@isaacthoman/pulpo 0.138.0 → 0.139.0

package/README.md

@@ -72,23 +72,18 @@ microdollars: `inputPriceMicros`/`outputPriceMicros` per million respective toke
 audio minute. Token billing requires SSE usage support. Credentials remain in
 the referenced provider connection.
 
-Manage each voice's optional sample independently:
-
-```bash
-pulpo speech-model preview upload read-aloud coral ./coral.wav
-pulpo speech-model preview download read-aloud coral --output ./coral-preview.wav
-pulpo --yes speech-model preview delete read-aloud coral
-pulpo --yes speech-model delete read-aloud
-```
-
-Uploading again replaces that voice's clip. MP3 and WAV files must be nonempty,
-at most 5 MiB, and no longer than 30 seconds; the server validates the audio.
-Deleting a preview leaves its voice configured. Model updates retain clips for
-unchanged voice IDs and remove clips for deleted voices. Model deletion removes
-its clips. Listening to an uploaded sample does not generate speech or incur
-generation charges.
-
-Speech catalog and preview commands use `/api/management/v1/speech-models` and
+Set each voice's `previewText` in the model document to customize generated
+previews (up to 500 characters). Use `null` to reset to the shared default;
+updates from older clients that omit this field preserve existing overrides.
+Users generate previews from **Preview speech** at the bottom of speech settings,
+using their selected controls and normal speech billing.
+
+`speech-model test-voice <id> <voice> -o clip.mp3` defaults to the configured text;
+`--text` supplies a one-off override. Admin tests incur provider costs without
+charging a user balance. Saved-preview upload/download/delete commands and
+`--save-preview` are retired.
+
+Speech catalog and voice asset commands use `/api/management/v1/speech-models` and
 require a current administrator with `catalog:read` for reads/downloads or
 `catalog:write` for mutations. Older servers without the `speechModels`
 capability report that an upgrade is required. User preferences continue to use
@@ -169,7 +164,7 @@ totals. Active or queued model work must finish before deletion can proceed.
 Use `speech-model preset --adapter mistral` for the Voxtral preset. The
 `provider-voices`, `provider-sample`, `clone`, `watermark`, `test-voice`, and
 `cleanup` subcommands manage provider discovery, private reference uploads,
-repair, looping watermark settings, synthesized previews, and retryable cleanup.
+repair, looping watermark settings, speech tests, and retryable cleanup.
 See [the speech administration guide](../../docs/speech.md) for complete commands,
 upload limits, billing behavior, and FFmpeg setup.
 

package/dist/index.js

@@ -14615,11 +14615,13 @@ var imageGenerationInputSchema = external_exports.object({
 // ../../packages/contracts/dist/speech.js
 var SPEECH_REQUEST_MAX_INPUT_LENGTH = 16384;
 var SPEECH_MAX_INSTRUCTIONS_LENGTH = 4096;
+var SPEECH_MAX_PREVIEW_TEXT_LENGTH = 500;
 var SPEECH_ASSET_MAX_BYTES = 10 * 1024 * 1024;
 var speechWatermarkSchema = external_exports.object({ enabled: external_exports.boolean().default(false), volume: external_exports.number().min(0.01).max(1).default(0.15) });
 var speechVoiceSchema = external_exports.object({
   id: external_exports.string().trim().min(1).max(200),
   label: external_exports.string().trim().min(1).max(120),
+  previewText: external_exports.string().trim().max(SPEECH_MAX_PREVIEW_TEXT_LENGTH).transform((text) => text || null).nullable().optional(),
   kind: external_exports.enum(["provider", "cloned"]).optional(),
   watermark: speechWatermarkSchema.optional()
 });
@@ -20800,7 +20802,7 @@ async function runModelTest(input) {
 }
 
