@zythum02/fello-server 0.1.52 → 0.1.54

package/out/server/main.js

@@ -58,6 +58,10 @@ var DEFAULT_SETTINGS = {
     muted: false,
     theme: "soft"
   },
+  voiceInput: {
+    altDoublePress: true
+  },
+  shortcuts: {},
   proxy: {
     mode: "off"
   }
@@ -71,6 +75,17 @@ function readSettings() {
     const raw = JSON.parse(readFileSync(settingsPath(), "utf-8"));
     const isObject = (value) => typeof value === "object" && value !== null && !Array.isArray(value);
     const rawObj = isObject(raw) ? raw : null;
+    const shortcuts = (() => {
+      const rawShortcuts = rawObj && isObject(rawObj.shortcuts) ? rawObj.shortcuts : null;
+      if (!rawShortcuts) return {};
+      const next = {};
+      for (const [commandId, value] of Object.entries(rawShortcuts)) {
+        if (Array.isArray(value)) {
+          next[commandId] = value.filter((shortcut) => typeof shortcut === "string").map((shortcut) => shortcut.trim()).filter(Boolean);
+        }
+      }
+      return next;
+    })();
     const rawAgents = rawObj && isObject(rawObj.agents) ? rawObj.agents : null;
     const agents = (() => {
       if (!rawAgents) return DEFAULT_SETTINGS.agents;
@@ -191,6 +206,12 @@ function readSettings() {
         theme: raw2.theme === "soft" || raw2.theme === "crisp" ? raw2.theme : DEFAULT_SETTINGS.sound.theme
       };
     })();
+    const voiceInput = (() => {
+      const raw2 = rawObj && isObject(rawObj.voiceInput) ? rawObj.voiceInput : null;
+      return {
+        altDoublePress: typeof raw2?.altDoublePress === "boolean" ? raw2.altDoublePress : DEFAULT_SETTINGS.voiceInput.altDoublePress
+      };
+    })();
     const proxy = (() => {
       const raw2 = rawObj && isObject(rawObj.proxy) ? rawObj.proxy : null;
       if (!raw2) return DEFAULT_SETTINGS.proxy;
@@ -231,6 +252,25 @@ function readSettings() {
       model: p.model,
       active: typeof p.active === "boolean" ? p.active : false
     })) : [];
+    const speechToText = Array.isArray(rawObj?.speechToText) ? rawObj.speechToText.filter((value) => isObject(value)).filter((value) => {
+      const provider = value.provider;
+      return typeof value.id === "string" && typeof value.name === "string" && typeof value.apiKey === "string" && (provider === "volcengine" || provider === "dashscope" || provider === "openai" || provider === "iflytek");
+    }).map((value) => ({
+      id: value.id,
+      name: value.name,
+      provider: value.provider,
+      apiKey: value.apiKey,
+      appId: typeof value.appId === "string" ? value.appId : void 0,
+      apiSecret: typeof value.apiSecret === "string" ? value.apiSecret : void 0,
+      resourceId: typeof value.resourceId === "string" ? value.resourceId : void 0,
+      model: typeof value.model === "string" ? value.model : void 0,
+      baseUrl: typeof value.baseUrl === "string" ? value.baseUrl : void 0,
+      workspaceId: typeof value.workspaceId === "string" ? value.workspaceId : void 0,
+      region: value.region === "cn-beijing" || value.region === "ap-southeast-1" ? value.region : void 0,
+      workspace: typeof value.workspace === "string" ? value.workspace : void 0,
+      language: typeof value.language === "string" ? value.language : void 0,
+      active: typeof value.active === "boolean" ? value.active : false
+    })) : [];
     const editor = (() => {
       const raw2 = rawObj && isObject(rawObj.editor) ? rawObj.editor : null;
       if (raw2 && typeof raw2.name === "string" && raw2.name.trim()) {
@@ -247,9 +287,12 @@ function readSettings() {
       ilink,
       editor,
       sound,
