@gobing-ai/knowledge-kit 0.0.17 → 0.0.19

package/plugins/generations/voice-gen/dist/index.js

@@ -14642,72 +14642,6 @@ var init_zod = __esm(() => {
   init_external();
 });
 
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/errors.js
-var ErrorCode;
-var init_errors3 = __esm(() => {
-  ErrorCode = {
-    NotFound: "NOT_FOUND",
-    Validation: "VALIDATION",
-    Conflict: "CONFLICT",
-    Internal: "INTERNAL"
-  };
-});
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/api-response.js
-var API_ERROR_CODES, ERROR_CODE_TO_HTTP, CLIENT_SAFE_CODES;
-var init_api_response = __esm(() => {
-  init_errors3();
-  API_ERROR_CODES = {
-    SUCCESS: 0,
-    NOT_FOUND: 404,
-    VALIDATION_ERROR: 422,
-    BAD_REQUEST: 400,
-    UNAUTHORIZED: 401,
-    FORBIDDEN: 403,
-    CONFLICT: 409,
-    INTERNAL_ERROR: 500
-  };
-  ERROR_CODE_TO_HTTP = {
-    [ErrorCode.NotFound]: API_ERROR_CODES.NOT_FOUND,
-    [ErrorCode.Validation]: API_ERROR_CODES.VALIDATION_ERROR,
-    [ErrorCode.Conflict]: API_ERROR_CODES.CONFLICT,
-    [ErrorCode.Internal]: API_ERROR_CODES.INTERNAL_ERROR
-  };
-  CLIENT_SAFE_CODES = new Set([ErrorCode.NotFound, ErrorCode.Validation, ErrorCode.Conflict]);
-});
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/cursor.js
-var init_cursor = () => {};
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/event-severity.js
-var init_event_severity = () => {};
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/output.js
-function processStream(name) {
-  const proc = globalThis.process;
-  const stream = proc?.[name];
-  if (stream === undefined) {
-    throw new Error(`No ${name} target available: set one via setDefaultOutputTargets or pass an explicit target`);
-  }
-  return stream;
-}
-function writeLine(message, target) {
-  target.write(`${message}
-`);
-}
-function echoError(message, target = defaultStderrTarget ?? processStream("stderr")) {
-  writeLine(message, target);
-}
-var defaultStderrTarget;
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/index.js
-var init_dist = __esm(() => {
-  init_api_response();
-  init_cursor();
-  init_errors3();
-  init_event_severity();
-});
-
 // ../../node_modules/.bun/yaml@2.9.0/node_modules/yaml/dist/nodes/identity.js
 var require_identity = __commonJS((exports) => {
   var ALIAS = Symbol.for("yaml.alias");
@@ -17612,15 +17546,15 @@ var require_timestamp = __commonJS((exports) => {
         throw new Error("!!timestamp expects a date, starting with yyyy-mm-dd");
       const [, year, month, day, hour, minute, second] = match.map(Number);
       const millisec = match[7] ? Number((match[7] + "00").substr(1, 3)) : 0;
-      let date6 = Date.UTC(year, month - 1, day, hour || 0, minute || 0, second || 0, millisec);
+      let date5 = Date.UTC(year, month - 1, day, hour || 0, minute || 0, second || 0, millisec);
       const tz = match[8];
       if (tz && tz !== "Z") {
         let d = parseSexagesimal(tz, false);
         if (Math.abs(d) < 30)
           d *= 60;
-        date6 -= 60000 * d;
+        date5 -= 60000 * d;
       }
-      return new Date(date6);
+      return new Date(date5);
     },
     stringify: ({ value }) => value?.toISOString().replace(/(T00:00:00)?\.000Z$/, "") ?? ""
   };
@@ -19480,7 +19414,7 @@ var require_composer = __commonJS((exports) => {
   var node_process = __require("process");
   var directives = require_directives();
   var Document = require_Document();
-  var errors4 = require_errors();
+  var errors3 = require_errors();
   var identity = require_identity();
   var composeDoc = require_compose_doc();
   var resolveEnd = require_resolve_end();
@@ -19531,9 +19465,9 @@ var require_composer = __commonJS((exports) => {
       this.onError = (source, code, message, warning) => {
         const pos = getErrorPos(source);
         if (warning)
-          this.warnings.push(new errors4.YAMLWarning(pos, code, message));
+          this.warnings.push(new errors3.YAMLWarning(pos, code, message));
         else
-          this.errors.push(new errors4.YAMLParseError(pos, code, message));
+          this.errors.push(new errors3.YAMLParseError(pos, code, message));
       };
       this.directives = new directives.Directives({ version: options.version || "1.2" });
