@craftedxp/sdk-node 0.20.0 → 0.20.1

package/README.md

@@ -198,6 +198,7 @@ const job = await client.speech.enqueue({
   agentId: 'agt_…',
   context: { goals: ['run 5k'], weather: 'sunny, 18°C' },
   instructions: '60–90 seconds, upbeat, finish by handing over to the live conversation',
+  emotion: 'hype', // emotional delivery on emotion-capable voices (see below)
   idempotencyKey: `${deviceId}:${alarmId}:${yyyymmdd}`,
   userTags: ['premium'],
   ttlSeconds: 6 * 3600,
@@ -209,7 +210,11 @@ const job = await client.speech.enqueue({
 
 // Sync path for short clips / dev: resolves with the ready asset.
 try {
-  const asset = await client.speech.synthesize({ agentId: 'agt_…', text: 'Good morning!' })
+  const asset = await client.speech.synthesize({
+    agentId: 'agt_…',
+    text: 'Good morning!',
+    emotion: 'hype',
+  })
   console.log(asset.url, asset.durationMs)
 } catch (err) {
   if (err instanceof PlatformError && err.code === 'accepted') {
@@ -220,6 +225,21 @@ try {
 
 Webhooks: `speech.ready` (`data: SpeechAsset`) and `speech.failed` (`data: SpeechJob`). Subscribe them on the agent's webhook like any other event. `format: 'ogg'` is reserved (400 today); `sampleRate` 24k/48k are upsampled from the provider's 16 kHz.
 
+### Emotion (0.20.0+)
+
+Pass an `emotion` preset on `synthesize()` / `enqueue()` to control delivery:
+
+```ts
+type Emotion = 'hype' | 'shout' | 'disappointed' | 'warm' | 'calm'
+
+await client.speech.enqueue({ agentId: 'agt_…', text: 'Wake up — big day!', emotion: 'hype' })
+```
+
+- Applies to **emotion-capable voices** (IndexTTS2, Hume); other voices silently ignore it (no error).
+- The emotion is applied **consistently across the whole message** (every sentence), so a long briefing keeps one delivery style.
+- Optional and backward-compatible — omit it for neutral delivery.
+- Typical use: pre-generate an emotional wake / notification message with `enqueue()` ahead of the trigger time, then play the ready asset. Add new presets server-side without an SDK change.
+
 ## Webhook signature verification
 
 ```ts
@@ -283,6 +303,7 @@ for await (const call of client.calls.listAll({ agentId })) {
 
 ## Changelog
 
+- **0.20.1** — docs: added the "Emotion" usage section + examples to the README (no API change; the `emotion` field ships from 0.20.0).
 - **0.20.0** — `speech.synthesize()` / `enqueue()` accept an `emotion` preset (`hype | shout | disappointed | warm | calm`) for emotional TTS voices.
 - **0.19.0** — type-only: `CallTokenMintResult.transport` widened to `'ws' | 'webrtc' | 'livekit'` (server returns `'livekit'` when the platform media edge is LiveKit Cloud; client SDKs that don't know it fall back to `'ws'`); `CallRecord.transport` added. No runtime change; drop-in for 0.18.0 consumers.
 - **Unreleased** _(forward-looking)_ — `isPersonal?: boolean` on `Agent` and `AgentCreateInput` (and so on `AgentUpdateInput`). Assistant-only: when `true`, the learning tools (`remember_fact`, `add_correction`, `add_skill_hint`) are live on every `ct_`-token call rather than only on `mode=learning`. Server-side field — no runtime change in this package.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@craftedxp/sdk-node",
-  "version": "0.20.0",
+  "version": "0.20.1",
   "description": "Node.js / TypeScript SDK for the voice agent platform. Server-side API client — mint call tokens, manage agents, query calls, upload knowledge-base docs.",
   "author": "Crafted XP",
   "license": "MIT",