: { session_id: string };
+    result: Record<string, never>;
+  };
 };

 // ---------------------------------------------------------------------------
@@ -94,6 +130,11 @@ export interface IIncomingCallEvent {
  from_uri: string;
  to_number: string;
  provider_id: string;
+  /** Whether registered browsers should see a `webrtc-incoming` toast for
+   *  this call. Set by the Rust engine from the matched inbound route's
+   *  `ringBrowsers` flag (defaults to `true` when no route matches, so
+   *  deployments without explicit routes preserve pre-routing behavior). */
+  ring_browsers?: boolean;
 }

 export interface IOutboundCallEvent {
@@ -517,7 +558,7 @@ export async function sendProxyCommand<K extends keyof TProxyCommands>(
  params: TProxyCommands[K]['params'],
 ): Promise<TProxyCommands[K]['result']> {
  if (!bridge || !initialized) throw new Error('proxy engine not initialized');
-  return bridge.sendCommand(method as string, params as any) as any;
+  return bridge.sendCommand(method, params) as Promise<TProxyCommands[K]['result']>;
 }

 /** Shut down the proxy engine. */
--- a/ts/sipproxy.ts
+++ b/ts/sipproxy.ts
@@ -24,8 +24,6 @@ import {
  getAllBrowserDeviceIds,
  getBrowserDeviceWs,
 } from './webrtcbridge.ts';
-import { initAnnouncement } from './announcement.ts';
-import { PromptCache } from './call/prompt-cache.ts';
 import { VoiceboxManager } from './voicebox.ts';
 import {
  initProxyEngine,
@@ -170,7 +168,6 @@ for (const d of appConfig.devices) {
 // Initialize subsystems
 // ---------------------------------------------------------------------------

-const promptCache = new PromptCache(log);
 const voiceboxManager = new VoiceboxManager(log);
 voiceboxManager.init(appConfig.voiceboxes ?? []);

@@ -273,7 +270,14 @@ async function startProxyEngine(): Promise<void> {
      legs: new Map(),
    });

-    // Notify browsers of incoming call.
+    // Notify browsers of the incoming call, but only if the matched inbound
+    // route asked for it. `ring_browsers !== false` preserves today's
+    // ring-by-default behavior for any Rust release that predates this
+    // field or for the fallback "no route matched" case (where Rust still
+    // sends `true`). Note: this is an informational toast — browsers do
+    // NOT race the SIP device to answer. First-to-answer-wins requires
+    // a multi-leg fork which is not yet implemented.
+    if (data.ring_browsers !== false) {
      const browserIds = getAllBrowserDeviceIds();
      for (const bid of browserIds) {
        sendToBrowserDevice(bid, {
@@ -283,6 +287,7 @@ async function startProxyEngine(): Promise<void> {
          deviceId: bid,
        });
      }
+    }
  });

  onProxyEvent('outbound_device_call', (data: IOutboundCallEvent) => {
@@ -493,7 +498,7 @@ async function startProxyEngine(): Promise<void> {
  onProxyEvent('recording_done', (data: any) => {
    log(`[voicemail] recording done: ${data.file_path} (${data.duration_ms}ms) caller=${data.caller_number}`);
    // Save voicemail metadata via VoiceboxManager.
-    voiceboxManager.addMessage?.('default', {
+    voiceboxManager.addMessage('default', {
      callerNumber: data.caller_number || 'Unknown',
      callerName: null,
      fileName: data.file_path,
@@ -511,6 +516,8 @@ async function startProxyEngine(): Promise<void> {
    providers: appConfig.providers,
    devices: appConfig.devices,
    routing: appConfig.routing,
+    voiceboxes: appConfig.voiceboxes ?? [],
+    ivr: appConfig.ivr,
  });

  if (!configured) {
@@ -522,31 +529,8 @@ async function startProxyEngine(): Promise<void> {
  const deviceList = appConfig.devices.map((d) => d.displayName).join(', ');
  log(`proxy engine started | LAN ${appConfig.proxy.lanIp}:${appConfig.proxy.lanPort} | providers: ${providerList} | devices: ${deviceList}`);

