shipmyagent 1.0.229 → 1.0.236

package/bin/extensions/voice/Index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"Index.d.ts","sourceRoot":"","sources":["../../../src/extensions/voice/Index.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAKH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,sCAAsC,CAAC;AA2UtE,eAAO,MAAM,cAAc,EAAE,SAuf5B,CAAC"}

package/bin/extensions/voice/Index.js

@@ -0,0 +1,655 @@
+/**
+ * Voice extension。
+ *
+ * 关键点（中文）
+ * - 统一管理本地语音识别模型：选择、安装、启停、激活模型。
+ * - 通过 `extensions.voice` 落盘配置，作为 chat 渠道接入 STT 的单一事实源。
+ */
+import prompts from "prompts";
+import { VOICE_MODEL_CATALOG, getVoiceModelCatalogItem, resolveVoiceModelId, } from "./runtime/Catalog.js";
+import { dedupeVoiceModelIds, ensureVoiceExtensionConfig, persistShipConfig, resolveVoiceModelsRootDir, toPortableRelativePath, } from "./runtime/ConfigStore.js";
+import { installVoiceModelFromHuggingFace, } from "./runtime/Installer.js";
+import { installVoiceTranscribeDependencies, resolveVoiceRunnersByModels, resolveVoiceStrategyByModel, } from "./runtime/DependencyInstaller.js";
+import { transcribeVoiceAudio } from "./runtime/Transcriber.js";
+function getStringOpt(opts, key) {
+    const value = opts[key];
+    if (typeof value !== "string")
+        return undefined;
+    const trimmed = value.trim();
+    return trimmed || undefined;
+}
+function getBooleanOpt(opts, key, defaultValue) {
+    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;
+}
+function getNumberOpt(opts, key) {
+    const value = opts[key];
+    if (typeof value === "number") {
+        if (!Number.isFinite(value) || Number.isNaN(value))
+            return undefined;
+        return Math.floor(value);
+    }
+    if (typeof value === "string") {
+        const text = value.trim();
+        if (!text)
+            return undefined;
+        const parsed = Number.parseInt(text, 10);
+        if (!Number.isFinite(parsed) || Number.isNaN(parsed)) {
+            throw new Error(`Invalid numeric option "${key}": ${value}`);
+        }
+        return parsed;
+    }
+    return undefined;
+}
+function parseVoiceModelIdOrThrow(input) {
+    const modelId = resolveVoiceModelId(input);
+    if (!modelId) {
+        const supported = VOICE_MODEL_CATALOG.map((item) => item.id).join(", ");
+        throw new Error(`Unsupported voice model: ${input}. Supported: ${supported}`);
+    }
+    return modelId;
+}
+function parseVoiceModelArgs(args) {
+    const parsed = [];
+    for (const arg of args) {
+        const text = String(arg || "").trim();
+        if (!text)
+            continue;
+        parsed.push(parseVoiceModelIdOrThrow(text));
+    }
+    return dedupeVoiceModelIds(parsed);
+}
+async function selectVoiceModelsInteractively(params) {
+    const answer = await prompts({
+        type: "multiselect",
+        name: "selectedModels",
+        message: params.message,
+        hint: "- Space 选择 · Enter 确认",
+        instructions: false,
+        min: 1,
+        choices: VOICE_MODEL_CATALOG.map((item) => ({
+            title: `${item.label}  (${item.description})`,
+            value: item.id,
+            selected: item.id === "SenseVoiceSmall",
+        })),
+    });
+    const selected = Array.isArray(answer.selectedModels)
+        ? answer.selectedModels
+        : [];
+    if (selected.length === 0) {
+        throw new Error("No voice model selected");
+    }
+    return dedupeVoiceModelIds(selected);
+}
+function readInstalledModelIds(raw) {
+    if (!Array.isArray(raw))
+        return [];
+    const parsed = [];
+    for (const item of raw) {
+        const modelId = resolveVoiceModelId(String(item || ""));
+        if (!modelId)
+            continue;
+        parsed.push(modelId);
+    }
+    return dedupeVoiceModelIds(parsed);
+}
+function toVoiceTranscribeJson(value) {
+    if (!value)
+        return null;
+    const out = {};
+    if (typeof value.strategy === "string")
+        out.strategy = value.strategy;
