@kmchai/cortico-world-voice 0.1.0 → 0.2.1

package/src/config.ts

@@ -4,19 +4,19 @@ import { pick, type Language, type LanguageTable } from 'cortico/core/language.t
 import type { VoiceConfigSection } from './types.ts';
 
 export const VOICE_ID = 'voice';
-/** 配置里 `x-options` 用的动态音色表。 */
-export const VOICE_OPTIONS_KIND = 'voice.voices';
+/** World 定义、WebUI 与配置页共用同一个显示名。 */
+export const VOICE_LABEL = 'Kmch AI · 语音';
 
 export const VOICE_DEFAULTS: VoiceConfigSection = {
   enabled: false,
 
-  backend: 'audiocpp',
-  baseUrl: 'http://127.0.0.1:8000',
-  model: 'ming3-tts',
-  voice: '4016',
-  apiKeySecret: '',
+  serviceUrl: 'http://127.0.0.1:9883',
+  serviceTokenSecret: '',
+  model: '',
+  voice: '',
+  modelOptions: {},
   format: 'wav',
-  timeoutMs: 120_000,
+  timeoutMs: 180_000,
 
   emotionMode: 'instruct',
   defaultEmotion: '',
@@ -47,51 +47,39 @@ export const VOICE_DEFAULTS: VoiceConfigSection = {
   chunkMax: 120,
   gapMs: 120,
   shareAudio: false,
-
-  httpApi: false,
-  httpPort: 9883,
-  httpToken: '',
-  httpPlay: false,
-
-  singUrl: '',
-  singToken: '',
-
-  customUrl: '',
-  customMethod: 'POST',
-  customHeaders: '',
-  customBody: '',
-  customResponse: 'audio',
-  customAudioPath: 'audio',
 };
 