-  // Generate TTS audio (WAV files on disk, played by Rust audio_player).
-  try {
-    await initAnnouncement(log);
-
-    // Pre-generate prompts.
-    await promptCache.generateBeep('voicemail-beep', 1000, 500, 8000);
-    for (const vb of appConfig.voiceboxes ?? []) {
-      if (!vb.enabled) continue;
-      const promptId = `voicemail-greeting-${vb.id}`;
-      if (vb.greetingWavPath) {
-        await promptCache.loadWavPrompt(promptId, vb.greetingWavPath);
-      } else {
-        const text = vb.greetingText || 'The person you are trying to reach is not available. Please leave a message after the tone.';
-        await promptCache.generatePrompt(promptId, text, vb.greetingVoice || 'af_bella');
-      }
-    }
-    if (appConfig.ivr?.enabled) {
-      for (const menu of appConfig.ivr.menus) {
-        await promptCache.generatePrompt(`ivr-menu-${menu.id}`, menu.promptText, menu.promptVoice || 'af_bella');
-      }
-    }
-    log(`[startup] prompts cached: ${promptCache.listIds().join(', ') || 'none'}`);
-  } catch (e) {
-    log(`[tts] init failed: ${e}`);
-  }
+  // TTS prompts (voicemail greetings, IVR menus) are generated on-demand
+  // by the Rust TTS engine when first needed. No startup pre-generation.
 }

 // ---------------------------------------------------------------------------
@@ -612,6 +596,8 @@ initWebUi(
        providers: fresh.providers,
        devices: fresh.devices,
        routing: fresh.routing,
+        voiceboxes: fresh.voiceboxes ?? [],
+        ivr: fresh.ivr,
      }).then((ok) => {
        if (ok) log('[config] reloaded — proxy engine reconfigured');
        else log('[config] reload failed — proxy engine rejected config');
--- a/ts/voicebox.ts
+++ b/ts/voicebox.ts
@@ -29,12 +29,14 @@ export interface IVoiceboxConfig {
  greetingVoice?: string;
  /** Path to uploaded WAV greeting (overrides TTS). */
  greetingWavPath?: string;
-  /** Seconds to wait before routing to voicemail (default 25). */
-  noAnswerTimeoutSec: number;
-  /** Maximum recording duration in seconds (default 120). */
-  maxRecordingSec: number;
-  /** Maximum stored messages per box (default 50). */
-  maxMessages: number;
+  /** Seconds to wait before routing to voicemail. Defaults to 25 when
+   *  absent — both the config loader and `VoiceboxManager.init` apply
+   *  the default via `??=`. */
+  noAnswerTimeoutSec?: number;
+  /** Maximum recording duration in seconds. Defaults to 120. */
+  maxRecordingSec?: number;
+  /** Maximum stored messages per box. Defaults to 50. */
+  maxMessages?: number;
 }

 export interface IVoicemailMessage {
@@ -148,6 +150,35 @@ export class VoiceboxManager {
  // Message CRUD
  // -------------------------------------------------------------------------

