@absolutejs/voice 0.0.22-beta.617 → 0.0.22-beta.619

package/dist/core/types.d.ts

@@ -531,10 +531,18 @@ export type VoiceSessionHandle<TContext = unknown, TSession extends VoiceSession
         transcriptStabilityMs: number;
     }>;
 };
+export type VoiceLLMUsage = {
+    provider?: string;
+    model?: string;
+    inputTokens?: number;
+    outputTokens?: number;
+    cachedInputTokens?: number;
+};
 export type VoiceRouteResult<TResult = unknown> = {
     complete?: boolean;
     result?: TResult;
     assistantText?: string;
+    usage?: VoiceLLMUsage;
     citations?: ReadonlyArray<VoiceTurnCitation>;
     transfer?: {
         metadata?: Record<string, unknown>;

package/dist/index.js

@@ -5898,6 +5898,9 @@ var createVoiceSession = (options) => {
       });
       committedOutput = undefined;
     }
+    if (options.costAccountant && committedOutput?.usage) {
+      options.costAccountant.recordLLM(committedOutput.usage);
+    }
     const output = {
       assistantText: committedOutput?.assistantText,
       citations: committedOutput?.citations,
@@ -7612,6 +7615,20 @@ var createVoiceRAGTool = (collection, options = {}) => {
 };
 
