@voicethere/agent 0.7.8 → 0.7.10

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@voicethere/agent",
-  "version": "0.7.8",
+  "version": "0.7.10",
   "description": "VoiceThere customer agent SDK — IPC types and runtime helpers for sandboxed child bundles",
   "type": "module",
   "exports": {
@@ -95,9 +95,9 @@
     "@node-webrtc-rust/sdk": ">=0.9.0"
   },
   "devDependencies": {
-    "@node-webrtc-rust/helpers": "0.9.0",
-    "@node-webrtc-rust/sdk": "0.9.0",
-    "@node-webrtc-rust/signaling": "0.9.0",
+    "@node-webrtc-rust/helpers": "0.9.5",
+    "@node-webrtc-rust/sdk": "0.9.5",
+    "@node-webrtc-rust/signaling": "0.9.5",
     "@types/node": "22.20.1",
     "http-server": "14.1.1",
     "ioredis": "5.11.1",

package/templates/README.md

@@ -17,7 +17,7 @@ import {
 
 | Kind        | Dashboard create                                                                                                                             | Prebuilt seed bundle                 | Typical consumer        |
 | ----------- | -------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------ | ----------------------- |
-| **product** | Yes (`echo`, `echo-dc`, `voice-starter`, `game-sync`, `voice-showcase`, `recording-consent`, `positional-tts`, `webhooks`, `webhooks-redis`) | Yes — `dist/templates/<id>/agent.js` | Platform project create |
+| **product** | Yes (`echo`, `echo-dc`, `voice-starter`, `game-sync`, `voice-showcase`, `recording-consent`, `positional-tts`, `spatial-showcase`, `webhooks`, `webhooks-redis`) | Yes — `dist/templates/<id>/agent.js` | Platform project create |
 | **e2e**     | No                                                                                                                                           | No — build from sources at test time | `voicethere/e2e` smokes |
 
 Product templates always set `seedOnCreate: true`. CI fails if a product template is missing its prebuilt bundle after `npm run build`.
@@ -79,6 +79,22 @@ Voice+Data demo — enables positional mixing and orbits each listener’s TTS s
 
 Sources: `positional-tts/agent.ts` (defineAgent wiring), `positional-tts/orbit.ts` (pure circle helper for tests).
 
+### `spatial-showcase/` (`spatial-showcase`)
+
+Spatial audio showcase for `/showcase` — one agent template with three demos selected by `{ type: "join", demo }` on the control DataChannel:
+
+| Demo | Behavior |
+| ---- | -------- |
+| `orbit` | TTS speaker orbits the listener (`setTtsPose`); `{ type: "orbit", action: "set" \| "say" }`; agent emits `{ type: "orbit_pose" }` |
+| `soundboard` | Positional clip pads — browser sends `clipId` only (never URLs); agent resolves against an allowlisted `assetOrigin` |
+| `proximity` | Shared mix room (`createMixGroup` + `setClientPose`); `{ type: "move" }`, `{ type: "mute_peer" }`; `{ type: "room_state" }` broadcast |
+
+**Protocol:** inbound `join`, `orbit`, `pad`, `pad_stop`, `pad_status`, `move`, `mute_peer`, `leave`, `ping` — outbound `showcase_ack`, `orbit_pose`, `pad_progress`, `room_state`, `room_full`, `error`.
+
+**Limits:** `MAX_ROOM_PEERS = 8`, `MAX_ACTIVE_PLAYS = 4` per session. Clip ids: `chime`, `bell`, `laser`, `impact`, `footsteps`, `rain-loop`, `cafe-loop`, `jingle`.
+
+Sources: `spatial-showcase/agent.ts`, `protocol.ts`, `sounds.ts`, `room.ts`, `orbit.ts`.
+
 ### `webhooks.ts` (`webhooks`)
 
 Inbound HTTP webhook sample — verifies `x-agent-webhook-signature` HMAC on the **raw body** with `AGENT_WEBHOOK_SIGNING_SECRET`, then `JSON.parse` and fans out to connected sessions via DataChannel + `speak`.

package/templates/crash.ts

@@ -6,17 +6,20 @@
  * - `{ type: "crash_exit" }` → process.exit(1)
  * - `{ type: "ping", id }` → `{ type: "pong", id }`
  * - `onUserSpeechFinal` text starting with "crash" → throw
- * - other finals → speak(`echo. ${text}`)
+ * - other finals → speak(`Okay. ${text}`)
  * - onSessionStart → speak("ready") after 1s (voice ready waiter)
  */
 import { defineAgent, sendToClient, speak } from "@voicethere/agent";
 
-/** TTS echo prefix with a sentence boundary so Piper does not glue words. */
+/**
+ * TTS reply prefix with a sentence boundary so Piper does not glue words.
+ * Use a word local streaming STT reliably emits a token for (`Okay.`); `echo.`
+ * measured ~5% no-token Piper renders. Never glue `echo:` onto the payload.
+ */
 export function formatEchoSpeak(text: string): string {
   const trimmed = text.trim();
   if (!trimmed) return "";
-  // Never glue `echo:` onto the next word; Piper skips "echo colon" on `echo:One`.
-  return `echo. ${trimmed}`;
+  return `Okay. ${trimmed}`;
 }
 
 export const CRASH_AGENT_MESSAGE =

package/templates/echo-smoke.ts

@@ -5,11 +5,15 @@
  */
 import { defineAgent, parseChatText, speak } from "@voicethere/agent";
 
-/** TTS echo prefix with a sentence boundary so Piper does not glue words. */
+/**
+ * TTS reply prefix with a sentence boundary so Piper does not glue words.
+ * Use a word local streaming STT (Kroko Zipformer) reliably emits a token for;
+ * `echo.` measured ~5% no-token Piper renders (audio intact). `Okay.` does not.
+ */
 export function formatEchoSpeak(text: string): string {
   const trimmed = text.trim();
   if (!trimmed) return "";
-  return `echo. ${trimmed}`;
+  return `Okay. ${trimmed}`;
 }
 
 defineAgent({