@absolutejs/voice 0.0.22-beta.643 → 0.0.22-beta.645

package/dist/index.js

@@ -4574,6 +4574,8 @@ var createVoiceSession = (options) => {
     user: 0
   };
   const recordingFormats = {};
+  const recordingStartedAt = Date.now();
+  let assistantRecordingEndsAtMs = 0;
   let recordingPersisted = false;
   const captureRecordingChunk = (channel, bytes, format) => {
     if (!recordingConfig || recordingPersisted) {
@@ -4589,11 +4591,33 @@ var createVoiceSession = (options) => {
     if (currentTotal >= recordingMaxBytes) {
       return;
     }
-    const remaining = recordingMaxBytes - currentTotal;
+    const bytesPerSample = 2;
+    const bytesPerSecond = format.sampleRateHz * format.channels * bytesPerSample;
+    let chunkStartMs = currentTotal / bytesPerSecond * 1000;
+    if (channel === "assistant" || currentTotal === 0) {
+      const wallClockOffsetMs = Math.max(0, Date.now() - recordingStartedAt);
+      chunkStartMs = channel === "assistant" ? Math.max(wallClockOffsetMs, assistantRecordingEndsAtMs) : wallClockOffsetMs;
+      const targetByteOffset = Math.floor(chunkStartMs * bytesPerSecond / 1000 / bytesPerSample) * bytesPerSample;
+      const silenceByteLength = Math.min(Math.max(0, targetByteOffset - currentTotal), recordingMaxBytes - currentTotal);
+      if (silenceByteLength > 0) {
+        recordingBuffers[channel].push(new Uint8Array(silenceByteLength));
+        recordingByteTotals[channel] += silenceByteLength;
+      }
+      if (channel === "assistant") {
+        assistantRecordingEndsAtMs = chunkStartMs + bytes.byteLength / bytesPerSecond * 1000;
+      }
+    }
+    const updatedTotal = recordingByteTotals[channel];
+    if (updatedTotal >= recordingMaxBytes) {
+      recordingFormats[channel] = format;
+      return;
+    }
+    const remaining = recordingMaxBytes - updatedTotal;
     const slice = bytes.byteLength <= remaining ? bytes : bytes.subarray(0, remaining);
     recordingBuffers[channel].push(new Uint8Array(slice));
     recordingByteTotals[channel] += slice.byteLength;
     recordingFormats[channel] = format;
+    return { byteLength: slice.byteLength, startMs: chunkStartMs };
   };
   const pruneTurnAudio = () => {
     const replayWindowMs = sttFallback?.replayWindowMs ?? DEFAULT_FALLBACK_REPLAY_MS;
@@ -4921,7 +4945,7 @@ var createVoiceSession = (options) => {
   };
   const sendAssistantAudio = async (chunk, input) => {
     const normalizedChunk = chunk instanceof Uint8Array ? new Uint8Array(chunk) : chunk instanceof ArrayBuffer ? new Uint8Array(chunk.slice(0)) : new Uint8Array(chunk.buffer.slice(chunk.byteOffset, chunk.byteOffset + chunk.byteLength));
-    captureRecordingChunk("assistant", normalizedChunk, input.format);
+    const recordingChunk = captureRecordingChunk("assistant", normalizedChunk, input.format);
     kickCallSilenceWatchdog();
     await send({
       chunkBase64: encodeBase64(normalizedChunk),
@@ -4940,6 +4964,10 @@ var createVoiceSession = (options) => {
     if (activeTTSTurnId) {
       await appendTurnLatencyStage({
         at: input.receivedAt,
+        metadata: recordingChunk ? {
+          recordingByteLength: recordingChunk.byteLength,
+          recordingStartMs: recordingChunk.startMs
+        } : undefined,
         stage: "assistant_audio_received",
         turnId: activeTTSTurnId
       });