+    if (typeof value.command === "string")
+        out.command = value.command;
+    if (typeof value.timeoutMs === "number")
+        out.timeoutMs = value.timeoutMs;
+    if (typeof value.pythonBin === "string")
+        out.pythonBin = value.pythonBin;
+    if (typeof value.language === "string")
+        out.language = value.language;
+    return out;
+}
+function toVoiceConfigJson(value) {
+    if (!value)
+        return null;
+    const out = {};
+    if (typeof value.enabled === "boolean")
+        out.enabled = value.enabled;
+    if (typeof value.provider === "string")
+        out.provider = value.provider;
+    if (typeof value.activeModel === "string")
+        out.activeModel = value.activeModel;
+    if (typeof value.modelsDir === "string")
+        out.modelsDir = value.modelsDir;
+    if (Array.isArray(value.installedModels)) {
+        out.installedModels = value.installedModels.map((item) => String(item));
+    }
+    const transcribe = toVoiceTranscribeJson(value.transcribe);
+    if (transcribe)
+        out.transcribe = transcribe;
+    return out;
+}
+function toVoiceCatalogJson() {
+    return VOICE_MODEL_CATALOG.map((item) => ({
+        id: item.id,
+        label: item.label,
+        description: item.description,
+        huggingfaceRepo: item.huggingfaceRepo,
+        revision: item.revision,
+    }));
+}
+function toInstallProgressJson(event) {
+    return {
+        stage: event.stage,
+        ...(typeof event.filePath === "string" ? { filePath: event.filePath } : {}),
+        ...(typeof event.index === "number" ? { index: event.index } : {}),
+        ...(typeof event.totalFiles === "number"
+            ? { totalFiles: event.totalFiles }
+            : {}),
+    };
+}
+function toDependencyInstallJson(value) {
+    return {
+        pythonBin: value.pythonBin,
+        runners: value.runners,
+        usedVirtualEnv: value.usedVirtualEnv,
+        ...(typeof value.venvDir === "string" ? { venvDir: value.venvDir } : {}),
+        ...(typeof value.basePythonBin === "string"
+            ? { basePythonBin: value.basePythonBin }
+            : {}),
+        items: value.items.map((item) => ({
+            runner: item.runner,
+            pythonBin: item.pythonBin,
+            args: item.args,
+            packages: item.packages,
+            command: item.command,
+            elapsedMs: item.elapsedMs,
+            ...(typeof item.stdoutTail === "string" ? { stdoutTail: item.stdoutTail } : {}),
+            ...(typeof item.stderrTail === "string" ? { stderrTail: item.stderrTail } : {}),
+        })),
+    };
+}
+export const voiceExtension = {
+    name: "voice",
+    actions: {
+        models: {
+            command: {
+                description: "列出内置语音模型目录",
+                mapInput() {
+                    return {};
+                },
+            },
+            api: {
+                method: "GET",
+            },
+            execute() {
+                return {
+                    success: true,
+                    data: {
+                        models: toVoiceCatalogJson(),
+                    },
+                };
+            },
+        },
+        status: {
+            command: {
+                description: "查看 voice extension 当前配置",
+                mapInput() {
+                    return {};
+                },
+            },
+            api: {
+                method: "GET",
+            },
+            execute(params) {
+                return {
+                    success: true,
+                    data: {
+                        voice: toVoiceConfigJson(params.context.config.extensions?.voice || null),
+                    },
+                };
+            },
+        },
+        on: {
+            command: {
+                description: "启用 voice extension，并可交互选择模型安装",
+                configure(command) {
+                    command
+                        .argument("[models...]")