 const TEXT = {
   zh: {
-    label: '语音',
+    label: 'Kmch AI · 语音',
     envTitle: '语音 · 环境提示词',
-    lampBackend: '语音服务',
+    lampService: '语音服务',
     lampQueue: '朗读队列',
-    lampApi: '本地接口',
-    lampSinger: '唱歌',
+    servicePage: '服务设置',
+    modelPanel: '模型',
+    modelSaveShape: '保存失败:参数格式不对。',
+    modelSaveModel: '保存失败:模型 id 不能为空。',
+    modelSaveVoice: '保存失败:音色必须是字符串。',
+    modelSaveOptions: '保存失败:模型参数必须是对象。',
+    modelUnknownMethod: '未知的面板方法。',
     connection: '连接',
-    connectionDesc: '语音服务地址与协议。',
-    backend: '协议',
-    baseUrl: '服务地址',
-    model: '模型',
-    voice: '音色',
-    apiKeySecret: '密钥变量名',
-    apiKeySecretDesc: '部署 .env 里的变量名,留空表示不需要。',
+    connectionDesc: '语音服务的 HTTP 地址与令牌。',
+    serviceUrl: '服务地址',
+    serviceUrlDesc: 'Kmch AI · 服务 的网关地址。',
+    serviceTokenSecret: '服务令牌变量名',
+    serviceTokenSecretDesc: '部署 .env 里的变量名,留空表示服务不校验。',
     format: '音频格式',
     timeoutMs: '超时(毫秒)',
     style: '音色与情绪',
     styleDesc: '',
     emotionMode: '情绪控制',
-    emotionModeDesc: 'instruct 折成自然语言;categorical 只发情绪标签;basic 只发语速/音调/音量;tags 只用正文标签;none 不发。',
+    emotionModeDesc: 'instruct 折成自然语言指令;categorical 只发情绪标签;basic 只发语速;tags 只用正文标签;none 不发。',
     defaultEmotion: '默认情绪',
     defaultInstruct: '默认表演指令',
     speed: '语速',
     paralinguistic: '副语言标签',
-    paralinguisticDesc: 'auto 只在支持标签的后端保留;strip 一律从朗读文本里删掉。',
-    triggers: '触发与过滤',
+    paralinguisticDesc: 'auto 只在当前模型支持标签时保留;keep 一律保留;strip 一律从朗读文本里删掉。',
+    triggers: '触发',
     triggersDesc: '',
     autoSpeak: '自动朗读',
     autoProbability: '自动朗读概率',
@@ -106,6 +94,8 @@ const TEXT = {
     explicitWindowSec: '回看秒数',
     minChars: '最短字数',
     maxChars: '最长字数',
+    filter: '文本过滤',
+    filterDesc: '',
     filterCode: '过滤代码块',
     filterUrls: '过滤网址',
     filterMarkdown: '过滤 Markdown',
@@ -114,7 +104,7 @@ const TEXT = {
     filterEmoji: '过滤表情与符号',
     filterSymbols: '额外过滤字符',
     filterSymbolsDesc: '按字面删除,可用逗号/分号/换行分隔。',
-    output: '输出与接口',
+    output: '输出',
     outputDesc: '',
     playback: '播放方式',
     playbackDesc: 'local 用本机播放器;event 发 tts.voice 事件交给别的插件;none 只合成。',
@@ -122,52 +112,37 @@ const TEXT = {
     chunkMax: '单块字数',
     gapMs: '块间停顿(毫秒)',
     shareAudio: '回执带音频附件',
-    httpApi: '开放本地语音接口',
-    httpPort: '接口端口',
-    httpToken: '接口令牌',
-    httpTokenDesc: '留空不校验;只监听 127.0.0.1。',
-    httpPlay: '接口合成也本机播放',
-    sing: '唱歌',
-    singDesc: '接到唱歌插件的 HTTP 接口;留空则 voice_sing 回失败。',
-    singUrl: '唱歌服务地址',
-    singToken: '唱歌令牌',
-    custom: 'custom 协议',
-    customDesc: '自定义 JSON 请求;占位符 {{text}} {{voice}} {{emotion}} {{instruct}} {{model}} {{speed}} {{format}}。',
-    customUrl: '完整地址',
-    customMethod: '方法',
-    customHeaders: '请求头(JSON)',
-    customHeadersDesc: '值里的 ${KEY} 换成 apiKeySecret 读到的密钥。',
-    customBody: '请求体模板(JSON)',
-    customResponse: '音频位置',
-    customAudioPath: '音频字段路径',
   },
   en: {
-    label: 'Voice',
+    label: 'Kmch AI · 语音',
     envTitle: 'Voice · environment prompt',
-    lampBackend: 'Speech service',
+    lampService: 'Speech service',
     lampQueue: 'Speech queue',
-    lampApi: 'Local API',
-    lampSinger: 'Singing',
+    servicePage: 'Service settings',
+    modelPanel: '模型',
+    modelSaveShape: 'Save failed: invalid arguments.',
+    modelSaveModel: 'Save failed: the model id must not be empty.',
+    modelSaveVoice: 'Save failed: voice must be a string.',
+    modelSaveOptions: 'Save failed: model options must be an object.',
+    modelUnknownMethod: 'Unknown panel method.',
     connection: 'Connection',
-    connectionDesc: 'Where the speech service is and which protocol it speaks.',
-    backend: 'Protocol',
-    baseUrl: 'Base URL',
-    model: 'Model',
-    voice: 'Voice',
-    apiKeySecret: 'Secret variable',
-    apiKeySecretDesc: 'Name in the deployment .env; empty means none.',
+    connectionDesc: 'Where the speech service is and which token it expects.',
+    serviceUrl: 'Service URL',
+    serviceUrlDesc: 'Gateway address of Kmch AI · Service.',
+    serviceTokenSecret: 'Service token variable',
+    serviceTokenSecretDesc: 'Name in the deployment .env; empty means the service does not check.',
     format: 'Audio format',
     timeoutMs: 'Timeout (ms)',
     style: 'Voice and emotion',
     styleDesc: '',
     emotionMode: 'Emotion control',
-    emotionModeDesc: 'instruct turns it into natural language; categorical sends a label; basic sends speed/pitch/volume; tags uses inline tags only; none sends nothing.',
+    emotionModeDesc: 'instruct turns it into natural language; categorical sends a label; basic sends speed; tags uses inline tags only; none sends nothing.',
     defaultEmotion: 'Default emotion',
     defaultInstruct: 'Default direction',
     speed: 'Speed',
     paralinguistic: 'Paralinguistic tags',
-    paralinguisticDesc: 'auto keeps them only on backends that support them; strip removes them from the spoken text.',
-    triggers: 'Triggers and filters',
+    paralinguisticDesc: 'auto keeps them only when the current model supports them; keep always keeps; strip removes them from the spoken text.',
+    triggers: 'Triggers',
     triggersDesc: '',
     autoSpeak: 'Automatic speech',
     autoProbability: 'Automatic probability',
@@ -182,6 +157,8 @@ const TEXT = {
     explicitWindowSec: 'Look-back seconds',
     minChars: 'Minimum characters',
     maxChars: 'Maximum characters',
+    filter: 'Text filter',
+    filterDesc: '',
     filterCode: 'Filter code blocks',
     filterUrls: 'Filter URLs',
     filterMarkdown: 'Filter Markdown',
@@ -190,7 +167,7 @@ const TEXT = {
     filterEmoji: 'Filter emoji and symbols',
     filterSymbols: 'Extra filtered characters',
     filterSymbolsDesc: 'Removed literally; comma/semicolon/newline separated.',
-    output: 'Output and API',
+    output: 'Output',
     outputDesc: '',
     playback: 'Playback',
     playbackDesc: 'local uses the built-in player; event emits tts.voice for another plugin; none only synthesizes.',
@@ -198,24 +175,6 @@ const TEXT = {
     chunkMax: 'Characters per chunk',
     gapMs: 'Gap between chunks (ms)',
     shareAudio: 'Attach audio to tool results',
-    httpApi: 'Serve the local speech API',
-    httpPort: 'API port',
-    httpToken: 'API token',
-    httpTokenDesc: 'Empty means no check; loopback only.',
-    httpPlay: 'Play API requests locally too',
-    sing: 'Singing',
-    singDesc: 'HTTP endpoint of a singing plugin; empty makes voice_sing fail.',
-    singUrl: 'Singing service URL',
-    singToken: 'Singing token',
-    custom: 'custom protocol',
-    customDesc: 'Custom JSON request; placeholders {{text}} {{voice}} {{emotion}} {{instruct}} {{model}} {{speed}} {{format}}.',
-    customUrl: 'Full URL',
-    customMethod: 'Method',
-    customHeaders: 'Headers (JSON)',
-    customHeadersDesc: '${KEY} in a value is replaced with the secret named by apiKeySecret.',
-    customBody: 'Body template (JSON)',
-    customResponse: 'Audio location',
-    customAudioPath: 'Audio field path',
   },
 } as const;
 