+      voiceInput,
+      shortcuts,
       proxy,
       snippets,
-      imageGeneration
+      imageGeneration,
+      speechToText
     };
   } catch {
     return DEFAULT_SETTINGS;
@@ -339,6 +382,12 @@ function getSettings() {
       muted: meta.sound.muted,
       theme: meta.sound.theme
     },
+    voiceInput: {
+      altDoublePress: meta.voiceInput.altDoublePress
+    },
+    shortcuts: Object.fromEntries(
+      Object.entries(meta.shortcuts).map(([commandId, shortcuts]) => [commandId, shortcuts.slice()])
+    ),
     proxy: meta.proxy ?? DEFAULT_SETTINGS.proxy,
     snippets: meta.snippets ?? [],
     imageGeneration: (meta.imageGeneration ?? []).map((p) => ({
@@ -351,6 +400,22 @@ function getSettings() {
       extraBody: p.extraBody ?? void 0,
       model: p.model,
       active: p.active
+    })),
+    speechToText: (meta.speechToText ?? []).map((p) => ({
+      id: p.id,
+      name: p.name,
+      provider: p.provider,
+      apiKey: p.apiKey,
+      appId: p.appId,
+      apiSecret: p.apiSecret,
+      resourceId: p.resourceId,
+      model: p.model,
+      baseUrl: p.baseUrl,
+      workspaceId: p.workspaceId,
+      region: p.region,
+      workspace: p.workspace,
+      language: p.language,
+      active: p.active
     }))
   };
 }
@@ -480,6 +545,20 @@ function updateSettings(settings) {
         theme: settings.sound.theme === "soft" || settings.sound.theme === "crisp" ? settings.sound.theme : prevMeta.sound.theme
       };
     })(),
+    voiceInput: (() => {
+      if (!settings.voiceInput) {
+        return prevMeta.voiceInput;
+      }
+      return {
+        altDoublePress: typeof settings.voiceInput.altDoublePress === "boolean" ? settings.voiceInput.altDoublePress : prevMeta.voiceInput.altDoublePress
+      };
+    })(),
+    shortcuts: settings.shortcuts ? Object.fromEntries(
+      Object.entries(settings.shortcuts).map(([commandId, shortcuts]) => [
+        commandId,
+        shortcuts.map((shortcut) => shortcut.trim()).filter(Boolean)
+      ])
+    ) : prevMeta.shortcuts,
     proxy: (() => {
       if (!settings.proxy) {
         return prevMeta.proxy;
@@ -506,7 +585,23 @@ function updateSettings(settings) {
       extraBody: p.extraBody ?? void 0,
       model: p.model,
       active: p.active
-    })) : prevMeta.imageGeneration
+    })) : prevMeta.imageGeneration,
+    speechToText: settings.speechToText ? settings.speechToText.map((p) => ({
+      id: p.id,
+      name: p.name,
+      provider: p.provider,
+      apiKey: p.apiKey,
+      appId: p.appId?.trim() || void 0,
+      apiSecret: p.apiSecret?.trim() || void 0,
+      resourceId: p.resourceId?.trim() || void 0,
+      model: p.model?.trim() || void 0,
+      baseUrl: p.baseUrl?.trim() || void 0,
+      workspaceId: p.workspaceId?.trim() || void 0,
+      region: p.region,
+      workspace: p.workspace?.trim() || void 0,
+      language: p.language?.trim() || void 0,
+      active: p.active
+    })) : prevMeta.speechToText
   };
   writeSettings(meta);
 }
@@ -4949,7 +5044,7 @@ async function executeSubagent(params, subSessionId, taskName, taskPrompt, signa
     }
     return accumulatedText || "[No output produced]";
   } catch (error) {
-    const errorMessage = error instanceof Error ? error.message : String(error);
+    const errorMessage2 = error instanceof Error ? error.message : String(error);
     const conn3 = params.getConnection();
     if (conn3) {
       await sendSubagentStatusUpdate(
@@ -4967,7 +5062,7 @@ async function executeSubagent(params, subSessionId, taskName, taskPrompt, signa