+                        .option("--models-dir <path>", "模型目录（默认 ~/.ship/models/voice）")
+                        .option("--active-model <modelId>", "激活模型 ID（必须在所选模型中）")
+                        .option("--no-install", "仅写入配置，不执行下载")
+                        .option("--force", "强制覆盖并重下已存在模型文件")
+                        .option("--hf-token <token>", "HuggingFace token（私有/Gated 模型场景）");
+                },
+                async mapInput({ args, opts }) {
+                    const fromArgs = parseVoiceModelArgs(args);
+                    const modelIds = fromArgs.length > 0
+                        ? fromArgs
+                        : await selectVoiceModelsInteractively({
+                            message: "请选择要启用/安装的语音识别模型（可多选）",
+                        });
+                    const activeModel = getStringOpt(opts, "activeModel")
+                        ? parseVoiceModelIdOrThrow(String(getStringOpt(opts, "activeModel")))
+                        : modelIds[0];
+                    if (!modelIds.includes(activeModel)) {
+                        throw new Error(`active-model "${activeModel}" is not in selected models: ${modelIds.join(", ")}`);
+                    }
+                    return {
+                        modelIds,
+                        install: getBooleanOpt(opts, "install", true),
+                        force: getBooleanOpt(opts, "force", false),
+                        modelsDir: getStringOpt(opts, "modelsDir"),
+                        activeModel,
+                        hfToken: getStringOpt(opts, "hfToken"),
+                    };
+                },
+            },
+            async execute(params) {
+                const payload = params.payload;
+                const modelsRootDir = resolveVoiceModelsRootDir({
+                    projectRoot: params.context.rootPath,
+                    modelsDir: payload.modelsDir,
+                });
+                const installResults = [];
+                if (payload.install) {
+                    for (const modelId of payload.modelIds) {
+                        const model = getVoiceModelCatalogItem(modelId);
+                        if (!model) {
+                            throw new Error(`Voice model catalog entry not found: ${modelId}`);
+                        }
+                        const progressEvents = [];
+                        const installed = await installVoiceModelFromHuggingFace({
+                            model,
+                            modelsRootDir,
+                            force: payload.force,
+                            hfToken: payload.hfToken,
+                            onProgress: (event) => {
+                                progressEvents.push(toInstallProgressJson(event));
+                            },
+                        });
+                        installResults.push({
+                            modelId,
+                            modelDir: installed.modelDir,
+                            repoId: installed.repoId,
+                            revision: installed.revision,
+                            downloadedFiles: installed.downloadedFiles,
+                            skippedFiles: installed.skippedFiles,
+                            downloadedFilePaths: installed.downloadedFilePaths,
+                            skippedFilePaths: installed.skippedFilePaths,
+                            progressEvents,
+                        });
+                    }
+                }
+                const voiceConfig = ensureVoiceExtensionConfig(params.context.config);
+                const existingInstalled = readInstalledModelIds(voiceConfig.installedModels);
+                voiceConfig.enabled = true;
+                voiceConfig.provider = "local";
+                voiceConfig.activeModel = payload.activeModel;
+                voiceConfig.modelsDir = toPortableRelativePath(params.context.rootPath, modelsRootDir);
+                voiceConfig.installedModels = dedupeVoiceModelIds([
+                    ...existingInstalled,
+                    ...payload.modelIds,
+                ]);
+                const shipJsonPath = await persistShipConfig({
+                    projectRoot: params.context.rootPath,
+                    config: params.context.config,
+                });
+                params.context.logger.info("Voice extension enabled", {
+                    activeModel: voiceConfig.activeModel,
+                    modelsDir: voiceConfig.modelsDir,
+                    install: payload.install,
+                });
+                return {
+                    success: true,
+                    data: {
+                        shipJsonPath,
+                        voice: toVoiceConfigJson(voiceConfig),
+                        install: payload.install,
+                        installResults,
+                    },
+                };
+            },
+        },
+        init: {
+            command: {
+                description: "从零初始化 voice（启用 + 安装模型 + 自动安装转写依赖）",
+                configure(command) {
+                    command
+                        .argument("[models...]")
