@downcity/agent 1.1.35 → 1.1.43

package/src/plugin/builtins/tts/types/TtsPlugin.ts

@@ -1,164 +0,0 @@
-/**
- * TTS Plugin 类型定义。
- *
- * 关键点（中文）
- * - TTS 的行为配置与本地模型依赖统一收敛到 `plugins.tts`。
- * - Console 只需要理解 plugin setup，不需要理解底层 asset 细节。
- */
-
-import type { JsonValue } from "@/types/common/Json.js";
-import type { TtsAudioFormat, TtsProvider } from "@/plugin/builtins/tts/types/Tts.js";
-
-/**
- * TTS Plugin 配置。
- */
-export interface TtsPluginConfig {
-  /**
-   * 兼容统一结构化配置约束的索引签名。
-   */
-  [key: string]: JsonValue | undefined;
-  /**
-   * 语音合成 provider 类型。
-   *
-   * 说明（中文）
-   * - 当前固定为 `local`。
-   */
-  provider?: TtsProvider;
-  /**
-   * 当前激活的本地模型 ID。
-   */
-  modelId?: string;
-  /**
-   * 本地模型根目录（可选）。
-   */
-  modelsDir?: string;
-  /**
-   * Python 可执行文件（可选）。
-   */
-  pythonBin?: string;
-  /**
-   * 默认语言提示（可选）。
-   *
-   * 说明（中文）
-   * - `auto` / `zh` / `en` 等简短值即可。
-   */
-  language?: string;
-  /**
-   * 默认音色 ID（可选）。
-   *
-   * 说明（中文）
-   * - 主要给运行时做覆盖；Console 默认不要求手动填写。
-   */
-  voice?: string;
-  /**
-   * 默认输出格式。
-   */
-  format?: TtsAudioFormat;
-  /**
-   * 默认语速倍率。
-   */
-  speed?: number;
-  /**
-   * 输出目录。
-   *
-   * 说明（中文）
-   * - 可为相对项目根目录或绝对路径。
-   * - 为空时默认写到 `.downcity/.cache/tts/`。
-   */
-  outputDir?: string;
-  /**
-   * 单次合成超时时间（毫秒，可选）。
-   */
-  timeoutMs?: number;
-  /**
-   * 已安装模型列表（可选）。
-   *
-   * 说明（中文）
-   * - 仅作为状态快照使用，不应作为真实安装判断唯一依据。
-   */
-  installedModels?: string[];
-}
-
-/**
- * TTS 安装输入。
- */
-export interface TtsInstallInput {
-  /**
-   * 兼容统一结构化配置约束的索引签名。
-   */
-  [key: string]: JsonValue | undefined;
-  /**
-   * 需要安装的模型 ID 列表（可选）。
-   */
-  modelIds?: string[];
-  /**
-   * 安装完成后激活的模型 ID（可选）。
-   */
-  activeModel?: string;
-  /**
-   * 是否强制覆盖已存在资源。
-   */
-  force?: boolean;
-  /**
-   * 模型目录（可选）。
-   */
-  modelsDir?: string;
-  /**
-   * Python 可执行文件（可选）。
-   */
-  pythonBin?: string;
-  /**
-   * 是否同时安装 Python 依赖。
-   */
-  installDeps?: boolean;
-  /**
-   * HuggingFace Token（可选）。
-   */
-  hfToken?: string;
-  /**
-   * 安装完成后默认输出格式（可选）。
-   */
-  format?: TtsAudioFormat;
-}
-
-/**
- * TTS 合成输入。
- */
-export interface TtsSynthesizeInput {
-  /**
-   * 兼容统一结构化配置约束的索引签名。
-   */
-  [key: string]: JsonValue | undefined;
-  /**
-   * 需要合成的文本内容。
-   */
-  text?: string;
-  /**
-   * 覆盖当前使用的模型 ID。
-   */
-  modelId?: string;
-  /**
-   * 覆盖当前语言提示（可选）。
-   */
-  language?: string;
-  /**
-   * 覆盖当前音色（可选）。
-   */
-  voice?: string;
-  /**
-   * 覆盖当前输出格式。
-   */
-  format?: TtsAudioFormat;
-  /**
-   * 覆盖当前语速倍率。
-   */
-  speed?: number;
-  /**
-   * 覆盖当前输出路径或目录。
-   *
-   * 说明（中文）
-   * - 若传入带扩展名的路径，则直接写该文件。
-   * - 若传入目录，则按默认命名生成文件。
-   */
-  output?: string;
-}

package/src/plugin/builtins/voice/Config.ts

@@ -1,135 +0,0 @@
-/**
- * Voice plugin 配置读写工具。
- *
- * 关键点（中文）
- * - 行为配置与转写依赖配置统一收敛到 `plugins.voice`。
- * - 写入前将输入规范化为 JSON object，避免 undefined 进入项目配置。
- */
-
-import type { VoicePluginConfig } from "@/plugin/builtins/voice/types/VoicePlugin.js";
-import type { AgentPluginConfigRuntime } from "@/types/runtime/host/AgentHost.js";
-import type { JsonObject, JsonValue } from "@/types/common/Json.js";
-
-/**
- * 将普通对象转为可持久化 JSON object。
- */
-export function toJsonObject(
-  input: Record<string, unknown> | null | undefined,
-): JsonObject | null {
-  if (!input) return null;
-  const out: JsonObject = {};
-  for (const [key, value] of Object.entries(input)) {
-    if (value === undefined) continue;
-    if (
-      value === null ||
-      typeof value === "string" ||
-      typeof value === "number" ||
-      typeof value === "boolean"
-    ) {
-      out[key] = value;
-      continue;
-    }