package/dist/manifest.d.ts

@@ -1,2 +1,17 @@
 import type { VoicePluginConfig, VoiceSessionStore } from "./core/types";
-export declare const manifest: import("@absolutejs/manifest").PackageManifest<VoicePluginConfig, VoiceSessionStore>;
+export declare const manifest: Omit<import("@absolutejs/manifest").PackageManifest<VoicePluginConfig, VoiceSessionStore>, "tools" | "contract" | "settings"> & (({
+    contract: 1;
+    tools?: Record<string, import("@absolutejs/manifest").LegacyManifestTool<VoiceSessionStore>> | undefined;
+} | {
+    contract: 2;
+    tools?: Record<string, import("@absolutejs/manifest").AuthorizedManifestTool<VoiceSessionStore>> | undefined;
+}) & {
+    settings: import("@sinclair/typebox").TObject<{
+        backchannelBargeInGuard: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TBoolean>;
+        greeting: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        normalizeNumbers: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TBoolean>;
+        path: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        preset: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"default">, import("@sinclair/typebox").TLiteral<"chat">, import("@sinclair/typebox").TLiteral<"guided-intake">, import("@sinclair/typebox").TLiteral<"dictation">, import("@sinclair/typebox").TLiteral<"noisy-room">, import("@sinclair/typebox").TLiteral<"pstn-balanced">, import("@sinclair/typebox").TLiteral<"pstn-fast">, import("@sinclair/typebox").TLiteral<"reliability">]>>;
+        resumeGreeting: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    }>;
+});