+                        .option("--models-dir <path>", "模型目录（默认 ~/.ship/models/voice）")
+                        .option("--active-model <modelId>", "激活模型 ID（必须在所选模型中）")
+                        .option("--no-install-model", "跳过模型下载，仅写入配置")
+                        .option("--no-install-deps", "跳过 Python 转写依赖安装")
+                        .option("--python <bin>", "Python 可执行文件（默认 python3）")
+                        .option("--venv-dir <path>", "PEP 668 回退虚拟环境目录（默认 ~/.ship/venvs/voice）")
+                        .option("--no-pip-upgrade", "安装依赖时不带 pip -U")
+                        .option("--pip-timeout-ms <ms>", "pip 安装超时毫秒（默认 300000）")
+                        .option("--force", "强制覆盖并重下已存在模型文件")
+                        .option("--hf-token <token>", "HuggingFace token（私有/Gated 模型场景）");
+                },
+                mapInput({ args, opts }) {
+                    const fromArgs = parseVoiceModelArgs(args);
+                    const modelIds = fromArgs.length > 0 ? fromArgs : ["SenseVoiceSmall"];
+                    const activeModel = getStringOpt(opts, "activeModel")
+                        ? parseVoiceModelIdOrThrow(String(getStringOpt(opts, "activeModel")))
+                        : modelIds[0];
+                    if (!modelIds.includes(activeModel)) {
+                        throw new Error(`active-model "${activeModel}" is not in selected models: ${modelIds.join(", ")}`);
+                    }
+                    return {
+                        modelIds,
+                        installModel: getBooleanOpt(opts, "installModel", true),
+                        installDeps: getBooleanOpt(opts, "installDeps", true),
+                        force: getBooleanOpt(opts, "force", false),
+                        modelsDir: getStringOpt(opts, "modelsDir"),
+                        activeModel,
+                        hfToken: getStringOpt(opts, "hfToken"),
+                        pythonBin: getStringOpt(opts, "python"),
+                        pipUpgrade: getBooleanOpt(opts, "pipUpgrade", true),
+                        pipTimeoutMs: getNumberOpt(opts, "pipTimeoutMs"),
+                        venvDir: getStringOpt(opts, "venvDir"),
+                    };
+                },
+            },
+            async execute(params) {
+                const payload = params.payload;
+                const modelsRootDir = resolveVoiceModelsRootDir({
+                    projectRoot: params.context.rootPath,
+                    modelsDir: payload.modelsDir,
+                });
+                const installResults = [];
+                if (payload.installModel) {
+                    for (const modelId of payload.modelIds) {
+                        const model = getVoiceModelCatalogItem(modelId);
+                        if (!model) {
+                            throw new Error(`Voice model catalog entry not found: ${modelId}`);
+                        }
+                        const progressEvents = [];
+                        const installed = await installVoiceModelFromHuggingFace({
+                            model,
+                            modelsRootDir,
+                            force: payload.force,
+                            hfToken: payload.hfToken,
+                            onProgress: (event) => {
+                                progressEvents.push(toInstallProgressJson(event));
+                            },
+                        });
+                        installResults.push({
+                            modelId,
+                            modelDir: installed.modelDir,
+                            repoId: installed.repoId,
+                            revision: installed.revision,
+                            downloadedFiles: installed.downloadedFiles,
+                            skippedFiles: installed.skippedFiles,
+                            downloadedFilePaths: installed.downloadedFilePaths,
+                            skippedFilePaths: installed.skippedFilePaths,
+                            progressEvents,
+                        });
+                    }
+                }
+                const voiceConfig = ensureVoiceExtensionConfig(params.context.config);
+                const existingInstalled = readInstalledModelIds(voiceConfig.installedModels);
+                if (!payload.installModel &&
+                    !existingInstalled.includes(payload.activeModel)) {
+                    throw new Error(`active-model "${payload.activeModel}" is not installed. Remove --no-install-model or run "sma voice install ${payload.activeModel}" first.`);
+                }
+                const installedFromInit = payload.installModel ? payload.modelIds : [];
+                voiceConfig.enabled = true;
+                voiceConfig.provider = "local";
+                voiceConfig.activeModel = payload.activeModel;
+                voiceConfig.modelsDir = toPortableRelativePath(params.context.rootPath, modelsRootDir);
+                voiceConfig.installedModels = dedupeVoiceModelIds([
+                    ...existingInstalled,
+                    ...installedFromInit,
+                ]);
+                let dependencyInstall;
+                if (payload.installDeps) {