-    if (Array.isArray(value)) {
-      out[key] = value
-        .filter((item) => item !== undefined)
-        .map((item) => item as JsonValue);
-      continue;
-    }
-    if (typeof value === "object") {
-      out[key] = toJsonObject(value as Record<string, unknown>) || {};
-    }
-  }
-  return out;
-}
-
-/**
- * 读取字符串选项。
- */
-export function getStringOpt(
-  opts: Record<string, JsonValue>,
-  key: string,
-): string | undefined {
-  const value = opts[key];
-  if (typeof value !== "string") return undefined;
-  const trimmed = value.trim();
-  return trimmed || undefined;
-}
-
-/**
- * 读取布尔选项。
- */
-export function getBooleanOpt(
-  opts: Record<string, JsonValue>,
-  key: string,
-  defaultValue: boolean,
-): boolean {
-  const value = opts[key];
-  if (typeof value === "boolean") return value;
-  if (typeof value === "string") {
-    const normalized = value.trim().toLowerCase();
-    if (["1", "true", "yes", "y", "on"].includes(normalized)) return true;
-    if (["0", "false", "no", "n", "off"].includes(normalized)) return false;
-  }
-  return defaultValue;
-}
-
-/**
- * 读取 Voice Plugin 配置。
- */
-export function readVoicePluginConfig(runtime: {
-  config: {
-    plugins?: Record<string, unknown>;
-  };
-}): VoicePluginConfig {
-  const current = runtime.config.plugins?.voice;
-  const normalized =
-    current && typeof current === "object" && !Array.isArray(current)
-      ? (current as VoicePluginConfig)
-      : {};
-  return {
-    enabled: normalized.enabled === true,
-    injectPrompt:
-      typeof normalized.injectPrompt === "boolean"
-        ? normalized.injectPrompt
-        : true,
-    augmentMessage:
-      typeof normalized.augmentMessage === "boolean"
-        ? normalized.augmentMessage
-        : true,
-    provider:
-      normalized.provider === "command" || normalized.provider === "local"
-        ? normalized.provider
-        : "local",
-    ...(typeof normalized.modelId === "string" ? { modelId: normalized.modelId } : {}),
-    ...(typeof normalized.modelsDir === "string" ? { modelsDir: normalized.modelsDir } : {}),
-    ...(typeof normalized.pythonBin === "string" ? { pythonBin: normalized.pythonBin } : {}),
-    ...(typeof normalized.command === "string" ? { command: normalized.command } : {}),
-    ...(typeof normalized.language === "string" ? { language: normalized.language } : {}),
-    ...(typeof normalized.timeoutMs === "number" ? { timeoutMs: normalized.timeoutMs } : {}),
-    ...(typeof normalized.strategy === "string" ? { strategy: normalized.strategy } : {}),
-    ...(Array.isArray(normalized.installedModels)
-      ? { installedModels: normalized.installedModels }
-      : {}),
-  };
-}
-
-/**
- * 写入完整 voice plugin 配置。
- */
-export async function writeVoicePluginConfig(params: {
-  agentState: {
-    config: {
-      plugins?: Record<string, unknown>;
-    };
-    pluginConfig: AgentPluginConfigRuntime;
-  };
-  value: VoicePluginConfig;
-}): Promise<void> {
-  if (!params.agentState.config.plugins) {
-    params.agentState.config.plugins = {};
-  }
-  params.agentState.config.plugins.voice = (toJsonObject(params.value) || {}) as JsonObject;
-  await params.agentState.pluginConfig.persistProjectPlugins(
-    params.agentState.config.plugins as Record<string, JsonObject>,
-  );
-}

package/src/plugin/builtins/voice/Dependency.ts

@@ -1,329 +0,0 @@
-/**
- * Voice Plugin Dependency Helper。
- *
- * 关键点（中文）
- * - voice 的转写依赖现在内聚在 plugin 内部，不再暴露独立依赖端口心智。
- * - 这里统一管理检查、安装、配置读写与转写句柄解析。
- */
-
-import { execFile as execFileCb } from "node:child_process";
-import { promisify } from "node:util";