       this.options = options;
@@ -19619,7 +19553,7 @@ ${cb}` : comment;
           break;
         case "error": {
           const msg = token.source ? `${token.message}: ${JSON.stringify(token.source)}` : token.message;
-          const error51 = new errors4.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg);
+          const error51 = new errors3.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg);
           if (this.atDirectives || !this.doc)
             this.errors.push(error51);
           else
@@ -19629,7 +19563,7 @@ ${cb}` : comment;
         case "doc-end": {
           if (!this.doc) {
             const msg = "Unexpected doc-end without preceding document";
-            this.errors.push(new errors4.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg));
+            this.errors.push(new errors3.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg));
             break;
           }
           this.doc.directives.docEnd = true;
@@ -19644,7 +19578,7 @@ ${end.comment}` : end.comment;
           break;
         }
         default:
-          this.errors.push(new errors4.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", `Unsupported token ${token.type}`));
+          this.errors.push(new errors3.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", `Unsupported token ${token.type}`));
       }
     }
     *end(forceDoc = false, endOffset = -1) {
@@ -19670,7 +19604,7 @@ ${end.comment}` : end.comment;
 var require_cst_scalar = __commonJS((exports) => {
   var resolveBlockScalar = require_resolve_block_scalar();
   var resolveFlowScalar = require_resolve_flow_scalar();
-  var errors4 = require_errors();
+  var errors3 = require_errors();
   var stringifyString = require_stringifyString();
   function resolveAsScalar(token, strict = true, onError) {
     if (token) {
@@ -19679,7 +19613,7 @@ var require_cst_scalar = __commonJS((exports) => {
         if (onError)
           onError(offset, code, message);
         else
-          throw new errors4.YAMLParseError([offset, offset + 1], code, message);
+          throw new errors3.YAMLParseError([offset, offset + 1], code, message);
       };
       switch (token.type) {
         case "scalar":
@@ -21559,7 +21493,7 @@ var require_parser = __commonJS((exports) => {
 var require_public_api = __commonJS((exports) => {
   var composer = require_composer();
   var Document = require_Document();
-  var errors4 = require_errors();
+  var errors3 = require_errors();
   var log = require_log();
   var identity = require_identity();
   var lineCounter = require_line_counter();
@@ -21576,8 +21510,8 @@ var require_public_api = __commonJS((exports) => {
     const docs = Array.from(composer$1.compose(parser$1.parse(source)));
     if (prettyErrors && lineCounter2)
       for (const doc2 of docs) {
-        doc2.errors.forEach(errors4.prettifyError(source, lineCounter2));
-        doc2.warnings.forEach(errors4.prettifyError(source, lineCounter2));
+        doc2.errors.forEach(errors3.prettifyError(source, lineCounter2));
+        doc2.warnings.forEach(errors3.prettifyError(source, lineCounter2));
       }
     if (docs.length > 0)
       return docs;
@@ -21592,13 +21526,13 @@ var require_public_api = __commonJS((exports) => {
       if (!doc2)
         doc2 = _doc;
       else if (doc2.options.logLevel !== "silent") {
-        doc2.errors.push(new errors4.YAMLParseError(_doc.range.slice(0, 2), "MULTIPLE_DOCS", "Source contains multiple documents; please use YAML.parseAllDocuments()"));
+        doc2.errors.push(new errors3.YAMLParseError(_doc.range.slice(0, 2), "MULTIPLE_DOCS", "Source contains multiple documents; please use YAML.parseAllDocuments()"));
         break;
       }
     }
     if (prettyErrors && lineCounter2) {
-      doc2.errors.forEach(errors4.prettifyError(source, lineCounter2));
-      doc2.warnings.forEach(errors4.prettifyError(source, lineCounter2));
+      doc2.errors.forEach(errors3.prettifyError(source, lineCounter2));
+      doc2.warnings.forEach(errors3.prettifyError(source, lineCounter2));
     }
     return doc2;
   }
@@ -21650,12 +21584,12 @@ var require_public_api = __commonJS((exports) => {
 });
 