+  /**
+   * Convenience wrapper around `saveMessage` — used by the `recording_done`
+   * event handler, which has a raw recording path + caller info and needs
+   * to persist metadata. Generates `id`, sets `timestamp = now`, defaults
+   * `heard = false`, and normalizes `fileName` to a basename (the WAV is
+   * expected to already live in the box's directory).
+   */
+  addMessage(
+    boxId: string,
+    info: {
+      callerNumber: string;
+      callerName?: string | null;
+      fileName: string;
+      durationMs: number;
+    },
+  ): void {
+    const msg: IVoicemailMessage = {
+      id: crypto.randomUUID(),
+      boxId,
+      callerNumber: info.callerNumber,
+      callerName: info.callerName ?? undefined,
+      timestamp: Date.now(),
+      durationMs: info.durationMs,
+      fileName: path.basename(info.fileName),
+      heard: false,
+    };
+    this.saveMessage(msg);
+  }
+
  /**
   * Save a new voicemail message.
   * The WAV file should already exist at the expected path.
--- a/ts_web/00_commitinfo_data.ts
+++ b/ts_web/00_commitinfo_data.ts
@@ -3,6 +3,6 @@
 */
 export const commitinfo = {
  name: 'siprouter',
-  version: '1.20.0',
+  version: '1.22.0',
  description: 'undefined'
 }
--- a/ts_web/elements/sipproxy-view-providers.ts
+++ b/ts_web/elements/sipproxy-view-providers.ts
@@ -164,162 +164,269 @@ export class SipproxyViewProviders extends DeesElement {
        iconName: 'lucide:plus',
        type: ['header'] as any,
        actionFunc: async () => {
-          await this.openAddModal();
-        },
-      },
-      {
-        name: 'Add Sipgate',
-        iconName: 'lucide:phone',
-        type: ['header'] as any,
-        actionFunc: async () => {
-          await this.openAddModal(PROVIDER_TEMPLATES.sipgate, 'Sipgate');
-        },
-      },
-      {
-        name: 'Add O2/Alice',
-        iconName: 'lucide:phone',
-        type: ['header'] as any,
-        actionFunc: async () => {
-          await this.openAddModal(PROVIDER_TEMPLATES.o2, 'O2/Alice');
+          await this.openAddStepper();
        },
      },
    ];
  }

-  // ---- add provider modal --------------------------------------------------
+  // ---- add provider stepper ------------------------------------------------

-  private async openAddModal(
-    template?: typeof PROVIDER_TEMPLATES.sipgate,
-    templateName?: string,
-  ) {
-    const { DeesModal } = await import('@design.estate/dees-catalog');
+  private async openAddStepper() {
+    const { DeesStepper } = await import('@design.estate/dees-catalog');
+    type TDeesStepper = InstanceType<typeof DeesStepper>;
+    // IStep / menuOptions types: we keep content typing loose (`any[]`) to
+    // avoid having to import tsclass IMenuItem just for one parameter annotation.

-    const formData = {
-      displayName: templateName || '',
-      domain: template?.domain || '',
-      outboundProxyAddress: template?.outboundProxy?.address || '',
-      outboundProxyPort: String(template?.outboundProxy?.port ?? 5060),
+    type TProviderType = 'Custom' | 'Sipgate' | 'O2/Alice';
+    interface IAccumulator {
+      providerType: TProviderType;
+      displayName: string;
+      domain: string;
+      outboundProxyAddress: string;
+      outboundProxyPort: string;
+      username: string;
+      password: string;
+      // Advanced — exposed in step 4
+      registerIntervalSec: string;
+      codecs: string;
+      earlyMediaSilence: boolean;
+    }
+
+    const accumulator: IAccumulator = {
+      providerType: 'Custom',
+      displayName: '',
+      domain: '',
+      outboundProxyAddress: '',
+      outboundProxyPort: '5060',
      username: '',
      password: '',
-      registerIntervalSec: String(template?.registerIntervalSec ?? 300),
-      codecs: template?.codecs ? template.codecs.join(', ') : '9, 0, 8, 101',
-      earlyMediaSilence: template?.quirks?.earlyMediaSilence ?? false,
+      registerIntervalSec: '300',
+      codecs: '9, 0, 8, 101',
+      earlyMediaSilence: false,
    };

-    const heading = template
-      ? `Add ${templateName} Provider`
-      : 'Add Provider';
+    // Snapshot the currently-selected step's form (if any) into accumulator.
+    const snapshotActiveForm = async (stepper: TDeesStepper) => {
+      const form = stepper.activeForm;
+      if (!form) return;
+      const data: Record<string, any> = await form.collectFormData();
+      Object.assign(accumulator, data);
+    };