-import type { JsonObject, JsonValue } from "@/types/common/Json.js";
-import type { PluginCommandContext } from "@/plugin/types/Plugin.js";
-import type {
-  VoicePluginConfig,
-  VoiceTranscriberConfig,
-  VoiceTranscriberInstallInput,
-  VoiceTranscriberHandle,
-} from "@/plugin/builtins/voice/types/VoicePlugin.js";
-import { transcribeVoiceAudio } from "@/plugin/builtins/voice/runtime/Transcriber.js";
-import {
-  VOICE_MODEL_CATALOG,
-  resolveVoiceModelId,
-} from "@/plugin/builtins/voice/runtime/Catalog.js";
-import {
-  detectLocalVoiceModelInstallState,
-  installVoiceModelFromHuggingFace,
-} from "@/plugin/builtins/voice/runtime/Installer.js";
-import {
-  installVoiceTranscribeDependencies,
-  resolveVoiceRunnersByModels,
-  resolveVoiceStrategyByModel,
-} from "@/plugin/builtins/voice/runtime/DependencyInstaller.js";
-import { resolveVoiceModelsRootDir } from "@/plugin/builtins/voice/runtime/Paths.js";
-import type { VoiceModelId } from "@/plugin/builtins/voice/types/Voice.js";
-
-const execFileAsync = promisify(execFileCb);
-
-export interface VoiceDependencyCheckResult {
-  /**
-   * 当前依赖是否可用。
-   */
-  available: boolean;
-  /**
-   * 不可用原因列表。
-   */
-  reasons: string[];
-  /**
-   * 结构化附加数据（可选）。
-   */
-  details?: JsonValue;
-}
-
-export interface VoiceDependencyInstallResult {
-  /**
-   * 安装是否成功。
-   */
-  success: boolean;
-  /**
-   * 人类可读消息（可选）。
-   */
-  message?: string;
-  /**
-   * 结构化附加数据（可选）。
-   */
-  details?: JsonValue;
-}
-
-function readVoicePluginRecord(context: PluginCommandContext): Record<string, unknown> {
-  const current = context.config.plugins?.voice;
-  if (!current || typeof current !== "object" || Array.isArray(current)) {
-    return {};
-  }
-  return current as Record<string, unknown>;
-}
-
-function toJsonObject(input: Record<string, unknown> | null | undefined): JsonObject {
-  const out: JsonObject = {};
-  if (!input) return out;
-  for (const [key, value] of Object.entries(input)) {
-    if (value === undefined) continue;
-    if (
-      value === null ||
-      typeof value === "string" ||
-      typeof value === "number" ||
-      typeof value === "boolean"
-    ) {
-      out[key] = value;
-      continue;
-    }
-    if (Array.isArray(value)) {
-      out[key] = value
-        .filter((item) => item !== undefined)
-        .map((item) => item as JsonValue);
-      continue;
-    }
-    if (typeof value === "object") {
-      out[key] = toJsonObject(value as Record<string, unknown>);
-    }
-  }
-  return out;
-}
-
-/**
- * 读取 voice 转写依赖配置。
- */
-export function readVoiceTranscriberConfig(
-  context: PluginCommandContext,
-): VoiceTranscriberConfig {
-  const current = readVoicePluginRecord(context);
-  return {
-    provider: "local",
-    ...(current as VoiceTranscriberConfig),
-  };
-}
-
-/**
- * 写入 voice 转写依赖配置。
- */
-export async function writeVoiceTranscriberConfig(params: {
-  context: PluginCommandContext;
-  value: Partial<VoiceTranscriberConfig>;
-}): Promise<VoicePluginConfig> {
-  if (!params.context.config.plugins) {
-    params.context.config.plugins = {};
-  }
-  const current = readVoicePluginRecord(params.context);
-  const next = {
-    ...current,
-    ...params.value,
-  };
-  params.context.config.plugins.voice = toJsonObject(next);
-  await params.context.pluginConfig.persistProjectPlugins(params.context.config.plugins);
-  return params.context.config.plugins.voice as VoicePluginConfig;
-}
-
-async function checkLocalProviderAvailability(
-  context: PluginCommandContext,
-  config: VoiceTranscriberConfig,
-): Promise<VoiceDependencyCheckResult> {
-  const reasons: string[] = [];