 // ../../node_modules/.bun/yaml@2.9.0/node_modules/yaml/dist/index.js
-var composer, Document, Schema, errors4, Alias, identity, Pair, Scalar, YAMLMap, YAMLSeq, cst, lexer, lineCounter, parser, publicApi, visit, $Composer, $Document, $Schema, $YAMLError, $YAMLParseError, $YAMLWarning, $Alias, $isAlias, $isCollection, $isDocument, $isMap, $isNode, $isPair, $isScalar, $isSeq, $Pair, $Scalar, $YAMLMap, $YAMLSeq, $Lexer, $LineCounter, $Parser, $parse, $parseAllDocuments, $parseDocument, $stringify, $visit, $visitAsync;
-var init_dist2 = __esm(() => {
+var composer, Document, Schema, errors3, Alias, identity, Pair, Scalar, YAMLMap, YAMLSeq, cst, lexer, lineCounter, parser, publicApi, visit, $Composer, $Document, $Schema, $YAMLError, $YAMLParseError, $YAMLWarning, $Alias, $isAlias, $isCollection, $isDocument, $isMap, $isNode, $isPair, $isScalar, $isSeq, $Pair, $Scalar, $YAMLMap, $YAMLSeq, $Lexer, $LineCounter, $Parser, $parse, $parseAllDocuments, $parseDocument, $stringify, $visit, $visitAsync;
+var init_dist = __esm(() => {
   composer = require_composer();
   Document = require_Document();
   Schema = require_Schema();
-  errors4 = require_errors();
+  errors3 = require_errors();
   Alias = require_Alias();
   identity = require_identity();
   Pair = require_Pair();
@@ -21671,9 +21605,9 @@ var init_dist2 = __esm(() => {
   $Composer = composer.Composer;
   $Document = Document.Document;
   $Schema = Schema.Schema;
-  $YAMLError = errors4.YAMLError;
-  $YAMLParseError = errors4.YAMLParseError;
-  $YAMLWarning = errors4.YAMLWarning;
+  $YAMLError = errors3.YAMLError;
+  $YAMLParseError = errors3.YAMLParseError;
+  $YAMLWarning = errors3.YAMLWarning;
   $Alias = Alias.Alias;
   $isAlias = identity.isAlias;
   $isCollection = identity.isCollection;
@@ -21701,7 +21635,7 @@ var init_dist2 = __esm(() => {
 // ../../node_modules/.bun/@gobing-ai+ts-runtime@0.4.39/node_modules/@gobing-ai/ts-runtime/dist/config.js
 var configSchema;
 var init_config = __esm(() => {
-  init_dist2();
+  init_dist();
   init_zod();
   configSchema = exports_external.object({
     app: exports_external.object({
@@ -21810,7 +21744,7 @@ function findProjectRoot(startDir) {
 var init_file_system_node = () => {};
 