@@ -238,11 +197,8 @@ export function voiceConfigGroups(language: Language): ConfigGroup[] {
         title: t.connection,
         description: t.connectionDesc,
         properties: {
-          [`${p}backend`]: { type: 'string', title: t.backend, enum: ['audiocpp', 'openai', 'custom'], 'x-hot': true },
-          [`${p}baseUrl`]: { type: 'string', title: t.baseUrl, 'x-hot': true },
-          [`${p}model`]: { type: 'string', title: t.model, 'x-hot': true },
-          [`${p}voice`]: { type: 'string', title: t.voice, 'x-options': VOICE_OPTIONS_KIND, 'x-hot': true },
-          [`${p}apiKeySecret`]: { type: 'string', title: t.apiKeySecret, description: t.apiKeySecretDesc, 'x-hot': false },
+          [`${p}serviceUrl`]: { type: 'string', title: t.serviceUrl, description: t.serviceUrlDesc, 'x-hot': true },
+          [`${p}serviceTokenSecret`]: { type: 'string', title: t.serviceTokenSecret, description: t.serviceTokenSecretDesc, 'x-hot': true },
           [`${p}format`]: { type: 'string', title: t.format, enum: ['wav', 'mp3'], 'x-hot': true },
           [`${p}timeoutMs`]: { type: 'integer', title: t.timeoutMs, minimum: 1000, maximum: 600_000, 'x-hot': true },
         },
@@ -280,6 +236,16 @@ export function voiceConfigGroups(language: Language): ConfigGroup[] {
           [`${p}explicitWindowSec`]: { type: 'integer', title: t.explicitWindowSec, minimum: 0, maximum: 3600, 'x-hot': true },
           [`${p}minChars`]: { type: 'integer', title: t.minChars, minimum: 1, maximum: 100, 'x-hot': true },
           [`${p}maxChars`]: { type: 'integer', title: t.maxChars, minimum: 20, maximum: 8000, 'x-hot': true },
+        },
+      },
+    },
+    {
+      id: 'world:voice-filter',
+      owner: 'world:voice',
+      schema: {
+        type: 'object',
+        title: t.filter,
+        properties: {
           [`${p}filterCode`]: { type: 'boolean', title: t.filterCode, 'x-hot': true },
           [`${p}filterUrls`]: { type: 'boolean', title: t.filterUrls, 'x-hot': true },
           [`${p}filterMarkdown`]: { type: 'boolean', title: t.filterMarkdown, 'x-hot': true },
@@ -301,29 +267,6 @@ export function voiceConfigGroups(language: Language): ConfigGroup[] {
           [`${p}chunkMax`]: { type: 'integer', title: t.chunkMax, minimum: 20, maximum: 500, 'x-hot': true },
           [`${p}gapMs`]: { type: 'integer', title: t.gapMs, minimum: 0, maximum: 5000, 'x-hot': true },
           [`${p}shareAudio`]: { type: 'boolean', title: t.shareAudio, 'x-hot': true },
-          [`${p}httpApi`]: { type: 'boolean', title: t.httpApi, 'x-hot': false },
-          [`${p}httpPort`]: { type: 'integer', title: t.httpPort, minimum: 1, maximum: 65535, 'x-hot': false },
-          [`${p}httpToken`]: { type: 'string', title: t.httpToken, description: t.httpTokenDesc, 'x-hot': false },
-          [`${p}httpPlay`]: { type: 'boolean', title: t.httpPlay, 'x-hot': true },
-          [`${p}singUrl`]: { type: 'string', title: t.singUrl, description: t.singDesc, 'x-hot': true },
-          [`${p}singToken`]: { type: 'string', title: t.singToken, 'x-hot': true },
-        },
-      },
-    },
-    {
-      id: 'world:voice-custom',
-      owner: 'world:voice',
-      schema: {
-        type: 'object',
-        title: t.custom,
-        description: t.customDesc,
-        properties: {
-          [`${p}customUrl`]: { type: 'string', title: t.customUrl, 'x-hot': true },
-          [`${p}customMethod`]: { type: 'string', title: t.customMethod, enum: ['POST', 'GET'], 'x-hot': true },
-          [`${p}customHeaders`]: { type: 'string', title: t.customHeaders, description: t.customHeadersDesc, 'x-hot': true },
-          [`${p}customBody`]: { type: 'string', title: t.customBody, 'x-hot': true },
-          [`${p}customResponse`]: { type: 'string', title: t.customResponse, enum: ['audio', 'base64', 'url'], 'x-hot': true },
-          [`${p}customAudioPath`]: { type: 'string', title: t.customAudioPath, 'x-hot': true },
         },
       },
     },