package/dist/manifest.js

@@ -102,14 +102,19 @@ var defineImplementation = () => (implementation) => {
   const defined = implementation;
   return defined;
 };
-var defineManifest = () => (manifest) => {
-  const defined = manifest;
-  return defined;
+var defineManifest = () => (manifest) => manifest;
+var literal = (value) => value;
+var RUNTIME_KIND = literal("runtime");
+var WORKSPACE_KIND = literal("workspace");
+var toolFactory = () => {
+  function runtime(definition) {
+    return { kind: RUNTIME_KIND, ...definition };
+  }
+  function workspace(definition) {
+    return { kind: WORKSPACE_KIND, ...definition };
+  }
+  return { runtime, workspace };
 };
-var toolFactory = () => ({
-  runtime: (definition) => ({ kind: "runtime", ...definition }),
-  workspace: (definition) => ({ kind: "workspace", ...definition })
-});
 function IsAsyncIterator(value) {
   return IsObject(value) && globalThis.Symbol.asyncIterator in value;
 }
@@ -847,7 +852,7 @@ function String2(options) {
 function* FromUnion(syntax) {
   const trim = syntax.trim().replace(/"|'/g, "");
   return trim === "boolean" ? yield Boolean() : trim === "number" ? yield Number2() : trim === "bigint" ? yield BigInt2() : trim === "string" ? yield String2() : yield (() => {
-    const literals = trim.split("|").map((literal) => Literal(literal.trim()));
+    const literals = trim.split("|").map((literal2) => Literal(literal2.trim()));
     return literals.length === 0 ? Never() : literals.length === 1 ? literals[0] : UnionEvaluated(literals);
   })();
 }
@@ -3555,10 +3560,10 @@ function ScoreUnion(schema, references, value) {
     const keys = Object.getOwnPropertyNames(value);
     const entries = Object.entries(object.properties);
     return entries.reduce((acc, [key, schema2]) => {
-      const literal = schema2[Kind] === "Literal" && schema2.const === value[key] ? 100 : 0;
+      const literal2 = schema2[Kind] === "Literal" && schema2.const === value[key] ? 100 : 0;
       const checks = Check(schema2, references, value[key]) ? 10 : 0;
       const exists = keys.includes(key) ? 1 : 0;
-      return acc + (literal + checks + exists);
+      return acc + (literal2 + checks + exists);
     }, 0);
   } else if (schema[Kind] === "Union") {
     const schemas = schema.anyOf.map((schema2) => Deref(schema2, references));
@@ -5895,7 +5900,13 @@ var wiringSnippet = Type.Object({
     Type.Literal("server-plugin")
   ]))
 });
-var clientFrameworks = ["angular", "client", "react", "svelte", "vue"];
+var clientFrameworks = [
+  "angular",
+  "client",
+  "react",
+  "svelte",
+  "vue"
+];
 var wiringRecipe = Type.Object({
   client: Type.Optional(Type.Partial(Type.Object(Object.fromEntries(clientFrameworks.map((framework) => [framework, wiringSnippet]))))),
   description: Type.Optional(Type.String()),
@@ -5960,8 +5971,47 @@ var toolAnnotations = Type.Object({
   readOnlyHint: Type.Optional(Type.Boolean()),
   title: Type.Optional(Type.String())
 });
+var toolAuthorization = Type.Object({
+  approval: Type.Union([
+    Type.Literal("always"),
+    Type.Literal("never"),
+    Type.Literal("policy")
+  ]),
+  audience: Type.Union([
+    Type.Literal("admin"),
+    Type.Literal("authenticated"),
+    Type.Literal("owner"),
+    Type.Literal("public")
+  ]),
+  compensatingTool: Type.Optional(Type.String({ pattern: TOOL_NAME_PATTERN.source })),
+  destinationFields: Type.Optional(Type.Array(Type.String({ minLength: 1 }), { minItems: 1 })),
+  destinations: Type.Optional(Type.Array(Type.String({ minLength: 1 }), { minItems: 1 })),
+  effects: Type.Array(Type.String({ minLength: 1 }), { minItems: 1 }),
+  idempotency: Type.Optional(Type.Union([
+    Type.Object({
+      field: Type.String({ minLength: 1 }),
+      mode: Type.Literal("field")
+    }),
+    Type.Object({ mode: Type.Literal("host") }),
+    Type.Object({ mode: Type.Literal("resource") })
+  ])),
+  requiredScopes: Type.Optional(Type.Array(Type.String({ minLength: 1 }), { minItems: 1 })),
+  resource: Type.Optional(Type.Object({
+    idField: Type.Optional(Type.String({ minLength: 1 })),
+    ownerIdField: Type.Optional(Type.String({ minLength: 1 })),
+    tenantIdField: Type.Optional(Type.String({ minLength: 1 })),
+    type: Type.String({ minLength: 1 })
+  })),
+  reversible: Type.Optional(Type.Boolean()),
+  spend: Type.Optional(Type.Object({
+    amountMinorField: Type.String({ minLength: 1 }),
+    currencyField: Type.String({ minLength: 1 }),
+    maximumAmountMinor: Type.Optional(Type.Integer({ minimum: 0 }))
+  }))
+});
 var serializedTool = Type.Object({
   annotations: Type.Optional(toolAnnotations),
+  authorization: Type.Optional(toolAuthorization),
   capabilities: Type.Optional(Type.Array(Type.Union([
     Type.Literal("exec"),
     Type.Literal("glob"),
@@ -5973,7 +6023,15 @@ var serializedTool = Type.Object({
   kind: Type.Union([Type.Literal("runtime"), Type.Literal("workspace")])
 });
 var manifestSchema = Type.Object({
-  contract: Type.Literal(1),
+  contract: Type.Union([Type.Literal(1), Type.Literal(2)]),
+  discovery: Type.Optional(Type.Object({
+    audiences: Type.Optional(Type.Array(Type.String({ minLength: 1 }))),
+    certificationUrl: Type.Optional(Type.String()),
+    intents: Type.Optional(Type.Array(Type.String({ minLength: 1 }))),
+    keywords: Type.Optional(Type.Array(Type.String({ minLength: 1 }))),
+    protocols: Type.Optional(Type.Array(Type.String({ minLength: 1 }))),
+    url: Type.Optional(Type.String())
+  })),
   identity: Type.Object({
     accent: Type.Optional(Type.String({ pattern: "^#[0-9a-fA-F]{3,8}$" })),
     category: Type.String({ minLength: 1 }),
@@ -5983,7 +6041,9 @@ var manifestSchema = Type.Object({
       svg: Type.Optional(Type.String()),
       url: Type.Optional(Type.String())
     })),
-    name: Type.String({ pattern: "^(@[a-z0-9-~][a-z0-9-._~]*/)?[a-z0-9-~][a-z0-9-._~]*$" }),
+    name: Type.String({
+      pattern: "^(@[a-z0-9-~][a-z0-9-._~]*/)?[a-z0-9-~][a-z0-9-._~]*$"
+    }),
     tagline: Type.String({ minLength: 1 })
   }),
   implements: Type.Optional(Type.Array(adapterImplementation)),
@@ -6956,7 +7016,7 @@ function String3(options) {
 function* FromUnion21(syntax) {
   const trim = syntax.trim().replace(/"|'/g, "");
   return trim === "boolean" ? yield Boolean2() : trim === "number" ? yield Number3() : trim === "bigint" ? yield BigInt3() : trim === "string" ? yield String3() : yield (() => {
-    const literals = trim.split("|").map((literal) => Literal2(literal.trim()));
+    const literals = trim.split("|").map((literal2) => Literal2(literal2.trim()));
     return literals.length === 0 ? Never2() : literals.length === 1 ? literals[0] : UnionEvaluated2(literals);
   })();
 }
@@ -8557,7 +8617,7 @@ var tool = toolFactory();
 var MAX_LIST_LIMIT = 200;
 var MAX_TRANSCRIPT_TURNS = 20;
 var manifest = defineManifest()({
-  contract: 1,
+  contract: 2,
   identity: {
     accent: "#f43f5e",
     category: "voice",
@@ -8760,6 +8820,15 @@ var manifest = defineManifest()({
   tools: {
     delete_session: tool.runtime({
       annotations: { destructiveHint: true, idempotentHint: true },
+      authorization: {
+        approval: "policy",
+        audience: "owner",
+        effects: ["delete"],
+        idempotency: { mode: "resource" },
+        requiredScopes: ["voice:sessions:delete"],
+        resource: { idField: "id", type: "voice-session" },
+        reversible: false
+      },
       description: "Delete one call session (its transcript and state) by id. Deleting an unknown id succeeds.",
       input: Type2.Object({ id: Type2.String({ minLength: 1 }) }),
       handler: async ({ id }, store) => {
@@ -8769,6 +8838,13 @@ var manifest = defineManifest()({
     }),
     get_session: tool.runtime({
       annotations: { readOnlyHint: true },
+      authorization: {
+        approval: "never",
+        audience: "owner",
+        effects: ["read"],
+        requiredScopes: ["voice:sessions:read"],
+        resource: { idField: "id", type: "voice-session" }
+      },
       description: "Fetch one call session by id: status, timing, and the committed conversation turns (caller text and assistant reply), most recent last.",
       input: Type2.Object({
         id: Type2.String({ minLength: 1 }),
@@ -8796,6 +8872,12 @@ var manifest = defineManifest()({
     }),
     list_sessions: tool.runtime({
       annotations: { readOnlyHint: true },
+      authorization: {
+        approval: "never",
+        audience: "owner",
+        effects: ["read"],
+        requiredScopes: ["voice:sessions:read"]
+      },
       description: "List call sessions with status, creation time, last activity, and turn count \u2014 most recently active first.",
       input: Type2.Object({
         limit: Type2.Optional(Type2.Integer({ maximum: MAX_LIST_LIMIT, minimum: 1 }))