 // src/core/agent.ts
+var addVoiceUsage = (acc, next) => {
+  if (!next)
+    return acc;
+  if (!acc)
+    return next;
+  const sum = (left, right) => left === undefined && right === undefined ? undefined : (left ?? 0) + (right ?? 0);
+  return {
+    cachedInputTokens: sum(acc.cachedInputTokens, next.cachedInputTokens),
+    inputTokens: sum(acc.inputTokens, next.inputTokens),
+    model: next.model ?? acc.model,
+    outputTokens: sum(acc.outputTokens, next.outputTokens),
+    provider: next.provider ?? acc.provider
+  };
+};
 var normalizeText3 = (value) => typeof value === "string" ? value.trim() : "";
 var toErrorMessage3 = (error) => error instanceof Error ? error.message : String(error);
 var createHistoryMessages = (session, turn) => {
@@ -7830,6 +7847,7 @@ var createVoiceAgent = (options) => {
 `) || undefined;
     stamp("agent.system-resolved", { systemChars: system?.length ?? 0 });
     let output = {};
+    let turnUsage;
     for (let round = 0;round <= maxToolRounds; round += 1) {
       const modelStartedAt = Date.now();
       stamp(`agent.round${round}.generate-start`);
@@ -7848,6 +7866,7 @@ var createVoiceAgent = (options) => {
           })),
           turn: input.turn
         });
+        turnUsage = addVoiceUsage(turnUsage, output.usage);
         stamp(`agent.round${round}.generate-done`, {
           ms: Date.now() - modelStartedAt,
           textChars: output.assistantText?.length ?? 0,
@@ -8111,6 +8130,7 @@ var createVoiceAgent = (options) => {
       result: output.result,
       toolResults,
       transfer: output.transfer,
+      ...turnUsage ? { usage: turnUsage } : {},
       voicemail: output.voicemail
     };
   };
@@ -39579,6 +39599,7 @@ var voice = (config) => {
   const monitorBindings = new Map;
   const runtime = {
     activeSessions: new Map,
+    pendingSessions: new Map,
     logger: resolveLogger(config.logger),
     profileSwitchGuardAutoSwitchCounts: new Map,
     profileSwitchGuardedSessions: new Set,
@@ -39738,6 +39759,31 @@ var voice = (config) => {
       turnDetection: sessionOptions.turnDetection
     });
   };
+  const createAndConnectSession = async (ws, sessionId, scenarioId) => {
+    const session = await createManagedSession(ws, sessionId, scenarioId);
+    const typedSession = session;
+    runtime.activeSessions.set(sessionId, typedSession);
+    registerMonitorSession(sessionId, typedSession);
+    await session.connect(buildSocketAdapter(ws, sessionId));
+    return typedSession;
+  };
+  const ensureManagedSession = async (ws, sessionId, scenarioId) => {
+    const active = runtime.activeSessions.get(sessionId);
+    if (active) {
+      return active;
+    }
+    const inFlight = runtime.pendingSessions.get(sessionId);
+    if (inFlight) {
+      return inFlight;
+    }
+    const creation = createAndConnectSession(ws, sessionId, scenarioId);
+    runtime.pendingSessions.set(sessionId, creation);
+    try {
+      return await creation;
+    } finally {
+      runtime.pendingSessions.delete(sessionId);
+    }
+  };
   const mountSurface = (app, value, factory) => {
     if (value === undefined || value === false) {
       return app;
@@ -39950,13 +39996,7 @@ var voice = (config) => {
       if (!audio) {
         return;
       }
-      const session = current ?? await createManagedSession(ws, sessionState.sessionId, sessionState.scenarioId ?? undefined);
-      if (!current) {
-        const typedSession = session;
-        runtime.activeSessions.set(sessionState.sessionId, typedSession);
-        registerMonitorSession(sessionState.sessionId, typedSession);
-        await session.connect(buildSocketAdapter(ws, sessionState.sessionId));
-      }
+      const session = current ?? await ensureManagedSession(ws, sessionState.sessionId, sessionState.scenarioId ?? undefined);
       await session.receiveAudio(audio);
     },
     open: async (ws) => {
@@ -39967,11 +40007,7 @@ var voice = (config) => {
         runtime.activeSessions.delete(sessionState.sessionId);
         deregisterMonitorSession(sessionState.sessionId, "superseded");
       }
-      const session = await createManagedSession(ws, sessionState.sessionId, sessionState.scenarioId ?? undefined);
-      const typedSession = session;
-      runtime.activeSessions.set(sessionState.sessionId, typedSession);
-      registerMonitorSession(sessionState.sessionId, typedSession);
-      await session.connect(buildSocketAdapter(ws, sessionState.sessionId));
+      await ensureManagedSession(ws, sessionState.sessionId, sessionState.scenarioId ?? undefined);
     }
   }).use(htmxRoutes()).use(surfaceRoutes());
 };
@@ -41796,6 +41832,13 @@ var shouldRetainVoiceRecording = (policy) => policy.retain.recordings;
 var shouldRetainVoiceTranscript = (policy) => policy.retain.transcriptText;
 // src/core/costAccounting.ts
 var DEFAULT_VOICE_PRICE_BOOK = {
+  "anthropic:claude-haiku-4-5-20251001": {
+    llm: {
+      cachedInputPerMillionTokensUsd: 0.1,
+      inputPerMillionTokensUsd: 1,
+      outputPerMillionTokensUsd: 5
+    }
+  },
   "anthropic:claude-opus-4-5": {
     llm: {
       cachedInputPerMillionTokensUsd: 1.5,
@@ -41803,6 +41846,13 @@ var DEFAULT_VOICE_PRICE_BOOK = {
       outputPerMillionTokensUsd: 75
     }
   },
+  "anthropic:claude-opus-4-6": {
+    llm: {
+      cachedInputPerMillionTokensUsd: 0.5,
+      inputPerMillionTokensUsd: 5,
+      outputPerMillionTokensUsd: 25
+    }
+  },
   "anthropic:claude-sonnet-4-5": {
     llm: {
       cachedInputPerMillionTokensUsd: 0.3,
@@ -41810,11 +41860,25 @@ var DEFAULT_VOICE_PRICE_BOOK = {
       outputPerMillionTokensUsd: 15
     }
   },
+  "anthropic:claude-sonnet-4-6": {
+    llm: {
+      cachedInputPerMillionTokensUsd: 0.3,
+      inputPerMillionTokensUsd: 3,
+      outputPerMillionTokensUsd: 15
+    }
+  },
   "assemblyai:streaming": { stt: { perSecondUsd: 0.00018 } },
   "azure:tts-neural": { tts: { perMillionCharactersUsd: 16 } },
   "cartesia:sonic-2": { tts: { perMillionCharactersUsd: 65 } },
   "deepgram:nova-3": { stt: { perSecondUsd: 0.000077 } },
   "elevenlabs:flash-v2-5": { tts: { perMillionCharactersUsd: 50 } },
+  "openai:gpt-4.1": {
+    llm: {
+      cachedInputPerMillionTokensUsd: 0.5,
+      inputPerMillionTokensUsd: 2,
+      outputPerMillionTokensUsd: 8
+    }
+  },
   "openai:gpt-4o-mini": {
     llm: {
       cachedInputPerMillionTokensUsd: 0.075,
@@ -45262,6 +45326,16 @@ var createVoiceFileRecordingStore = (options) => {
   return { get, list, put };
 };
 // src/core/modelAdapters.ts
+var normalizeVoiceUsage = (provider, model, raw) => {
+  if (!raw)
+    return;
+  const toNum = (value) => typeof value === "number" && Number.isFinite(value) ? value : undefined;
+  const details = typeof raw.input_tokens_details === "object" && raw.input_tokens_details !== null ? raw.input_tokens_details : undefined;
+  const inputTokens = toNum(raw.input_tokens) ?? toNum(raw.prompt_tokens) ?? toNum(raw.promptTokenCount);
+  const outputTokens = toNum(raw.output_tokens) ?? toNum(raw.completion_tokens) ?? toNum(raw.candidatesTokenCount);
+  const cachedInputTokens = toNum(raw.cache_read_input_tokens) ?? toNum(details?.cached_tokens) ?? toNum(raw.cachedContentTokenCount);
+  return { cachedInputTokens, inputTokens, model, outputTokens, provider };
+};
 var isVoiceProviderRoutingPolicyPreset = (value) => value === "balanced" || value === "cost-cap" || value === "cost-first" || value === "latency-first" || value === "quality-first";
 var resolveVoiceProviderRoutingPolicyPreset = (preset, options = {}) => {
   switch (preset) {
@@ -46122,7 +46196,8 @@ var createOpenAIVoiceAssistantModel = (options) => {
       }
       return {
         ...assistantText ? { assistantText } : {},
-        ...toolCalls.length ? { toolCalls } : {}
+        ...toolCalls.length ? { toolCalls } : {},
+        ...usage ? { usage: normalizeVoiceUsage("openai", options.model, usage) } : {}
       };
     }
   };
@@ -46205,7 +46280,8 @@ var createAnthropicVoiceAssistantModel = (options) => {
       }
       return {
         ...assistantText ? { assistantText } : {},
-        ...toolCalls.length ? { toolCalls } : {}
+        ...toolCalls.length ? { toolCalls } : {},
+        ...usage ? { usage: normalizeVoiceUsage("anthropic", options.model, usage) } : {}
       };
     }
   };
@@ -46311,7 +46387,8 @@ var createGeminiVoiceAssistantModel = (options) => {
       }
       return {
         ...assistantText ? { assistantText } : {},
-        ...toolCalls.length ? { toolCalls } : {}
+        ...toolCalls.length ? { toolCalls } : {},
+        ...usage ? { usage: normalizeVoiceUsage("google", options.model, usage) } : {}
       };
     }
   };