-  const modelId = resolveVoiceModelId(String(config.modelId || ""));
-  if (!modelId) {
-    reasons.push("voice transcriber modelId is missing");
-    return {
-      available: false,
-      reasons,
-    };
-  }
-
-  const modelsRootDir = resolveVoiceModelsRootDir({
-    projectRoot: context.rootPath,
-    modelsDir: config.modelsDir,
-  });
-  const installState = await detectLocalVoiceModelInstallState({
-    modelId,
-    modelsRootDir,
-  });
-  if (!installState.installed) {
-    reasons.push(`voice model is not installed: ${modelId}`);
-  }
-
-  const pythonBin = String(config.pythonBin || "").trim() || "python3";
-  try {
-    await execFileAsync(pythonBin, ["--version"], {
-      timeout: 15_000,
-      maxBuffer: 1024 * 1024,
-    });
-  } catch (error) {
-    reasons.push(`python runtime missing: ${String(error)}`);
-  }
-
-  return {
-    available: reasons.length === 0,
-    reasons,
-  };
-}
-
-/**
- * 检查 voice 转写依赖可用性。
- */
-export async function checkVoiceTranscriber(
-  context: PluginCommandContext,
-): Promise<VoiceDependencyCheckResult> {
-  const config = readVoiceTranscriberConfig(context);
-  if (config.provider === "command") {
-    const command = String(config.command || "").trim();
-    return {
-      available: Boolean(command),
-      reasons: command ? [] : ["voice transcriber command is missing"],
-    };
-  }
-  return checkLocalProviderAvailability(context, config);
-}
-
-/**
- * 安装或修复 voice 转写依赖。
- */
-export async function installVoiceTranscriber(params: {
-  context: PluginCommandContext;
-  input?: VoiceTranscriberInstallInput;
-}): Promise<VoiceDependencyInstallResult> {
-  const context = params.context;
-  const input = params.input;
-  const config = readVoiceTranscriberConfig(context);
-  if (config.provider === "command") {
-    return {
-      success: true,
-      message: "voice command provider does not require install",
-    };
-  }
-
-  const requestedModelIds =
-    (Array.isArray(input?.modelIds) ? input?.modelIds : [])
-      .map((item) => resolveVoiceModelId(String(item || "").trim()))
-      .filter((item): item is VoiceModelId => Boolean(item));
-  const selectedModelIds: VoiceModelId[] =
-    requestedModelIds.length > 0
-      ? requestedModelIds
-      : [resolveVoiceModelId(String(config.modelId || "SenseVoiceSmall")) || "SenseVoiceSmall"];
-
-  const modelsRootDir = resolveVoiceModelsRootDir({
-    projectRoot: context.rootPath,
-    modelsDir: input?.modelsDir || config.modelsDir,
-  });
-
-  for (const modelId of selectedModelIds) {
-    const model = VOICE_MODEL_CATALOG.find((item) => item.id === modelId);
-    if (!model) {
-      throw new Error(`Unsupported voice model: ${modelId}`);
-    }
-    const installState = await detectLocalVoiceModelInstallState({
-      modelId,
-      modelsRootDir,
-    });
-    if (!installState.installed || input?.force === true) {
-      await installVoiceModelFromHuggingFace({
-        model,
-        modelsRootDir,
-        force: input?.force === true,
-        hfToken: input?.hfToken,
-      });
-    }
-  }
-
-  if (input?.installDeps !== false) {
-    await installVoiceTranscribeDependencies({
-      pythonBin:
-        String(input?.pythonBin || config.pythonBin || "").trim() || "python3",
-      runners: resolveVoiceRunnersByModels(selectedModelIds),
-    });
-  }
-
-  const activeModelId =
-    resolveVoiceModelId(String(input?.activeModel || "")) ||
-    selectedModelIds[0];
-
-  const nextConfig: VoiceTranscriberConfig = {
-    ...config,
-    provider: "local",
-    modelId: activeModelId,
-    modelsDir: modelsRootDir,
-    pythonBin:
-      String(input?.pythonBin || config.pythonBin || "").trim() || undefined,