package/dist/manifest.json

@@ -1,5 +1,5 @@
 {
-	"contract": 1,
+	"contract": 2,
 	"identity": {
 		"accent": "#f43f5e",
 		"category": "voice",
@@ -325,6 +325,24 @@
 				"destructiveHint": true,
 				"idempotentHint": true
 			},
+			"authorization": {
+				"approval": "policy",
+				"audience": "owner",
+				"effects": [
+					"delete"
+				],
+				"idempotency": {
+					"mode": "resource"
+				},
+				"requiredScopes": [
+					"voice:sessions:delete"
+				],
+				"resource": {
+					"idField": "id",
+					"type": "voice-session"
+				},
+				"reversible": false
+			},
 			"description": "Delete one call session (its transcript and state) by id. Deleting an unknown id succeeds.",
 			"input": {
 				"type": "object",
@@ -344,6 +362,20 @@
 			"annotations": {
 				"readOnlyHint": true
 			},
+			"authorization": {
+				"approval": "never",
+				"audience": "owner",
+				"effects": [
+					"read"
+				],
+				"requiredScopes": [
+					"voice:sessions:read"
+				],
+				"resource": {
+					"idField": "id",
+					"type": "voice-session"
+				}
+			},
 			"description": "Fetch one call session by id: status, timing, and the committed conversation turns (caller text and assistant reply), most recent last.",
 			"input": {
 				"type": "object",
@@ -368,6 +400,16 @@
 			"annotations": {
 				"readOnlyHint": true
 			},
+			"authorization": {
+				"approval": "never",
+				"audience": "owner",
+				"effects": [
+					"read"
+				],
+				"requiredScopes": [
+					"voice:sessions:read"
+				]
+			},
 			"description": "List call sessions with status, creation time, last activity, and turn count — most recently active first.",
 			"input": {
 				"type": "object",

