@kmchai/cortico-world-voice 0.1.0 → 0.2.0

package/src/speech/tags.ts

@@ -5,7 +5,7 @@
  * 情绪标记只是控制注释,合成前从文本里删掉。
  */
 
-/** Ming 3.0 / audiocpp 支持的副语言标签。 */
+/** Ming 3.0 系模型支持的副语言标签。 */
 export const PARALINGUISTIC_TAGS: readonly string[] = [
   'Breathing',
   'Cough',
@@ -57,13 +57,10 @@ export function stripUnknownTags(text: string): string {
 }
 
 /**
- * 自动模式只在支持标签的后端保留标签。`audiocpp` 的 Ming 3.0 支持;
- * OpenAI 兼容服务把它们当正文念,所以删掉。
+ * auto 保留标签,由服务里的模型决定演不演;keep 一律保留;strip 一律删掉。
  */
-export function keepsParalinguistic(backend: string, mode: 'auto' | 'keep' | 'strip'): boolean {
-  if (mode === 'keep') return true;
-  if (mode === 'strip') return false;
-  return backend === 'audiocpp';
+export function keepsParalinguistic(mode: 'auto' | 'keep' | 'strip'): boolean {
+  return mode !== 'strip';
 }
 
 const EMOTION_MARKERS: ReadonlyArray<RegExp> = [

package/src/speech/text.ts

@@ -86,7 +86,7 @@ export function filterSpeech(raw: string, cfg: VoiceConfigSection): FilteredSpee
 
   text = canonicalizeTags(text);
   text = stripUnknownTags(text);
-  if (!keepsParalinguistic(cfg.backend, cfg.paralinguistic)) text = stripParalinguisticTags(text);
+  if (!keepsParalinguistic(cfg.paralinguistic)) text = stripParalinguisticTags(text);
 
   if (cfg.filterEmoji) text = text.replace(EMOJI_RE, '').replace(SYMBOL_RE, '');
 

package/src/tools.ts

@@ -8,7 +8,7 @@ export const VOICE_TOOL_DECLS: readonly ToolSpec[] = [
       'Say a line aloud in your voice through the person\'s speakers, or only synthesize it when silent is true. '
       + 'Use it when the person asks for voice, when this line needs a specific emotion or voice, or when automatic reading is off. '
       + 'It only makes sound; the visible reply still goes through the channel\'s own send tool. '
-      + 'When the backend supports paralinguistic tags, keep them inline in the text: [Laughter], [Sigh], [Breathing], [Uhm], [Question-ei] and the like are performed, not read out.',
+      + 'When the speech service supports paralinguistic tags, keep them inline in the text: [Laughter], [Sigh], [Breathing], [Uhm], [Question-ei] and the like are performed, not read out.',
     tags: ['speak'],
     parameters: {
       type: 'object',
@@ -24,21 +24,6 @@ export const VOICE_TOOL_DECLS: readonly ToolSpec[] = [
       required: ['text'],
     },
   },
-  {
-    name: 'voice_sing',
-    description:
-      'Sing a lyric line through the configured singing plugin. It only makes sound; the visible reply still goes through the channel\'s own send tool. '
-      + 'If no singing plugin is configured, or it is not ready, the call fails instead of speaking the lyrics.',
-    tags: ['speak'],
-    parameters: {
-      type: 'object',
-      additionalProperties: false,
-      properties: {
-        lyrics: { type: 'string', maxLength: 2000, description: 'The lyric line to sing.' },
-      },
-      required: ['lyrics'],
-    },
-  },
   {
     name: 'voice_hush',
     description: 'Stop the voice now: the line playing is cut and queued lines are dropped. It does not affect the visible reply.',

package/src/types.ts

@@ -1,9 +1,7 @@
-/** 共享类型:配置段、合成请求与音频、协议适配器接口。 */
-import type { Logger } from 'cortico/core/types.ts';
+/** 共享类型:配置段、合成请求与音频、服务目录与健康。 */
 
-export type BackendId = 'audiocpp' | 'openai' | 'custom';
 export type AudioFormat = 'wav' | 'mp3';
-/** 情绪怎么发给后端。 */
+/** 情绪怎么发给服务。 */
 export type EmotionMode = 'instruct' | 'categorical' | 'basic' | 'tags' | 'none';
 /** 副语言标签(`[Laughter]` 等)怎么处理。 */
 export type ParalinguisticMode = 'auto' | 'keep' | 'strip';
@@ -13,13 +11,13 @@ export type PlaybackMode = 'local' | 'event' | 'none';
 export interface VoiceConfigSection {
   enabled: boolean;
 
-  // 连接与协议
-  backend: BackendId;
-  baseUrl: string;
+  // 连接
+  /** Kmch AI · 服务 的网关地址。 */
+  serviceUrl: string;
+  /** 服务令牌的环境变量名;空表示服务不校验。 */
+  serviceTokenSecret: string;
   model: string;
   voice: string;
-  /** 密钥的环境变量名;空表示不需要。 */
-  apiKeySecret: string;
   format: AudioFormat;
   timeoutMs: number;
 
@@ -65,28 +63,6 @@ export interface VoiceConfigSection {
   gapMs: number;
   /** 工具回执带音频附件,渠道工具可用 `[blob …]` 句柄发送。 */
   shareAudio: boolean;
-
-  // 本地语音接口
-  httpApi: boolean;
-  httpPort: number;
-  httpToken: string;
-  /** 经 HTTP 接口合成时是否也在本机播放。 */
-  httpPlay: boolean;
-
-  // 唱歌
-  singUrl: string;
-  singToken: string;
-
-  // custom 协议
-  customUrl: string;
-  customMethod: 'POST' | 'GET';
-  /** JSON 对象字符串,值里的 `${KEY}` 换成 apiKeySecret 读到的密钥。 */
-  customHeaders: string;
-  /** JSON 模板;占位符见 README。 */
-  customBody: string;
-  customResponse: 'audio' | 'base64' | 'url';
-  /** base64/url 回应里音频字段的点路径。 */
-  customAudioPath: string;
 }
 
 /** 一次合成。`text` 已经过过滤,不再含控制标记。 */
@@ -107,30 +83,20 @@ export interface TtsAudio {
   seconds?: number;
 }
 
-export interface VoiceOption {
+/** 配置下拉的一项;`value` 写回配置,`label` 显示。 */
+export interface SelectOption {
   value: string;
   label: string;
 }
 
-export interface BackendHealth {
+export type VoiceOption = SelectOption;
+export type ModelOption = SelectOption;
+
+/** `GET /health` 的摘要;`ok` 只表示网关有回应,`ready` 表示至少有一个服务能合成。 */
+export interface ServiceHealth {
   ok: boolean;
+  ready: boolean;
+  status: string;
   detail?: string;
-}
-
-/** 适配器运行时可用的东西。 */
-export interface BackendContext {
-  cfg: VoiceConfigSection;
-  secret(name: string): string;
-  fetch: typeof fetch;
-  log: Logger;
-}
-
-/** 一种语音合成协议。新增协议 = 实现这个接口,再在 registry 里登记。 */
-export interface TtsBackend {
-  readonly id: BackendId;
-  readonly label: string;
-  synthesize(req: TtsRequest, ctx: BackendContext): Promise<TtsAudio>;
-  /** 后端能列出的音色;不支持时返回空数组。 */
-  listVoices(ctx: BackendContext): Promise<VoiceOption[]>;
-  health(ctx: BackendContext): Promise<BackendHealth>;
+  models?: number;
 }