-    await DeesModal.createAndShow({
-      heading,
-      width: 'small',
-      showCloseButton: true,
+    // Overwrite template-owned fields. Keep user-owned fields (username,
+    // password) untouched. displayName is replaced only when empty or still
+    // holds a branded auto-fill.
+    const applyTemplate = (type: TProviderType) => {
+      const tpl =
+        type === 'Sipgate' ? PROVIDER_TEMPLATES.sipgate
+        : type === 'O2/Alice' ? PROVIDER_TEMPLATES.o2
+        : null;
+      if (!tpl) return;
+      accumulator.domain = tpl.domain;
+      accumulator.outboundProxyAddress = tpl.outboundProxy.address;
+      accumulator.outboundProxyPort = String(tpl.outboundProxy.port);
+      accumulator.registerIntervalSec = String(tpl.registerIntervalSec);
+      accumulator.codecs = tpl.codecs.join(', ');
+      accumulator.earlyMediaSilence = tpl.quirks.earlyMediaSilence;
+      if (
+        !accumulator.displayName ||
+        accumulator.displayName === 'Sipgate' ||
+        accumulator.displayName === 'O2/Alice'
+      ) {
+        accumulator.displayName = type;
+      }
+    };
+
+    // --- Step builders (called after step 1 so accumulator is populated) ---
+
+    const buildConnectionStep = (): any => ({
+      title: 'Connection',
      content: html`
-        <div style="display:flex;flex-direction:column;gap:12px;padding:4px 0;">
+        <dees-form>
          <dees-input-text
            .key=${'displayName'}
            .label=${'Display Name'}
-            .value=${formData.displayName}
-            @input=${(e: Event) => { formData.displayName = (e.target as any).value; }}
+            .value=${accumulator.displayName}
+            .required=${true}
          ></dees-input-text>
          <dees-input-text
            .key=${'domain'}
            .label=${'Domain'}
-            .value=${formData.domain}
-            @input=${(e: Event) => { formData.domain = (e.target as any).value; }}
+            .value=${accumulator.domain}
+            .required=${true}
          ></dees-input-text>
          <dees-input-text
            .key=${'outboundProxyAddress'}
            .label=${'Outbound Proxy Address'}
-            .value=${formData.outboundProxyAddress}
-            @input=${(e: Event) => { formData.outboundProxyAddress = (e.target as any).value; }}
+            .value=${accumulator.outboundProxyAddress}
          ></dees-input-text>
          <dees-input-text
            .key=${'outboundProxyPort'}
            .label=${'Outbound Proxy Port'}
-            .value=${formData.outboundProxyPort}
-            @input=${(e: Event) => { formData.outboundProxyPort = (e.target as any).value; }}
+            .value=${accumulator.outboundProxyPort}
          ></dees-input-text>
+        </dees-form>
+      `,
+      menuOptions: [
+        {
+          name: 'Continue',
+          iconName: 'lucide:arrow-right',
+          action: async (stepper: TDeesStepper) => {
+            await snapshotActiveForm(stepper);
+            stepper.goNext();
+          },
+        },
+      ],
+    });
+
+    const buildCredentialsStep = (): any => ({
+      title: 'Credentials',
+      content: html`
+        <dees-form>
          <dees-input-text
            .key=${'username'}
            .label=${'Username / Auth ID'}
-            .value=${formData.username}
-            @input=${(e: Event) => { formData.username = (e.target as any).value; }}
+            .value=${accumulator.username}
+            .required=${true}
          ></dees-input-text>
          <dees-input-text
            .key=${'password'}
            .label=${'Password'}
            .isPasswordBool=${true}
-            .value=${formData.password}
-            @input=${(e: Event) => { formData.password = (e.target as any).value; }}
+            .value=${accumulator.password}
+            .required=${true}
          ></dees-input-text>
+        </dees-form>
+      `,
+      menuOptions: [
+        {
+          name: 'Continue',
+          iconName: 'lucide:arrow-right',
+          action: async (stepper: TDeesStepper) => {
+            await snapshotActiveForm(stepper);
+            stepper.goNext();
+          },
+        },
+      ],
+    });
+
+    const buildAdvancedStep = (): any => ({
+      title: 'Advanced',
+      content: html`
+        <dees-form>
          <dees-input-text
            .key=${'registerIntervalSec'}
            .label=${'Register Interval (sec)'}
