@ai-matrx/media 0.15.47 → 0.15.49

package/dist/speech.d.cts

@@ -1,4 +1,4 @@
-import { ProviderSessionFailure, ProviderSessionFailureVerdict } from '@ai-matrx/agents/matrx';
+import { ProviderSessionFailure, ProviderSessionFailureVerdict, ProviderSessionUsage } from '@ai-matrx/agents/matrx';
 import { b as SpeechPronunciation, V as VoicePurpose } from './tts-config-JbF86KIt.cjs';
 import { GenerationConfig } from '@cartesia/cartesia-js/resources/tts';
 import { SupportedLanguage } from '@cartesia/cartesia-js/resources/voices';
@@ -124,6 +124,11 @@ interface SpeechPorts {
      * person's sentence and whether retrying can help.
      */
     reportProviderFailure?: (failure: ProviderSessionFailure) => Promise<ProviderSessionFailureVerdict | null>;
+    /**
+     * Report what a browser-held provider session spent (`POST /broker/usage`, once per `usage_id`).
+     * The host owns retry and delivery; it must never throw.
+     */
+    reportProviderUsage?: (usage: ProviderSessionUsage) => void;
     /** Tell the person speech stopped (a toast on the web). Without it the error is only on the item. */
     notifyError?: (title: string, description: string) => void;
     /**

package/dist/speech.d.ts

@@ -1,4 +1,4 @@
-import { ProviderSessionFailure, ProviderSessionFailureVerdict } from '@ai-matrx/agents/matrx';
+import { ProviderSessionFailure, ProviderSessionFailureVerdict, ProviderSessionUsage } from '@ai-matrx/agents/matrx';
 import { b as SpeechPronunciation, V as VoicePurpose } from './tts-config-JbF86KIt.js';
 import { GenerationConfig } from '@cartesia/cartesia-js/resources/tts';
 import { SupportedLanguage } from '@cartesia/cartesia-js/resources/voices';
@@ -124,6 +124,11 @@ interface SpeechPorts {
      * person's sentence and whether retrying can help.
      */
     reportProviderFailure?: (failure: ProviderSessionFailure) => Promise<ProviderSessionFailureVerdict | null>;
+    /**
+     * Report what a browser-held provider session spent (`POST /broker/usage`, once per `usage_id`).
+     * The host owns retry and delivery; it must never throw.
+     */
+    reportProviderUsage?: (usage: ProviderSessionUsage) => void;
     /** Tell the person speech stopped (a toast on the web). Without it the error is only on the item. */
     notifyError?: (title: string, description: string) => void;
     /**

package/dist/speech.js

@@ -795,6 +795,8 @@ var init_cartesia_socket = __esm({
       #listeners = /* @__PURE__ */ new Map();
       /** Set when the provider refused this account and reconnecting cannot help. */
       #refused = null;
+      /** Sends accepted per context, for the usage idempotency key. */
+      #sendCounts = /* @__PURE__ */ new Map();
       constructor(native, options) {
         this.#native = native;
         this.#options = options;
@@ -864,7 +866,7 @@ var init_cartesia_socket = __esm({
           source = new CartesiaAudioSource(this.#options.sampleRate);
           this.#sources.set(contextId, source);
         }
-        await this.#native.send({
+        const sent = this.#native.send({
           context_id: contextId,
           model_id: request.modelId,
           transcript: request.transcript,
@@ -881,6 +883,8 @@ var init_cartesia_socket = __esm({
           max_buffer_delay_ms: request.maxBufferDelayMs,
           generation_config: request.generationConfig
         });
+        await sent;
+        this.#reportUsage(contextId, request);
         return {
           source,
           on: (_event, listener) => {
@@ -890,6 +894,25 @@ var init_cartesia_socket = __esm({
           }
         };
       }
+      /**
+       * Cartesia bills the characters it is sent. Report each accepted send once: the context id plus a
+       * per-context sequence is the idempotency key, so the host's retries can never double-bill.
+       */
+      #reportUsage(contextId, request) {
+        const characters = request.transcript?.length ?? 0;
+        if (characters <= 0) return;
+        const sequence = (this.#sendCounts.get(contextId) ?? 0) + 1;
+        this.#sendCounts.set(contextId, sequence);
+        try {
+          speechPorts().reportProviderUsage?.({
+            provider: "cartesia",
+            model: request.modelId,
+            usage_id: `${contextId}:${sequence}`,
+            characters
+          });
+        } catch {
+        }
+      }
       continue(request) {
         return this.send(request);
       }