@avakado.ai/schemas 1.1.32 → 1.1.33

package/README.md

@@ -123,3 +123,19 @@ await Subscription.createIndexes();
 ```
 
 That is why an index or enum change is a major version bump.
+
+### Realtime speech providers
+
+Realtime agents may store `speechConfig` separately from the model's
+`responseConfig`. Omit it for existing/native voice behavior, or explicitly use
+`{ provider: "native" }`. An external voice is configured as follows:
+
+```js
+speechConfig: {
+  provider: "elevenlabs",
+  voiceId: "your-voice-id",
+  voiceName: "My voice", // Optional display metadata
+  modelId: "eleven_flash_v2_5",
+  settings: { stability: 0.5, similarityBoost: 0.75 }
+}
+```

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@avakado.ai/schemas",
-  "version": "1.1.32",
+  "version": "1.1.33",
   "description": "Canonical Mongo schemas, models, enums and index declarations shared by the Avakado services.",
   "license": "UNLICENSED",
   "type": "module",

package/src/models/Agent.js

@@ -83,6 +83,22 @@ const ModelConfigSchema = new Schema({
 const ResponseConfigSchema = new Schema({
     provider: { type: String, enum: [...Object.keys(ProviderConfig), 'elevenlabs', 'custom'], required: true },
 }, { _id: false, discriminatorKey: 'provider' });
+const SpeechConfigSchema = new Schema({
+    provider: { type: String, enum: ['native', 'elevenlabs'], required: true },
+}, { _id: false, discriminatorKey: 'provider', strict: 'throw' });
+SpeechConfigSchema.discriminator('native', new Schema({}, { _id: false, strict: 'throw' }));
+SpeechConfigSchema.discriminator('elevenlabs', new Schema({
+    voiceId: { type: String, required: true, trim: true },
+    voiceName: { type: String, trim: true },
+    modelId: { type: String, required: true, trim: true },
+    settings: {
+        type: new Schema({
+            stability: { type: Number, min: 0, max: 1 },
+            similarityBoost: { type: Number, min: 0, max: 1 },
+        }, { _id: false, strict: 'throw' }),
+        default: undefined,
+    },
+}, { _id: false, strict: 'throw' }));
 const RealtimeConfigSchema = new Schema({
     modality: {
         type: String, required: true,
@@ -95,7 +111,9 @@ const RealtimeConfigSchema = new Schema({
             message: props => `Invalid modality "${props.value}" for this provider`,
         }
     },
-    responseConfig: ResponseConfigSchema
+    responseConfig: ResponseConfigSchema,
+    // Missing on existing agents means native speech; no migration required.
+    speechConfig: { type: SpeechConfigSchema, default: undefined },
 }, { _id: false });
 const responseConfigPath = RealtimeConfigSchema.path('responseConfig');
 responseConfigPath.discriminator('openai', new Schema({

package/types.d.ts

@@ -46,6 +46,7 @@ export interface AgentModel {
     personalInfo?: { name?: string; description?: string; avatar?: string; systemPrompt?: { role?: string; objective?: string; instructionsAndWorkflow?: string; constraintsAndRules?: string; }; };
     modality?: string;
     responseConfig?: { provider: "openai" | "gemini" | "anthropic" | "elevenlabs" | "custom"; };
+    speechConfig?: { provider: "native"; } | { voiceId: string; voiceName?: string; modelId: string; settings?: { stability?: number; similarityBoost?: number; }; provider: "elevenlabs"; };
 }
 
 export interface Amount {

package/src/media.test.js

@@ -1,85 +0,0 @@
-import assert from "node:assert/strict";
-import test from "node:test";
-import { MediaPlanError, negotiateCall, selectChannelMedia } from "./media.js";
-
-test("exotel 8 kHz pcm is resampled to openai 24 kHz", () => {
-    const plan = negotiateCall({ channelName: "Exotel", provider: "openai" });
-    assert.deepEqual(plan.channel, { encoding: "pcm16", rate: 8000 });
-    assert.deepEqual(plan.provider.input, { encoding: "pcm16", rate: 24000 });
-    assert.deepEqual(plan.provider.output, { encoding: "pcm16", rate: 24000 });
-    assert.equal(plan.uplink, "pcm16:8000->pcm16:24000");
-    assert.equal(plan.downlink, "pcm16:24000->pcm16:8000");
-});
-
-test("exotel 16 kHz pcm is resampled to openai 24 kHz", () => {
-    const plan = negotiateCall({
-        channelName: "exotel",
-        channelConfig: { media: { encoding: "pcm16", rate: 16000 } },
-        provider: "openai",
-    });
-    assert.equal(plan.channel.rate, 16000);
-    assert.equal(plan.provider.input.rate, 24000);
-    assert.equal(plan.uplink, "pcm16:16000->pcm16:24000");
-    assert.equal(plan.downlink, "pcm16:24000->pcm16:16000");
-});
-
-test("exotel 8 kHz and gemini bridge 16 kHz in and 24 kHz out", () => {
-    const plan = negotiateCall({ channelName: "exotel", provider: "gemini" });
-    assert.deepEqual(plan.provider.input, { encoding: "pcm16", rate: 16000 });
-    assert.deepEqual(plan.provider.output, { encoding: "pcm16", rate: 24000 });
-    assert.equal(plan.uplink, "pcm16:8000->pcm16:16000");
-    assert.equal(plan.downlink, "pcm16:24000->pcm16:8000");
-});
-
-test("exotel 16 kHz into gemini only downsamples the model output", () => {
-    const plan = negotiateCall({
-        channelName: "exotel",
-        channelConfig: { media: { encoding: "pcm16", rate: 16000 } },
-        provider: "google",
-    });
-    assert.equal(plan.uplink, "identity");
-    assert.equal(plan.downlink, "pcm16:24000->pcm16:16000");
-});
-
-test("exotel 24 kHz into gemini only downsamples the caller audio", () => {
-    const plan = negotiateCall({
-        channelName: "exotel",
-        channelConfig: { media: { encoding: "pcm16", rate: 24000 } },
-        provider: "gemini",
-    });
-    assert.equal(plan.uplink, "pcm16:24000->pcm16:16000");
-    assert.equal(plan.downlink, "identity");
-});
-
-test("whatsapp and twilio default to g711 ulaw and match openai", () => {
-    for (const channelName of ["Whatsapp", "twilio", "tataTele"]) {
-        const plan = negotiateCall({ channelName, provider: "openai" });
-        assert.deepEqual(plan.channel, { encoding: "g711_ulaw", rate: 8000 });
-        assert.equal(plan.uplink, "identity");
-        assert.equal(plan.downlink, "identity");
-    }
-});
-
-test("whatsapp ulaw into gemini uses the registered converters", () => {
-    const plan = negotiateCall({ channelName: "whatsapp", provider: "gemini" });
-    assert.equal(plan.uplink, "g711_ulaw:8000->pcm16:16000");
-    assert.equal(plan.downlink, "pcm16:24000->g711_ulaw:8000");
-});
-
-test("a wire outside the channel menu is rejected", () => {
-    assert.throws(
-        () => selectChannelMedia("exotel", { encoding: "pcm16", rate: 48000 }),
-        MediaPlanError
-    );
-});
-
-test("alaw into gemini fails until a converter exists", () => {
-    assert.throws(
-        () => negotiateCall({
-            channelName: "whatsapp",
-            channelConfig: { media: { encoding: "g711_alaw", rate: 8000 } },
-            provider: "gemini",
-        }),
-        /No converter/
-    );
-});