             sessionUpdate: "agent_message_chunk",
             content: { type: "text", text: `
 
-**Error:** ${errorMessage}` }
+**Error:** ${errorMessage2}` }
           }
         });
       }
@@ -4975,7 +5070,7 @@ async function executeSubagent(params, subSessionId, taskName, taskPrompt, signa
     if (signal.aborted) {
       return accumulatedText || "[Cancelled]";
     }
-    return accumulatedText || `[Error: ${errorMessage}]`;
+    return accumulatedText || `[Error: ${errorMessage2}]`;
   } finally {
     await closeACPClientTools(subAcp);
   }
@@ -10716,13 +10811,13 @@ function createAutomationModule(ctx, deps) {
       sendEvent("task-update", { scheduleId, task });
       return task;
     } catch (err) {
-      const errorMessage = err instanceof Error ? err.message : String(err);
+      const errorMessage2 = err instanceof Error ? err.message : String(err);
       task.completedAt = Date.now();
       task.status = "error";
-      task.error = errorMessage;
+      task.error = errorMessage2;
       store.saveTask(scheduleId, task);
       sendEvent("task-update", { scheduleId, task });
-      console.error(`[Automation] Task failed:`, errorMessage);
+      console.error(`[Automation] Task failed:`, errorMessage2);
       return task;
     } finally {
       runningTasks.delete(scheduleId);
@@ -11369,7 +11464,7 @@ async function callOpenAICompatibleApi(provider, request) {
 }
 function createImageGenerationModule(ctx, deps) {
   const { storage } = ctx;
-  function getActiveProvider() {
+  function getActiveProvider2() {
     const settings = storage.getSettings();
     const provider = settings.imageGeneration?.find((p) => p.active);
     if (!provider) {
@@ -11380,7 +11475,7 @@ function createImageGenerationModule(ctx, deps) {
     return provider;
   }
   async function generateImage(sessionId, request) {
-    const provider = getActiveProvider();
+    const provider = getActiveProvider2();
     const apiResult = await callOpenAICompatibleApi(provider, request);
     const shareDir = storage.sessionShareDir(sessionId);
     if (!shareDir) throw new Error(`Session not found: ${sessionId}`);
@@ -11778,6 +11873,178 @@ function createToolboxModule(_ctx) {
   };
 }
 
+// src/backend/speech/manager.ts
+import { Buffer as Buffer2 } from "node:buffer";
+import {
+  createASRClient
+} from "unified-realtime-asr";
+var DEFAULT_DASHSCOPE_MODEL = "fun-asr-flash-8k-realtime";
+var DEFAULT_OPENAI_MODEL = "gpt-4o-transcribe";
+function optionalString(value) {
+  const trimmed = value?.trim();
+  return trimmed || void 0;
+}
+function required(value, name) {
+  const result = optionalString(value);
+  if (!result) throw new Error(`\u5B9E\u65F6\u8BED\u97F3\u8BC6\u522B\u914D\u7F6E\u7F3A\u5C11 ${name}\u3002`);
+  return result;
+}
+function buildOptions(provider) {
+  return {
+    language: optionalString(provider.language) ?? "zh-CN",
+    sampleRate: 16e3,
+    channels: 1,
+    format: "pcm",
+    interimResults: true,
+    punctuation: true,
+    autoReconnect: false,
+    transcriptionModel: provider.provider === "openai" ? optionalString(provider.model) ?? DEFAULT_OPENAI_MODEL : void 0
+  };
+}
+function buildConfig(provider) {
+  const options = buildOptions(provider);
+  const url = optionalString(provider.baseUrl);
+  switch (provider.provider) {
+    case "volcengine":
+      return {
+        provider: "volcengine",
+        apiKey: required(provider.apiKey, "API Key"),
+        resourceId: optionalString(provider.resourceId),
+        appId: optionalString(provider.appId),
+        url,
+        options