+                    const runners = resolveVoiceRunnersByModels(payload.modelIds);
+                    const installedDeps = await installVoiceTranscribeDependencies({
+                        pythonBin: payload.pythonBin,
+                        runners,
+                        upgrade: payload.pipUpgrade,
+                        timeoutMs: payload.pipTimeoutMs,
+                        venvDir: payload.venvDir,
+                    });
+                    dependencyInstall = toDependencyInstallJson(installedDeps);
+                    voiceConfig.transcribe = {
+                        ...(voiceConfig.transcribe || {}),
+                        pythonBin: installedDeps.pythonBin,
+                    };
+                }
+                voiceConfig.transcribe = {
+                    ...(voiceConfig.transcribe || {}),
+                    strategy: resolveVoiceStrategyByModel(payload.activeModel),
+                };
+                const shipJsonPath = await persistShipConfig({
+                    projectRoot: params.context.rootPath,
+                    config: params.context.config,
+                });
+                params.context.logger.info("Voice extension initialized", {
+                    activeModel: voiceConfig.activeModel,
+                    modelsDir: voiceConfig.modelsDir,
+                    installModel: payload.installModel,
+                    installDeps: payload.installDeps,
+                });
+                return {
+                    success: true,
+                    data: {
+                        shipJsonPath,
+                        voice: toVoiceConfigJson(voiceConfig),
+                        installModel: payload.installModel,
+                        installDeps: payload.installDeps,
+                        installResults,
+                        ...(dependencyInstall ? { dependencyInstall } : {}),
+                    },
+                };
+            },
+        },
+        off: {
+            command: {
+                description: "关闭 voice extension（保留已安装模型记录）",
+                mapInput() {
+                    return {};
+                },
+            },
+            async execute(params) {
+                const voiceConfig = ensureVoiceExtensionConfig(params.context.config);
+                const previousEnabled = voiceConfig.enabled === true;
+                voiceConfig.enabled = false;
+                const shipJsonPath = await persistShipConfig({
+                    projectRoot: params.context.rootPath,
+                    config: params.context.config,
+                });
+                params.context.logger.info("Voice extension disabled");
+                return {
+                    success: true,
+                    data: {
+                        shipJsonPath,
+                        previousEnabled,
+                        voice: toVoiceConfigJson(voiceConfig),
+                    },
+                };
+            },
+        },
+        install: {
+            command: {
+                description: "安装 voice 模型（不改变 enabled 状态）",
+                configure(command) {
+                    command
+                        .argument("[models...]")
+                        .option("--models-dir <path>", "模型目录（默认 ~/.ship/models/voice）")
+                        .option("--force", "强制覆盖并重下已存在模型文件")
+                        .option("--hf-token <token>", "HuggingFace token（私有/Gated 模型场景）");
+                },
+                async mapInput({ args, opts }) {
+                    const fromArgs = parseVoiceModelArgs(args);
+                    const modelIds = fromArgs.length > 0
+                        ? fromArgs
+                        : await selectVoiceModelsInteractively({
+                            message: "请选择要安装的语音识别模型（可多选）",
+                        });
+                    return {
+                        modelIds,
+                        force: getBooleanOpt(opts, "force", false),
+                        modelsDir: getStringOpt(opts, "modelsDir"),
+                        hfToken: getStringOpt(opts, "hfToken"),
+                    };
+                },
+            },
+            async execute(params) {
+                const payload = params.payload;
+                const modelsRootDir = resolveVoiceModelsRootDir({
+                    projectRoot: params.context.rootPath,
+                    modelsDir: payload.modelsDir,
+                });
+                const installResults = [];
+                for (const modelId of payload.modelIds) {
+                    const model = getVoiceModelCatalogItem(modelId);
+                    if (!model) {
+                        throw new Error(`Voice model catalog entry not found: ${modelId}`);
+                    }
+                    const progressEvents = [];
+                    const installed = await installVoiceModelFromHuggingFace({
+                        model,
+                        modelsRootDir,
+                        force: payload.force,
+                        hfToken: payload.hfToken,
+                        onProgress: (event) => {
+                            progressEvents.push(toInstallProgressJson(event));
+                        },
+                    });
+                    installResults.push({