-            .value=${formData.registerIntervalSec}
-            @input=${(e: Event) => { formData.registerIntervalSec = (e.target as any).value; }}
+            .value=${accumulator.registerIntervalSec}
          ></dees-input-text>
          <dees-input-text
            .key=${'codecs'}
            .label=${'Codecs (comma-separated payload types)'}
-            .value=${formData.codecs}
-            @input=${(e: Event) => { formData.codecs = (e.target as any).value; }}
+            .value=${accumulator.codecs}
          ></dees-input-text>
          <dees-input-checkbox
            .key=${'earlyMediaSilence'}
            .label=${'Early Media Silence (quirk)'}
-            .value=${formData.earlyMediaSilence}
-            @newValue=${(e: CustomEvent) => { formData.earlyMediaSilence = e.detail; }}
+            .value=${accumulator.earlyMediaSilence}
          ></dees-input-checkbox>
-        </div>
+        </dees-form>
      `,
      menuOptions: [
        {
-          name: 'Cancel',
-          iconName: 'lucide:x',
-          action: async (modalRef: any) => {
-            modalRef.destroy();
+          name: 'Continue',
+          iconName: 'lucide:arrow-right',
+          action: async (stepper: TDeesStepper) => {
+            await snapshotActiveForm(stepper);
+            // Rebuild the review step so its rendering reflects the latest
+            // accumulator values (the review step captures values at build time).
+            stepper.steps = [...stepper.steps.slice(0, 4), buildReviewStep()];
+            await (stepper as any).updateComplete;
+            stepper.goNext();
          },
        },
+      ],
+    });
+
+    const buildReviewStep = (): any => {
+      const resolvedId = slugify(accumulator.displayName);
+      return {
+        title: 'Review & Create',
+        content: html`
+          <dees-panel>
+            <div
+              style="display:grid;grid-template-columns:auto 1fr;gap:6px 16px;font-size:.85rem;padding:8px 4px;"
+            >
+              <div style="color:#94a3b8;">Type</div>
+              <div>${accumulator.providerType}</div>
+              <div style="color:#94a3b8;">Display Name</div>
+              <div>${accumulator.displayName}</div>
+              <div style="color:#94a3b8;">ID</div>
+              <div style="font-family:'JetBrains Mono',monospace;">${resolvedId}</div>
+              <div style="color:#94a3b8;">Domain</div>
+              <div>${accumulator.domain}</div>
+              <div style="color:#94a3b8;">Outbound Proxy</div>
+              <div>
+                ${accumulator.outboundProxyAddress || accumulator.domain}:${accumulator.outboundProxyPort}
+              </div>
+              <div style="color:#94a3b8;">Username</div>
+              <div>${accumulator.username}</div>
+              <div style="color:#94a3b8;">Password</div>
+              <div>${'*'.repeat(Math.min(accumulator.password.length, 12))}</div>
+              <div style="color:#94a3b8;">Register Interval</div>
+              <div>${accumulator.registerIntervalSec}s</div>
+              <div style="color:#94a3b8;">Codecs</div>
+              <div>${accumulator.codecs}</div>
+              <div style="color:#94a3b8;">Early-Media Silence</div>
+              <div>${accumulator.earlyMediaSilence ? 'yes' : 'no'}</div>
+            </div>
+          </dees-panel>
+        `,
+        menuOptions: [
          {
-          name: 'Create',
+            name: 'Create Provider',
            iconName: 'lucide:check',
-          action: async (modalRef: any) => {
-            if (!formData.displayName.trim() || !formData.domain.trim()) {
-              deesCatalog.DeesToast.error('Display name and domain are required');
-              return;
+            action: async (stepper: TDeesStepper) => {
+              // Collision-resolve id against current state.
+              const existing = (this.appData.providers || []).map((p) => p.id);
+              let uniqueId = resolvedId;
+              let suffix = 2;
+              while (existing.includes(uniqueId)) {
+                uniqueId = `${resolvedId}-${suffix++}`;
              }
-            try {
-              const providerId = slugify(formData.displayName);
-              const codecs = formData.codecs
+
+              const parsedCodecs = accumulator.codecs
                .split(',')
                .map((s: string) => parseInt(s.trim(), 10))
                .filter((n: number) => !isNaN(n));