package/src/definition.ts

@@ -1,15 +1,13 @@
 import { describeTools, type WorldDefinition } from 'cortico/world.ts';
-import { VOICE_DEFAULTS, VOICE_ID, VOICE_OPTIONS_KIND } from './config.ts';
-import { currentVoiceOptions } from './options.ts';
+import { VOICE_DEFAULTS, VOICE_ID, VOICE_LABEL } from './config.ts';
 import { VOICE_TOOL_DECLS } from './tools.ts';
 import type { VoiceConfigSection } from './types.ts';
 import { VoiceWorld } from './world.ts';
 
 export const VOICE: WorldDefinition<VoiceConfigSection> = {
   id: VOICE_ID,
-  label: 'Voice',
-  defaults: () => ({ ...VOICE_DEFAULTS }),
-  configOptions: (kind) => (kind === VOICE_OPTIONS_KIND ? currentVoiceOptions() : []),
+  label: VOICE_LABEL,
+  defaults: () => ({ ...VOICE_DEFAULTS, modelOptions: {} }),
   create: (ctx) => new VoiceWorld({
     cfg: ctx.cfg,
     timezone: ctx.timezone,
@@ -20,7 +18,6 @@ export const VOICE: WorldDefinition<VoiceConfigSection> = {
   describe: {
     tools: describeTools(VOICE_TOOL_DECLS, {
       voice_speak: { zh: '现在念一句;可指定情绪、表演指令、音色与语速,也可以只合成不出声。', en: 'Say one line now, optionally with an emotion, direction, voice and speed, or synthesize it silently.' },
-      voice_sing: { zh: '把一句歌词交给唱歌插件唱出来。', en: 'Sing one lyric line through the configured singing service.' },
       voice_hush: { zh: '立刻停声并清空排队。', en: 'Stop the voice and drop queued lines.' },
       voice_style: { zh: '读或改自动朗读使用的长期音色与情绪。', en: 'Read or change the standing voice style used by automatic reading.' },
     }),

