@downcity/plugins 1.0.56 → 1.0.59

package/src/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 "@downcity/agent/internal/types/common/Json.js";
-import type { PluginCommandContext } from "@downcity/agent/internal/plugin/types/Plugin.js";
-import type {
-  VoicePluginConfig,
-  VoiceTranscriberConfig,
-  VoiceTranscriberInstallInput,
-  VoiceTranscriberHandle,
-} from "@/voice/types/VoicePlugin.js";
-import { transcribeVoiceAudio } from "@/voice/runtime/Transcriber.js";
-import {
-  VOICE_MODEL_CATALOG,
-  resolveVoiceModelId,
-} from "@/voice/runtime/Catalog.js";
-import {
-  detectLocalVoiceModelInstallState,
-  installVoiceModelFromHuggingFace,
-} from "@/voice/runtime/Installer.js";
-import {
-  installVoiceTranscribeDependencies,
-  resolveVoiceRunnersByModels,
-  resolveVoiceStrategyByModel,
-} from "@/voice/runtime/DependencyInstaller.js";
-import { resolveVoiceModelsRootDir } from "@/voice/runtime/Paths.js";
-import type { VoiceModelId } from "@/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/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 "@downcity/agent/internal/types/runtime/agent/AgentContext.js";
-import type { ChatInboundAugmentInput } from "@/chat/types/ChatPlugin.js";
-import type { JsonValue } from "@downcity/agent/internal/types/common/Json.js";
-import { transcribeWithVoiceDependency } from "@/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;
-}

package/src/voice/ModelCatalog.ts

@@ -1,43 +0,0 @@
-/**
- * Voice 插件模型目录封装。
- *
- * 关键点（中文）
- * - 对 plugin 层暴露稳定的 voice 领域接口。
- * - plugin 不直接依赖底层 runtime 文件布局，只依赖这里的领域封装。
- */
-
-import {
-  VOICE_MODEL_CATALOG,
-  resolveVoiceModelId,
-} from "@/voice/runtime/Catalog.js";
-
-/**
- * 列出内置 voice 模型目录。
- */
-export function listVoiceModels(): Array<{
-  /**
-   * 稳定模型 ID。
-   */
-  id: string;
-  /**
-   * 用户可读标签。
-   */
-  label: string;
-  /**
-   * 用户可读说明。
-   */
-  description: string;
-}> {
-  return VOICE_MODEL_CATALOG.map((item) => ({
-    id: item.id,
-    label: item.label,
-    description: item.description,
-  }));
-}
-
-/**
- * 解析用户输入的 voice 模型 ID。
- */
-export function resolveVoicePluginModelId(input: string): string | null {
-  return resolveVoiceModelId(input);
-}

package/src/voice/runtime/Catalog.ts

@@ -1,68 +0,0 @@
-import type { VoiceModelCatalogItem, VoiceModelId } from "@/voice/types/Voice.js";
-
-/**
- * Voice 内置模型目录。
- *
- * 关键点（中文）
- * - 统一维护 CLI 展示、下载源与配置落盘 ID。
- * - 只保留“稳定字段”，避免目录频繁变动影响外部调用。
- */
-export const VOICE_MODEL_CATALOG: readonly VoiceModelCatalogItem[] = [
-  {
-    id: "SenseVoiceSmall",
-    label: "SenseVoiceSmall",
-    description: "中文优先，体积较小，推荐默认选择。",
-    huggingfaceRepo: "FunAudioLLM/SenseVoiceSmall",
-    revision: "main",
-  },
-  {
-    id: "paraformer-zh-streaming",
-    label: "paraformer-zh-streaming",
-    description: "中文流式场景友好，低延迟。",
-    huggingfaceRepo: "funasr/paraformer-zh-streaming",
-    revision: "main",
-  },
-  {
-    id: "whisper-large-v3-turbo",
-    label: "whisper-large-v3-turbo",
-    description: "多语言能力稳定，泛化较强。",
-    huggingfaceRepo: "openai/whisper-large-v3-turbo",
-    revision: "main",
-  },
-] as const;
-
-const VOICE_MODEL_ALIAS_MAP: Record<string, VoiceModelId> = {
-  sensevoicesmall: "SenseVoiceSmall",
-  "sensevoice-small": "SenseVoiceSmall",
-  "paraformer-zh-streaming": "paraformer-zh-streaming",
-  paraformer_zh_streaming: "paraformer-zh-streaming",
-  "whisper-large-v3-turbo": "whisper-large-v3-turbo",
-  whisper_large_v3_turbo: "whisper-large-v3-turbo",
-};
-
-const CATALOG_BY_ID: Record<VoiceModelId, VoiceModelCatalogItem> =
-  VOICE_MODEL_CATALOG.reduce(
-    (acc, item) => {
-      acc[item.id] = item;
-      return acc;
-    },
-    {} as Record<VoiceModelId, VoiceModelCatalogItem>,
-  );
-
-/**
- * 解析用户输入为稳定模型 ID（支持别名与大小写）。
- */
-export function resolveVoiceModelId(input: string): VoiceModelId | null {
-  const raw = String(input || "").trim();
-  if (!raw) return null;
-  const exact = VOICE_MODEL_CATALOG.find((item) => item.id === raw);
-  if (exact) return exact.id;
-  return VOICE_MODEL_ALIAS_MAP[raw.toLowerCase()] || null;
-}
-
-/**
- * 读取目录项；不存在返回 null。
- */
-export function getVoiceModelCatalogItem(id: VoiceModelId): VoiceModelCatalogItem | null {
-  return CATALOG_BY_ID[id] || null;
-}