package/dist/testing/index.js

@@ -4285,6 +4285,16 @@ var hardenFetch = (baseFetch = globalThis.fetch) => Object.assign(async (input,
 }, { preconnect: baseFetch.preconnect.bind(baseFetch) });
 
 // src/core/modelAdapters.ts
+var normalizeVoiceUsage = (provider, model, raw) => {
+  if (!raw)
+    return;
+  const toNum = (value) => typeof value === "number" && Number.isFinite(value) ? value : undefined;
+  const details = typeof raw.input_tokens_details === "object" && raw.input_tokens_details !== null ? raw.input_tokens_details : undefined;
+  const inputTokens = toNum(raw.input_tokens) ?? toNum(raw.prompt_tokens) ?? toNum(raw.promptTokenCount);
+  const outputTokens = toNum(raw.output_tokens) ?? toNum(raw.completion_tokens) ?? toNum(raw.candidatesTokenCount);
+  const cachedInputTokens = toNum(raw.cache_read_input_tokens) ?? toNum(details?.cached_tokens) ?? toNum(raw.cachedContentTokenCount);
+  return { cachedInputTokens, inputTokens, model, outputTokens, provider };
+};
 var isVoiceProviderRoutingPolicyPreset = (value) => value === "balanced" || value === "cost-cap" || value === "cost-first" || value === "latency-first" || value === "quality-first";
 var resolveVoiceProviderRoutingPolicyPreset = (preset, options = {}) => {
   switch (preset) {
@@ -5145,7 +5155,8 @@ var createOpenAIVoiceAssistantModel = (options) => {
       }
       return {
         ...assistantText ? { assistantText } : {},
-        ...toolCalls.length ? { toolCalls } : {}
+        ...toolCalls.length ? { toolCalls } : {},
+        ...usage ? { usage: normalizeVoiceUsage("openai", options.model, usage) } : {}
       };
     }
   };
@@ -5228,7 +5239,8 @@ var createAnthropicVoiceAssistantModel = (options) => {
       }
       return {
         ...assistantText ? { assistantText } : {},
-        ...toolCalls.length ? { toolCalls } : {}
+        ...toolCalls.length ? { toolCalls } : {},
+        ...usage ? { usage: normalizeVoiceUsage("anthropic", options.model, usage) } : {}
       };
     }
   };
@@ -5334,7 +5346,8 @@ var createGeminiVoiceAssistantModel = (options) => {
       }
       return {
         ...assistantText ? { assistantText } : {},
-        ...toolCalls.length ? { toolCalls } : {}
+        ...toolCalls.length ? { toolCalls } : {},
+        ...usage ? { usage: normalizeVoiceUsage("google", options.model, usage) } : {}
       };
     }
   };
@@ -8199,6 +8212,9 @@ var createVoiceSession = (options) => {
       });
       committedOutput = undefined;
     }
+    if (options.costAccountant && committedOutput?.usage) {
+      options.costAccountant.recordLLM(committedOutput.usage);
+    }
     const output = {
       assistantText: committedOutput?.assistantText,
       citations: committedOutput?.citations,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@absolutejs/voice",
-  "version": "0.0.22-beta.617",
+  "version": "0.0.22-beta.619",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",