 // ../../plugins/generations/voice-gen/src/index.ts
-import { basename, dirname as dirname2, extname, join, resolve } from "path";
+import { basename, dirname as dirname2, extname, join as join2, resolve } from "path";
 import { parseArgs } from "util";
 
 // ../../packages/kk-core/src/kinds.ts
@@ -21887,11 +21821,26 @@ var PluginManifestSchema = exports_external.object({
 init_file_system_node();
 init_config();
 
+// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.67/node_modules/@gobing-ai/ts-utils/dist/output.js
+var defaultStderrTarget;
+function processStream(name) {
+  const proc = globalThis.process;
+  const stream = proc?.[name];
+  if (stream === undefined) {
+    throw new Error(`No ${name} target available: set one via setDefaultOutputTargets or pass an explicit target`);
+  }
+  return stream;
+}
+function writeLine(message, target) {
+  target.write(`${message}
+`);
+}
+function echoError(message, target = defaultStderrTarget ?? processStream("stderr")) {
+  writeLine(message, target);
+}
+
 // ../../packages/kk-core/src/render-md.ts
 if (false) {}
-// ../../plugins/generations/voice-gen/src/index.ts
-init_dist();
-
 // ../../plugins/generations/voice-gen/src/concat.ts
 function parseWav(buffer) {
   if (buffer.length < 44) {
@@ -22007,21 +21956,21 @@ function concatWavs(wavBuffers, gapsMs = []) {
   }
   const headerSize = 44;
   const totalFileSize = headerSize + totalDataLength;
-  const output2 = new Uint8Array(totalFileSize);
-  const view = new DataView(output2.buffer, output2.byteOffset, output2.byteLength);
-  output2[0] = 82;
-  output2[1] = 73;
-  output2[2] = 70;
-  output2[3] = 70;
+  const output = new Uint8Array(totalFileSize);
+  const view = new DataView(output.buffer, output.byteOffset, output.byteLength);
+  output[0] = 82;
+  output[1] = 73;
+  output[2] = 70;
+  output[3] = 70;
   view.setUint32(4, 36 + totalDataLength, true);
-  output2[8] = 87;
-  output2[9] = 65;
-  output2[10] = 86;
-  output2[11] = 69;
-  output2[12] = 102;
-  output2[13] = 109;
-  output2[14] = 116;
-  output2[15] = 32;
+  output[8] = 87;
+  output[9] = 65;
+  output[10] = 86;
+  output[11] = 69;
+  output[12] = 102;
+  output[13] = 109;
+  output[14] = 116;
+  output[15] = 32;
   view.setUint32(16, 16, true);
   view.setUint16(20, first.audioFormat, true);
   view.setUint16(22, first.numChannels, true);
@@ -22029,17 +21978,44 @@ function concatWavs(wavBuffers, gapsMs = []) {
   view.setUint32(28, first.byteRate, true);
   view.setUint16(32, first.blockAlign, true);
   view.setUint16(34, first.bitsPerSample, true);
-  output2[36] = 100;
-  output2[37] = 97;
-  output2[38] = 116;
-  output2[39] = 97;
+  output[36] = 100;
+  output[37] = 97;
+  output[38] = 116;
+  output[39] = 97;
   view.setUint32(40, totalDataLength, true);
   let currentOffset = headerSize;
   for (const piece of pieces) {
-    output2.set(piece, currentOffset);
+    output.set(piece, currentOffset);
     currentOffset += piece.length;
   }
-  return output2;
+  return output;
+}
+function applySpeed(wav, speed) {
+  if (speed === 1)
+    return wav;
+  const proc = Bun.spawnSync([
+    "ffmpeg",
+    "-hide_banner",
+    "-loglevel",
+    "error",
+    "-f",
+    "wav",
+    "-i",
+    "pipe:0",
+    "-filter:a",
+    `atempo=${speed}`,
+    "-f",
+    "wav",
+    "pipe:1"
+  ], {
+    stdin: Buffer.from(wav),
+    stdout: "pipe",
+    stderr: "pipe"
+  });
+  if (proc.exitCode !== 0 || !proc.stdout) {
+    throw new Error(`ffmpeg atempo failed (exit ${proc.exitCode}): ${proc.stderr?.toString().trim()}`);
+  }
+  return new Uint8Array(proc.stdout);
 }
 