              const newProvider: any = {
-                id: providerId,
-                displayName: formData.displayName.trim(),
-                domain: formData.domain.trim(),
+                id: uniqueId,
+                displayName: accumulator.displayName.trim(),
+                domain: accumulator.domain.trim(),
                outboundProxy: {
-                  address: formData.outboundProxyAddress.trim() || formData.domain.trim(),
-                  port: parseInt(formData.outboundProxyPort, 10) || 5060,
+                  address:
+                    accumulator.outboundProxyAddress.trim() || accumulator.domain.trim(),
+                  port: parseInt(accumulator.outboundProxyPort, 10) || 5060,
                },
-                username: formData.username.trim(),
-                password: formData.password,
-                registerIntervalSec: parseInt(formData.registerIntervalSec, 10) || 300,
-                codecs,
+                username: accumulator.username.trim(),
+                password: accumulator.password,
+                registerIntervalSec: parseInt(accumulator.registerIntervalSec, 10) || 300,
+                codecs: parsedCodecs.length ? parsedCodecs : [9, 0, 8, 101],
                quirks: {
-                  earlyMediaSilence: formData.earlyMediaSilence,
+                  earlyMediaSilence: accumulator.earlyMediaSilence,
                },
              };

+              try {
                const result = await appState.apiSaveConfig({
                  addProvider: newProvider,
                });
-
                if (result.ok) {
-                modalRef.destroy();
-                deesCatalog.DeesToast.success(`Provider "${formData.displayName}" created`);
+                  await stepper.destroy();
+                  deesCatalog.DeesToast.success(
+                    `Provider "${newProvider.displayName}" created`,
+                  );
                } else {
                  deesCatalog.DeesToast.error('Failed to save provider');
                }
@@ -330,7 +437,73 @@ export class SipproxyViewProviders extends DeesElement {
            },
          },
        ],
-    });
+      };
+    };
+
+    // --- Step 1: Provider Type ------------------------------------------------
+    //
+    // Note: `DeesStepper.createAndShow()` dismisses on backdrop click; a user
+    // mid-form could lose work. Acceptable for v1 — revisit if users complain.
+
+    const typeOptions: { option: string; key: TProviderType }[] = [
+      { option: 'Custom', key: 'Custom' },
+      { option: 'Sipgate', key: 'Sipgate' },
+      { option: 'O2 / Alice', key: 'O2/Alice' },
+    ];
+    const currentTypeOption =
+      typeOptions.find((o) => o.key === accumulator.providerType) || null;
+
+    const typeStep: any = {
+      title: 'Choose provider type',
+      content: html`
+        <dees-form>
+          <dees-input-dropdown
+            .key=${'providerType'}
+            .label=${'Provider Type'}
+            .options=${typeOptions}
+            .selectedOption=${currentTypeOption}
+            .enableSearch=${false}
+            .required=${true}
+          ></dees-input-dropdown>
+        </dees-form>
+      `,
+      menuOptions: [
+        {
+          name: 'Continue',
+          iconName: 'lucide:arrow-right',
+          action: async (stepper: TDeesStepper) => {
+            // `dees-input-dropdown.value` is an object `{option, key, payload?}`,
+            // not a plain string — extract the `key` directly instead of using
+            // the generic `snapshotActiveForm` helper (which would clobber
+            // `accumulator.providerType`'s string type via Object.assign).
+            const form = stepper.activeForm;
+            if (form) {
+              const data: Record<string, any> = await form.collectFormData();
+              const selected = data.providerType;
+              if (selected && typeof selected === 'object' && selected.key) {
+                accumulator.providerType = selected.key as TProviderType;
+              }
+            }
+            if (!accumulator.providerType) {
+              accumulator.providerType = 'Custom';
+            }
+            applyTemplate(accumulator.providerType);
+            // (Re)build steps 2-5 with current accumulator values.
+            stepper.steps = [
+              typeStep,
+              buildConnectionStep(),
+              buildCredentialsStep(),
+              buildAdvancedStep(),
+              buildReviewStep(),
+            ];
+            await (stepper as any).updateComplete;
+            stepper.goNext();
+          },
+        },
+      ],
+    };
+
+    await DeesStepper.createAndShow({ steps: [typeStep] });
  }