package/src/models.ts

@@ -0,0 +1,66 @@
+/**
+ * 模型能力与提示词片段。
+ *
+ * 服务返回的模型 id 决定两件事:正文里能不能写副语言标签,以及环境提示词
+ * 该怎么描述文本与情绪。按 id 前缀匹配,大小写不敏感;空模型与未知模型按
+ * 通用 OpenAI 兼容模型处理。
+ */
+
+interface ModelInfo {
+  /** 模型 id 的前缀,小写。 */
+  readonly match: readonly string[];
+  /** 给人看的名字;未知模型为空。 */
+  readonly label: string;
+  /** 是否支持 `[Laughter]` 这类副语言标签。 */
+  readonly tags: boolean;
+  /** 环境提示词里的文本与情绪写法。 */
+  readonly prompt: string;
+}
+
+const MING3: ModelInfo = {
+  match: ['ming3', 'ming-3', 'ming_3'],
+  label: 'Ming 3.0',
+  tags: true,
+  prompt: '支持 [Laughter] [Sigh] [Breathing] [Uhm] [Question-ei] 等副语言标签;情绪可用 instruct/emotion.',
+};
+
+const INDEXTTS2: ModelInfo = {
+  match: ['index-tts2', 'index_tts2', 'index-tts2.5', 'index_tts2_5', '2.5'],
+  label: 'IndexTTS2',
+  tags: false,
+  prompt: '情绪用自然语言(emotion_text),不要写 [Laughter] 这类方括号标签;采样/时长参数在语音模型页配置,不要写进正文.',
+};
+
+const GENERIC: ModelInfo = {
+  match: [],
+  label: '',
+  tags: false,
+  prompt: '用 instructions 描述语气;不要用方括号标签,除非确认模型支持.',
+};
+
+const KNOWN: readonly ModelInfo[] = [MING3, INDEXTTS2];
+
+function normalize(model: string): string {
+  return typeof model === 'string' ? model.trim().toLowerCase() : '';
+}
+
+function profileOf(model: string): ModelInfo | null {
+  const id = normalize(model);
+  if (!id) return null;
+  return KNOWN.find((info) => info.match.some((prefix) => id.startsWith(prefix))) ?? null;
+}
+
+/** 当前模型是否支持副语言标签;未知与空模型按通用模型处理,返回 false。 */
+export function modelSupportsTags(model: string): boolean {
+  return profileOf(model)?.tags ?? GENERIC.tags;
+}
+
+/** 给人看的模型名;未知或空模型返回空串,由调用方退回模型 id。 */
+export function modelLabel(model: string): string {
+  return profileOf(model)?.label ?? '';
+}
+
+/** 环境提示词里的文本与情绪写法;空模型与未知模型用通用片段。 */
+export function modelPrompt(model: string): string {
+  return (profileOf(model) ?? GENERIC).prompt;
+}

package/src/panel.ts

@@ -0,0 +1,34 @@
+/**
+ * 「模型」面板的数据形状。服务端经由 `World.webUi().invoke` 构造,浏览器端只
+ * `import type` 这些声明;面板不接触服务地址与令牌。
+ */
+import type { VoiceModel } from './types.ts';
+
+export interface PanelModelState {
+  /** `worlds.voice.model`。 */
+  model: string;
+  /** `worlds.voice.voice`。 */
+  voice: string;
+  /** 当前模型的环境提示词片段,来自 `models.ts`。 */
+  prompt: string;
+  /** 当前模型已保存的请求参数。 */
+  options: Record<string, unknown>;
+  /** 服务目录;服务不可达时是上一次可用的结果。 */
+  models: VoiceModel[];
+  /** 最近一次读取服务目录失败的原因。 */
+  error?: string;
+}
+
+/** `invoke('model','save', [save])` 的请求体。 */
+export interface PanelModelSave {
+  model: string;
+  voice: string;
+  options: Record<string, unknown>;
+}
+
+/** 只认普通对象:数组、null 与类对象都不算。 */
+export function isPlainObject(value: unknown): value is Record<string, unknown> {
+  if (!value || typeof value !== 'object' || Array.isArray(value)) return false;
+  const proto = Object.getPrototypeOf(value);
+  return proto === Object.prototype || proto === null;
+}