 // ../../plugins/generations/voice-gen/src/mp3.ts
@@ -22437,6 +22413,60 @@ function createVoiceboxClient(options = {}) {
   };
 }
 
+// ../../plugins/generations/voice-gen/src/voicebox-server.ts
+import { spawn as nodeSpawn } from "child_process";
+import { existsSync as existsSync2, openSync } from "fs";
+import { homedir } from "os";
+import { join } from "path";
+var BOOT_WAIT_MS = 30000;
+var POLL_MS = 500;
+var LOG_PATH = "/tmp/voicebox-server.log";
+async function probeVoiceboxHealth(url2, customFetch) {
+  try {
+    return (await customFetch(`${url2}/health`)).ok;
+  } catch {
+    return false;
+  }
+}
+async function ensureVoiceboxRunning(options = {}) {
+  const rawUrl = options.url ?? process.env.VOICEBOX_URL ?? "http://127.0.0.1:17493";
+  const customFetch = options.fetch ?? fetch;
+  const sleep = options.sleep ?? ((ms) => new Promise((r) => setTimeout(r, ms)));
+  const customSpawn = options.spawn ?? nodeSpawn;
+  if (await probeVoiceboxHealth(rawUrl, customFetch)) {
+    return;
+  }
+  const url2 = new URL(rawUrl);
+  const local = ["127.0.0.1", "localhost", "::1"].includes(url2.hostname);
+  if (!local) {
+    throw new Error(`voice-gen: Voicebox not reachable at ${rawUrl} \u2014 auto-start only applies to local URLs; is the remote server up? (fix VOICEBOX_URL)`);
+  }
+  const serverBin = options.serverBin ?? process.env.VOICEBOX_SERVER_BIN ?? "/Applications/Voicebox.app/Contents/MacOS/voicebox-server";
+  if (!existsSync2(serverBin)) {
+    throw new Error(`voice-gen: Voicebox not reachable at ${rawUrl} and VOICEBOX_SERVER_BIN is missing (${serverBin}) \u2014 start Voicebox.app or set VOICEBOX_SERVER_BIN`);
+  }
+  const dataDir = options.dataDir ?? process.env.VOICEBOX_DATA_DIR ?? join(homedir(), "Library/Application Support/sh.voicebox.app");
+  const logPath = options.logPath ?? LOG_PATH;
+  const port = url2.port === "" ? 17493 : Number(url2.port);
+  echoError(`voice-gen: Voicebox not reachable at ${rawUrl} \u2014 auto-starting ${serverBin} (log: ${logPath})`);
+  const child = customSpawn(serverBin, ["--port", String(port), "--data-dir", dataDir], {
+    detached: true,
+    stdio: ["ignore", "ignore", openSync(logPath, "a")]
+  });
+  child.unref?.();
+  const bootWaitMs = options.bootWaitMs ?? BOOT_WAIT_MS;
+  const pollMs = options.pollMs ?? POLL_MS;
+  const deadline = Date.now() + bootWaitMs;
+  while (Date.now() < deadline) {
+    await sleep(pollMs);
+    if (await probeVoiceboxHealth(rawUrl, customFetch)) {
+      echoError(`voice-gen: Voicebox ready at ${rawUrl}`);
+      return;
+    }
+  }
+  throw new Error(`voice-gen: Voicebox did not become healthy at ${rawUrl} within ${bootWaitMs}ms of auto-start \u2014 check ${logPath}`);
+}
+
 // ../../plugins/generations/voice-gen/src/voicescript.ts
 var VOICEBOX_TEXT_MAX = 50000;
 var VOICEBOX_INSTRUCT_MAX = 500;