package/dist/testing/index.js

@@ -7211,6 +7211,8 @@ var createVoiceSession = (options) => {
     user: 0
   };
   const recordingFormats = {};
+  const recordingStartedAt = Date.now();
+  let assistantRecordingEndsAtMs = 0;
   let recordingPersisted = false;
   const captureRecordingChunk = (channel, bytes2, format) => {
     if (!recordingConfig || recordingPersisted) {
@@ -7226,11 +7228,33 @@ var createVoiceSession = (options) => {
     if (currentTotal >= recordingMaxBytes) {
       return;
     }
-    const remaining = recordingMaxBytes - currentTotal;
+    const bytesPerSample = 2;
+    const bytesPerSecond = format.sampleRateHz * format.channels * bytesPerSample;
+    let chunkStartMs = currentTotal / bytesPerSecond * 1000;
+    if (channel === "assistant" || currentTotal === 0) {
+      const wallClockOffsetMs = Math.max(0, Date.now() - recordingStartedAt);
+      chunkStartMs = channel === "assistant" ? Math.max(wallClockOffsetMs, assistantRecordingEndsAtMs) : wallClockOffsetMs;
+      const targetByteOffset = Math.floor(chunkStartMs * bytesPerSecond / 1000 / bytesPerSample) * bytesPerSample;
+      const silenceByteLength = Math.min(Math.max(0, targetByteOffset - currentTotal), recordingMaxBytes - currentTotal);
+      if (silenceByteLength > 0) {
+        recordingBuffers[channel].push(new Uint8Array(silenceByteLength));
+        recordingByteTotals[channel] += silenceByteLength;
+      }
+      if (channel === "assistant") {
+        assistantRecordingEndsAtMs = chunkStartMs + bytes2.byteLength / bytesPerSecond * 1000;
+      }
+    }
+    const updatedTotal = recordingByteTotals[channel];
+    if (updatedTotal >= recordingMaxBytes) {
+      recordingFormats[channel] = format;
+      return;
+    }
+    const remaining = recordingMaxBytes - updatedTotal;
     const slice = bytes2.byteLength <= remaining ? bytes2 : bytes2.subarray(0, remaining);
     recordingBuffers[channel].push(new Uint8Array(slice));
     recordingByteTotals[channel] += slice.byteLength;
     recordingFormats[channel] = format;
+    return { byteLength: slice.byteLength, startMs: chunkStartMs };
   };
   const pruneTurnAudio = () => {
     const replayWindowMs = sttFallback?.replayWindowMs ?? DEFAULT_FALLBACK_REPLAY_MS;
@@ -7558,7 +7582,7 @@ var createVoiceSession = (options) => {
   };
   const sendAssistantAudio = async (chunk, input) => {
     const normalizedChunk = chunk instanceof Uint8Array ? new Uint8Array(chunk) : chunk instanceof ArrayBuffer ? new Uint8Array(chunk.slice(0)) : new Uint8Array(chunk.buffer.slice(chunk.byteOffset, chunk.byteOffset + chunk.byteLength));
-    captureRecordingChunk("assistant", normalizedChunk, input.format);
+    const recordingChunk = captureRecordingChunk("assistant", normalizedChunk, input.format);
     kickCallSilenceWatchdog();
     await send({
       chunkBase64: encodeBase64(normalizedChunk),
@@ -7577,6 +7601,10 @@ var createVoiceSession = (options) => {
     if (activeTTSTurnId) {
       await appendTurnLatencyStage({
         at: input.receivedAt,
+        metadata: recordingChunk ? {
+          recordingByteLength: recordingChunk.byteLength,
+          recordingStartMs: recordingChunk.startMs
+        } : undefined,
         stage: "assistant_audio_received",
         turnId: activeTTSTurnId
       });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@absolutejs/voice",
-  "version": "0.0.22-beta.643",
+  "version": "0.0.22-beta.645",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",
@@ -140,13 +140,13 @@
     }
   },
   "dependencies": {
-    "@absolutejs/manifest": "^0.1.0",
+    "@absolutejs/manifest": "^0.4.0",
     "@absolutejs/media": "0.0.1-beta.19",
     "@sinclair/typebox": "^0.34.0"
   },
   "devDependencies": {
     "@absolutejs/absolute": "0.19.0-beta.1009",
-    "@absolutejs/ai": "0.0.5",
+    "@absolutejs/ai": "0.0.46",
     "@angular/core": "^21.0.0",
     "@electric-sql/pglite": "^0.4.5",
     "@eslint/js": "^10.0.1",
@@ -169,6 +169,6 @@
     "vue": "3.5.27"
   },
   "absolutejs": {
-    "manifestContract": 1
+    "manifestContract": 2
   }
 }