+                        modelId,
+                        modelDir: installed.modelDir,
+                        repoId: installed.repoId,
+                        revision: installed.revision,
+                        downloadedFiles: installed.downloadedFiles,
+                        skippedFiles: installed.skippedFiles,
+                        downloadedFilePaths: installed.downloadedFilePaths,
+                        skippedFilePaths: installed.skippedFilePaths,
+                        progressEvents,
+                    });
+                }
+                const voiceConfig = ensureVoiceExtensionConfig(params.context.config);
+                const existingInstalled = readInstalledModelIds(voiceConfig.installedModels);
+                voiceConfig.provider = "local";
+                voiceConfig.modelsDir = toPortableRelativePath(params.context.rootPath, modelsRootDir);
+                voiceConfig.installedModels = dedupeVoiceModelIds([
+                    ...existingInstalled,
+                    ...payload.modelIds,
+                ]);
+                if (!voiceConfig.activeModel && payload.modelIds.length > 0) {
+                    voiceConfig.activeModel = payload.modelIds[0];
+                }
+                const shipJsonPath = await persistShipConfig({
+                    projectRoot: params.context.rootPath,
+                    config: params.context.config,
+                });
+                params.context.logger.info("Voice model install completed", {
+                    models: payload.modelIds,
+                    modelsDir: voiceConfig.modelsDir,
+                });
+                return {
+                    success: true,
+                    data: {
+                        shipJsonPath,
+                        voice: toVoiceConfigJson(voiceConfig),
+                        installResults,
+                    },
+                };
+            },
+        },
+        use: {
+            command: {
+                description: "切换 voice active 模型",
+                configure(command) {
+                    command.argument("<modelId>");
+                },
+                mapInput({ args }) {
+                    const modelId = parseVoiceModelIdOrThrow(String(args[0] || ""));
+                    return { modelId };
+                },
+            },
+            async execute(params) {
+                const payload = params.payload;
+                const voiceConfig = ensureVoiceExtensionConfig(params.context.config);
+                const installed = readInstalledModelIds(voiceConfig.installedModels);
+                if (!installed.includes(payload.modelId)) {
+                    return {
+                        success: false,
+                        error: `Model "${payload.modelId}" is not installed. Run "sma voice install ${payload.modelId}" first.`,
+                    };
+                }
+                voiceConfig.provider = "local";
+                voiceConfig.activeModel = payload.modelId;
+                const shipJsonPath = await persistShipConfig({
+                    projectRoot: params.context.rootPath,
+                    config: params.context.config,
+                });
+                params.context.logger.info("Voice active model switched", {
+                    activeModel: payload.modelId,
+                });
+                return {
+                    success: true,
+                    data: {
+                        shipJsonPath,
+                        voice: toVoiceConfigJson(voiceConfig),
+                    },
+                };
+            },
+        },
+        transcribe: {
+            command: {
+                description: "转写本地音频文件（用于联调）",
+                configure(command) {
+                    command
+                        .argument("<audioPath>")
+                        .option("--language <code>", "语言提示（可选，例如 zh / en）");
+                },
+                mapInput({ args, opts }) {
+                    const audioPath = String(args[0] || "").trim();
+                    if (!audioPath) {
+                        throw new Error("audioPath is required");
+                    }
+                    return {
+                        audioPath,
+                        language: getStringOpt(opts, "language"),
+                    };
+                },
+            },
+            api: {
+                method: "POST",
+            },
+            async execute(params) {
+                const payload = params.payload;
+                const result = await transcribeVoiceAudio({
+                    runtime: params.context,
+                    audioPath: payload.audioPath,
+                    language: payload.language,
+                });
+                return {
+                    success: true,
+                    data: {
+                        text: result.text,
+                        modelId: result.modelId,
+                        provider: result.provider,
+                        elapsedMs: result.elapsedMs,
+                        runner: result.runner,
+                        audioPath: result.audioPath,
+                    },
+                };
+            },
+        },
+    },
+};
+//# sourceMappingURL=Index.js.map