@@ -22817,6 +22847,11 @@ function validateVoiceScript(script) {
         throw new Error("gap_ms must be non-negative");
       }
     }
+    if (segment.speed !== undefined) {
+      if (typeof segment.speed !== "number" || segment.speed < 0.5 || segment.speed > 2) {
+        throw new Error("speed must be between 0.5 and 2.0");
+      }
+    }
     if (segment.seed !== undefined) {
       if (typeof segment.seed !== "number" || !Number.isInteger(segment.seed) || segment.seed < 0) {
         throw new Error("seed must be a non-negative integer");
@@ -22840,8 +22875,8 @@ async function processGeneratorIO(inputPath, outputPath, clientOverride, transco
   const fs = createNodeFileSystem();
   const outDir = dirname2(outputPath);
   const outStem = basename(outputPath, extname(outputPath));
-  const audioPath = resolve(join(outDir, `${outStem}.wav`));
-  const mp3Path = resolve(join(outDir, `${outStem}.mp3`));
+  const audioPath = resolve(join2(outDir, `${outStem}.wav`));
+  const mp3Path = resolve(join2(outDir, `${outStem}.mp3`));
   const wantMp3 = isMp3Requested();
   await fs.deleteFile(outputPath);
   await fs.deleteFile(audioPath);
@@ -22859,6 +22894,9 @@ async function processGeneratorIO(inputPath, outputPath, clientOverride, transco
     return;
   }
   const client = clientOverride ?? createVoiceboxClient();
+  if (!clientOverride) {
+    await ensureVoiceboxRunning();
+  }
   try {
     await client.health();
     const envDefaultProfile = process.env.VOICEBOX_DEFAULT_PROFILE;
@@ -22979,18 +23017,20 @@ async function processGeneratorIO(inputPath, outputPath, clientOverride, transco
         throw lastError instanceof Error ? lastError : new Error(String(lastError ?? "no audio"));
       }
       const gap = segment.gap_ms ?? 0;
-      segmentWavs.push(wavBytes);
+      const speed = segment.speed ?? 1;
+      segmentWavs.push(applySpeed(wavBytes, speed));
       segmentGaps.push(gap);
-      totalDuration += segDuration + gap / 1000;
+      totalDuration += segDuration / speed + gap / 1000;
       segmentMetadata.push({
         generationId,
         profile: resolvedProfile.name,
         speaker: segment.speaker,
         emotion: rawEmotion,
-        duration: segDuration,
+        duration: segDuration / speed,
         verifyRetries,
         transcription,
-        loudnessDip
+        loudnessDip,
+        ...speed !== 1 ? { speed } : {}
       });
       echoError(`voice-gen: segment ${segIndex + 1}/${combinedScript.segments.length} done in ${((Date.now() - segStartedAt) / 1000).toFixed(1)}s${verifyRetries ? ` (${verifyRetries} verify ${verifyRetries === 1 ? "retry" : "retries"})` : ""}`);
     }
@@ -23071,6 +23111,7 @@ export {
   concatWavs,
   computeExpectedDurationRange,
   auditVoiceSegments,
+  applySpeed,
   VOICEBOX_TEXT_MAX,
   VOICEBOX_LANGUAGES,
   VOICEBOX_INSTRUCT_MAX,