-    ...(config.command ? { command: config.command } : {}),
-    language:
-      typeof config.language === "string" && config.language.trim()
-        ? config.language.trim()
-        : undefined,
-    strategy: resolveVoiceStrategyByModel(activeModelId),
-    installedModels: selectedModelIds,
-  };
-  await writeVoiceTranscriberConfig({
-    context,
-    value: nextConfig,
-  });
-
-  return {
-    success: true,
-    message: "voice transcriber installed",
-    details: {
-      modelIds: selectedModelIds,
-      activeModel: activeModelId,
-      modelsRootDir,
-    },
-  };
-}
-
-/**
- * 解析 voice 转写句柄。
- */
-export async function resolveVoiceTranscriber(
-  context: PluginCommandContext,
-): Promise<VoiceTranscriberHandle> {
-  return {
-    async transcribe(input) {
-      try {
-        const result = await transcribeVoiceAudio({
-          context,
-          audioPath: input.audioPath,
-          language: input.language,
-        });
-        return {
-          success: true,
-          text: result.text,
-        };
-      } catch (error) {
-        return {
-          success: false,
-          error: String(error),
-        };
-      }
-    },
-  };
-}
-
-/**
- * 执行一次 voice 转写。
- */
-export async function transcribeWithVoiceDependency(params: {
-  context: PluginCommandContext;
-  audioPath: string;
-  language?: string;
-}): Promise<{ success: boolean; text?: string; error?: string }> {
-  const transcriber = await resolveVoiceTranscriber(params.context);
-  return transcriber.transcribe({
-    audioPath: params.audioPath,
-    ...(params.language ? { language: params.language } : {}),
-  });
-}

package/src/plugin/builtins/voice/InboundAugment.ts

@@ -1,59 +0,0 @@
-/**
- * Voice 入站消息增强 hook。
- *
- * 关键点（中文）
- * - 只处理 voice/audio 附件，转写结果作为 plugin section 注入。
- * - 转写失败不阻塞主链路，保持 best-effort。
- */
-
-import type { AgentContext } from "@/core/AgentContextTypes.js";
-import type { ChatInboundAugmentInput } from "@/plugin/builtins/chat/types/ChatPlugin.js";
-import type { JsonValue } from "@/types/common/Json.js";
-import { transcribeWithVoiceDependency } from "@/plugin/builtins/voice/Dependency.js";
-
-/**
- * 自动转写入站语音附件。
- */
-export async function augmentVoiceInboundMessage(params: {
-  context: AgentContext;
-  value: JsonValue;
-}): Promise<JsonValue> {
-  const input = params.value as unknown as ChatInboundAugmentInput;
-  const voiceAttachments = (Array.isArray(input.attachments) ? input.attachments : []).filter(
-    (item) =>
-      (item.kind === "voice" || item.kind === "audio") &&
-      typeof item.path === "string" &&
-      item.path.trim(),
-  );
-  if (voiceAttachments.length === 0) {
-    return input as unknown as JsonValue;
-  }
-
-  const pluginSections = Array.isArray(input.pluginSections)
-    ? [...input.pluginSections]
-    : [];
-
-  for (const attachment of voiceAttachments) {
-    try {
-      const result = await transcribeWithVoiceDependency({
-        context: params.context,
-        audioPath: String(attachment.path || "").trim(),
-      });
-      const text = typeof result.text === "string" ? result.text.trim() : "";
-      if (!text) continue;
-
-      const absPath = String(attachment.path || "").trim();
-      const rel = absPath.startsWith(`${params.context.rootPath}/`)
-        ? absPath.slice(params.context.rootPath.length + 1)
-        : absPath;
-      pluginSections.push(`【语音转写 ${attachment.kind}: ${rel}】\n${text}`);
-    } catch {
-      // 关键点（中文）：转写失败不阻塞主链路，保持 best-effort。
-    }
-  }
-
-  return {
-    ...input,
-    pluginSections,
-  } as unknown as JsonValue;
-}