 // src/index.ts
-var CLI_VERSION = true ? "0.138.0" : "0.1.0";
+var CLI_VERSION = true ? "0.139.0" : "0.1.0";
 var CLI_BUNDLED = true;
 var commandIo = /* @__PURE__ */ new WeakMap();
 var commandClientFactory = /* @__PURE__ */ new WeakMap();
@@ -21275,33 +21277,6 @@ function createProgram(io = processIo, dependencies = {}) {
   speechModel.command("preset <id>").requiredOption("--provider <id>", "existing provider connection ID").option("--adapter <adapter>", "openai or mistral", "openai").description("Print an editable OpenAI or Voxtral model document without creating it").action((id, options2) => {
     writeOutput(io, speechModelSchema.parse({ ...external_exports.enum(["openai", "mistral"]).parse(options2.adapter) === "mistral" ? VOXTRAL_SPEECH_PRESET : OPENAI_SPEECH_PRESET, id, providerConnectionId: options2.provider }), true);
   });
-  const speechPreview = speechModel.command("preview").description("Manage each voice\u2019s optional MP3/WAV preview");
-  const speechPreviewPath = (id, voice) => `/api/management/v1/speech-models/${encodeURIComponent(id)}/voices/${encodeURIComponent(voice)}/preview`;
-  speechPreview.command("upload <id> <voice> <path>").description("Upload or replace a voice preview (up to 30 seconds and 5 MiB)").action(async (id, voice, path, _options, command) => {
-    const filename = basename(path);
-    const extension = filename.split(".").at(-1)?.toLowerCase();
-    const contentType = extension === "mp3" ? "audio/mpeg" : extension === "wav" ? "audio/wav" : null;
-    if (!contentType) throw new Error("Speech previews must be MP3 or WAV files");
-    const file2 = await stat(path);
-    if (!file2.isFile() || !file2.size || file2.size > 5 * 1024 * 1024) throw new Error("Speech previews must be nonempty files of at most 5 MiB");
-    const bytes = new Uint8Array(await readFile2(path));
-    if (!bytes.length || bytes.length > 5 * 1024 * 1024) throw new Error("Speech previews must be nonempty and at most 5 MiB");
-    const { client } = await clientFor(command);
-    emit(io, command, await client.upload(speechPreviewPath(id, voice), { bytes, filename, contentType }));
-  });
-  speechPreview.command("download <id> <voice>").requiredOption("-o, --output <path>", "destination audio file").action(async (id, voice, options2, command) => {
-    const { client } = await clientFor(command);
-    const result = await client.download(speechPreviewPath(id, voice));
-    await writeFile2(options2.output, result.bytes, { mode: 384 });
-    emit(io, command, { id, voice, output: options2.output });
-  });
-  speechPreview.command("delete <id> <voice>").description("Remove a voice preview without deleting its voice").action(async (id, voice, _options, command) => {
-    const options2 = globalOptions(command);
-    await confirmExact(io, `${id}/${voice}`, Boolean(options2.yes), Boolean(options2.json));
-    const { client } = await clientFor(command);
-    await client.request(speechPreviewPath(id, voice), { method: "DELETE" });
-    emit(io, command, { id, voice, deleted: true });
-  });
   const voicePath = (id, voice) => `/api/management/v1/speech-models/${encodeURIComponent(id)}/voices/${encodeURIComponent(voice)}`;
   speechModel.command("provider-voices <id>").action(async (id, _options, command) => {
     const { client } = await clientFor(command);
@@ -21348,9 +21323,9 @@ function createProgram(io = processIo, dependencies = {}) {
       });
     }
   }
-  speechModel.command("test-voice <id> <voice>").requiredOption("-o, --output <path>").option("--text <text>", "text to synthesize", "Hello. This is a sample of my voice.").option("--save-preview", "save generated speech as the user preview").action(async (id, voice, options2, command) => {
+  speechModel.command("test-voice <id> <voice>").requiredOption("-o, --output <path>").option("--text <text>", "text to synthesize (defaults to the voice preview text)").action(async (id, voice, options2, command) => {
     const { client } = await clientFor(command);
-    const result = await client.download(`${voicePath(id, voice)}/test`, 12e4, { method: "POST", body: { input: options2.text, savePreview: Boolean(options2.savePreview) } });
+    const result = await client.download(`${voicePath(id, voice)}/test`, 12e4, { method: "POST", body: { ...options2.text !== void 0 ? { input: options2.text } : {} } });
     await writeFile2(options2.output, result.bytes, { mode: 384 });
     emit(io, command, { output: options2.output });
   });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@isaacthoman/pulpo",
-  "version": "0.138.0",
+  "version": "0.139.0",
   "description": "Operator-first command-line client for Pulpo",
   "type": "module",
   "bin": {