package/plugins/generations/voice-gen/src/concat.ts

@@ -218,3 +218,37 @@ export function concatWavs(wavBuffers: Uint8Array[], gapsMs: number[] = []): Uin
 
     return output;
 }
+
+/**
+ * Resample a segment's playback rate with ffmpeg `atempo` (in-memory stdin/stdout).
+ * Voicebox has no rate parameter, so per-segment `speed` is a local post-step; 1.0 skips.
+ */
+export function applySpeed(wav: Uint8Array, speed: number): Uint8Array {
+    if (speed === 1) return wav;
+    const proc = Bun.spawnSync(
+        [
+            'ffmpeg',
+            '-hide_banner',
+            '-loglevel',
+            'error',
+            '-f',
+            'wav',
+            '-i',
+            'pipe:0',
+            '-filter:a',
+            `atempo=${speed}`,
+            '-f',
+            'wav',
+            'pipe:1',
+        ],
+        {
+            stdin: Buffer.from(wav),
+            stdout: 'pipe',
+            stderr: 'pipe',
+        },
+    );
+    if (proc.exitCode !== 0 || !proc.stdout) {
+        throw new Error(`ffmpeg atempo failed (exit ${proc.exitCode}): ${proc.stderr?.toString().trim()}`);
+    }
+    return new Uint8Array(proc.stdout);
+}

package/plugins/generations/voice-gen/src/index.ts

@@ -3,10 +3,11 @@ import { parseArgs } from 'node:util';
 import { type Content, ContentSchema, type Doc, DocListSchema } from '@gobing-ai/kk-core';
 import { createNodeFileSystem } from '@gobing-ai/ts-runtime';
 import { echoError } from '@gobing-ai/ts-utils';
-import { concatWavs } from './concat';
+import { applySpeed, concatWavs } from './concat';
 import { isMp3Requested, type Mp3Transcoder, transcodeWavToMp3 } from './mp3';
 import { auditVoiceSegments, detectLoudnessDip, TRANSCRIPTION_FIDELITY_FLOOR, transcriptionFidelity } from './qc';
 import { createVoiceboxClient, type VoiceboxClient, type VoiceboxGenerateBody } from './voicebox-client';
+import { ensureVoiceboxRunning } from './voicebox-server';
 import {
     mergeVoiceScripts,
     parseDocToVoiceScript,
@@ -71,6 +72,13 @@ export async function processGeneratorIO(
 
     const client = clientOverride ?? createVoiceboxClient();
 
+    // Auto-start the local server when it is down — only on the default client path;
+    // tests inject clientOverride and skip the spawn entirely. With vdriver=ominivoice
+    // (the default) this plugin never runs, so no server is needed at all.
+    if (!clientOverride) {
+        await ensureVoiceboxRunning();
+    }
+
     try {
         // 1. Health check
         await client.health();
@@ -281,20 +289,22 @@ export async function processGeneratorIO(
             }
 
             const gap = segment.gap_ms ?? 0;
+            const speed = segment.speed ?? 1;
 
-            segmentWavs.push(wavBytes);
+            segmentWavs.push(applySpeed(wavBytes, speed));
             segmentGaps.push(gap);
 
-            totalDuration += segDuration + gap / 1000;
+            totalDuration += segDuration / speed + gap / 1000;
             segmentMetadata.push({
                 generationId,
                 profile: resolvedProfile.name,
                 speaker: segment.speaker,
                 emotion: rawEmotion,
-                duration: segDuration,
+                duration: segDuration / speed,
                 verifyRetries,
                 transcription,
                 loudnessDip,
+                ...(speed !== 1 ? { speed } : {}),
             });
             echoError(
                 `voice-gen: segment ${segIndex + 1}/${combinedScript.segments.length} done in ${((Date.now() - segStartedAt) / 1000).toFixed(1)}s${verifyRetries ? ` (${verifyRetries} verify ${verifyRetries === 1 ? 'retry' : 'retries'})` : ''}`,