  // ---- edit provider modal -------------------------------------------------
Author	SHA1	Message	Date
Juergen Kunz	5a280c5c41	v1.22.0 Some checks failed Docker (tags) / release (push) Failing after 2s Details	2026-04-12 20:45:08 +00:00
Juergen Kunz	59d8c2557c	feat(proxy-engine): add on-demand TTS caching for voicemail and IVR prompts	2026-04-12 20:45:08 +00:00
Juergen Kunz	cfadd7a2b6	v1.21.0 Some checks failed Docker (tags) / release (push) Failing after 3s Details	2026-04-11 20:04:56 +00:00
Juergen Kunz	80f710f6d8	feat(providers): replace provider creation modal with a guided multi-step setup flow	2026-04-11 20:04:56 +00:00
Juergen Kunz	9ea57cd659	v1.20.5 Some checks failed Docker (tags) / release (push) Failing after 3s Details	2026-04-11 19:20:14 +00:00
Juergen Kunz	c40c726dc3	fix(readme): improve architecture and call flow documentation with Mermaid diagrams	2026-04-11 19:20:14 +00:00
Juergen Kunz	37ba7501fa	v1.20.4 Some checks failed Docker (tags) / release (push) Failing after 3s Details	2026-04-11 19:17:39 +00:00
Juergen Kunz	24924a1aea	fix(deps): bump @design.estate/dees-catalog to ^3.71.1	2026-04-11 19:17:38 +00:00
Juergen Kunz	7ed76a9488	v1.20.3 Some checks failed Docker (tags) / release (push) Failing after 3s Details	2026-04-11 19:02:52 +00:00
Juergen Kunz	a9fdfe5733	fix(ts-config,proxybridge,voicebox): align voicebox config types and add missing proxy bridge command definitions	2026-04-11 19:02:52 +00:00
Juergen Kunz	6fcdf4291a	v1.20.2 Some checks failed Docker (tags) / release (push) Failing after 3s Details	2026-04-11 18:40:56 +00:00
Juergen Kunz	81441e7853	fix(proxy-engine): fix inbound route browser ringing and provider-facing SDP advertisement while preventing RTP port exhaustion	2026-04-11 18:40:56 +00:00
Juergen Kunz	21ffc1d017	v1.20.1 Some checks failed Docker (tags) / release (push) Failing after 3s Details	2026-04-11 12:32:46 +00:00
Juergen Kunz	2f16c5efae	fix(docker): install required native build tools for Rust dependencies in the build image	2026-04-11 12:32:46 +00:00