package/plugins/generations/voice-gen/src/voicebox-server.ts

@@ -0,0 +1,91 @@
+import { type ChildProcess, spawn as nodeSpawn } from 'node:child_process';
+import { existsSync, openSync } from 'node:fs';
+import { homedir } from 'node:os';
+import { join } from 'node:path';
+import { echoError } from '@gobing-ai/ts-utils';
+
+/** Boot wait / poll cadence for a freshly spawned local server (Voicebox cold start ≈ 2-5s). */
+const BOOT_WAIT_MS = 30_000;
+const POLL_MS = 500;
+const LOG_PATH = '/tmp/voicebox-server.log';
+
+export interface VoiceboxServerOptions {
+    url?: string;
+    serverBin?: string;
+    dataDir?: string;
+    logPath?: string;
+    bootWaitMs?: number;
+    pollMs?: number;
+    fetch?: typeof fetch;
+    sleep?: (ms: number) => Promise<void>;
+    spawn?: typeof nodeSpawn;
+}
+
+async function probeVoiceboxHealth(url: string, customFetch: typeof fetch): Promise<boolean> {
+    try {
+        return (await customFetch(`${url}/health`)).ok;
+    } catch {
+        return false;
+    }
+}
+
+/**
+ * Ensure a local Voicebox server is reachable: probe `/health`, and when it is down and the
+ * URL points at this machine, spawn the app's server binary detached (stderr appended to the
+ * log file, survives plugin exit) and poll until healthy — fail loud otherwise. Remote URLs
+ * and a missing binary are configuration errors, not auto-start candidates.
+ */
+export async function ensureVoiceboxRunning(options: VoiceboxServerOptions = {}): Promise<void> {
+    const rawUrl = options.url ?? process.env.VOICEBOX_URL ?? 'http://127.0.0.1:17493';
+    const customFetch = options.fetch ?? fetch;
+    const sleep = options.sleep ?? ((ms: number) => new Promise<void>((r) => setTimeout(r, ms)));
+    const customSpawn = options.spawn ?? nodeSpawn;
+
+    if (await probeVoiceboxHealth(rawUrl, customFetch)) {
+        return;
+    }
+
+    const url = new URL(rawUrl);
+    const local = ['127.0.0.1', 'localhost', '::1'].includes(url.hostname);
+    if (!local) {
+        throw new Error(
+            `voice-gen: Voicebox not reachable at ${rawUrl} — auto-start only applies to local URLs; is the remote server up? (fix VOICEBOX_URL)`,
+        );
+    }
+    const serverBin =
+        options.serverBin ??
+        process.env.VOICEBOX_SERVER_BIN ??
+        '/Applications/Voicebox.app/Contents/MacOS/voicebox-server';
+    if (!existsSync(serverBin)) {
+        throw new Error(
+            `voice-gen: Voicebox not reachable at ${rawUrl} and VOICEBOX_SERVER_BIN is missing (${serverBin}) — start Voicebox.app or set VOICEBOX_SERVER_BIN`,
+        );
+    }
+    const dataDir =
+        options.dataDir ??
+        process.env.VOICEBOX_DATA_DIR ??
+        join(homedir(), 'Library/Application Support/sh.voicebox.app');
+    const logPath = options.logPath ?? LOG_PATH;
+    const port = url.port === '' ? 17493 : Number(url.port);
+
+    echoError(`voice-gen: Voicebox not reachable at ${rawUrl} — auto-starting ${serverBin} (log: ${logPath})`);
+    const child: ChildProcess = customSpawn(serverBin, ['--port', String(port), '--data-dir', dataDir], {
+        detached: true,
+        stdio: ['ignore', 'ignore', openSync(logPath, 'a')],
+    });
+    child.unref?.();
+
+    const bootWaitMs = options.bootWaitMs ?? BOOT_WAIT_MS;
+    const pollMs = options.pollMs ?? POLL_MS;
+    const deadline = Date.now() + bootWaitMs;
+    while (Date.now() < deadline) {
+        await sleep(pollMs);
+        if (await probeVoiceboxHealth(rawUrl, customFetch)) {
+            echoError(`voice-gen: Voicebox ready at ${rawUrl}`);
+            return;
+        }
+    }
+    throw new Error(
+        `voice-gen: Voicebox did not become healthy at ${rawUrl} within ${bootWaitMs}ms of auto-start — check ${logPath}`,
+    );
+}