+      };
+    case "dashscope":
+      return {
+        provider: "dashscope",
+        apiKey: required(provider.apiKey, "API Key"),
+        model: optionalString(provider.model) ?? DEFAULT_DASHSCOPE_MODEL,
+        workspaceId: optionalString(provider.workspaceId),
+        region: provider.region,
+        workspace: optionalString(provider.workspace),
+        url,
+        options
+      };
+    case "openai":
+      return {
+        provider: "openai",
+        apiKey: required(provider.apiKey, "API Key"),
+        url,
+        options
+      };
+    case "iflytek":
+      return {
+        provider: "iflytek",
+        appId: required(provider.appId, "App ID"),
+        apiKey: required(provider.apiKey, "API Key"),
+        apiSecret: required(provider.apiSecret, "API Secret"),
+        url,
+        options
+      };
+  }
+}
+function getActiveProvider(ctx) {
+  const provider = ctx.storage.getSettings().speechToText.find((item) => item.active);
+  if (!provider) {
+    throw new Error("\u8BF7\u5148\u5728\u8BBE\u7F6E \u2192 \u8BED\u97F3\u8BC6\u522B\u4E2D\u914D\u7F6E\u5E76\u542F\u7528\u4E00\u4E2A Provider\u3002");
+  }
+  return provider;
+}
+function errorMessage(error) {
+  return error instanceof Error ? error.message : String(error);
+}
+function createAsrManager(ctx) {
+  const sessions = /* @__PURE__ */ new Map();
+  const keyOf2 = (clientId, asrSessionId) => `${clientId}:${asrSessionId}`;
+  const emitError = (session, error) => {
+    ctx.sendEvent("asr-error", {
+      clientId: session.clientId,
+      asrSessionId: session.asrSessionId,
+      message: errorMessage(error)
+    });
+  };
+  const emitTranscript = (session, transcript) => {
+    ctx.sendEvent("asr-transcript", {
+      clientId: session.clientId,
+      asrSessionId: session.asrSessionId,
+      text: transcript.text,
+      isFinal: transcript.isFinal,
+      id: transcript.id,
+      index: transcript.index,
+      speaker: transcript.speaker
+    });
+  };
+  return {
+    async start(clientId, asrSessionId) {
+      const key = keyOf2(clientId, asrSessionId);
+      if (sessions.has(key)) return { ok: true };
+      const provider = getActiveProvider(ctx);
+      const client2 = createASRClient(buildConfig(provider));
+      const active = { clientId, asrSessionId, client: client2 };
+      sessions.set(key, active);
+      client2.on("transcript", (transcript) => {
+        if (sessions.get(key) === active) emitTranscript(active, transcript);
+      });
+      client2.on("error", (error) => {
+        if (sessions.get(key) === active) emitError(active, error);
+      });
+      client2.on("close", (info) => {
+        if (sessions.get(key) !== active) return;
+        sessions.delete(key);
+        ctx.sendEvent("asr-closed", {
+          clientId,
+          asrSessionId,
+          code: info?.code,
+          reason: info?.reason
+        });
+      });
+      try {
+        await client2.connect();
+        return { ok: true };
+      } catch (error) {
+        sessions.delete(key);
+        client2.removeAllListeners();
+        throw new Error(errorMessage(error));
+      }
+    },
+    frame(clientId, asrSessionId, audioB64) {
+      const active = sessions.get(keyOf2(clientId, asrSessionId));
+      if (!active) return;
+      try {
+        active.client.sendAudio(Buffer2.from(audioB64, "base64"));
+      } catch (error) {
+        emitError(active, error);
+      }
+    },
+    async stop(clientId, asrSessionId) {
+      const key = keyOf2(clientId, asrSessionId);
+      const active = sessions.get(key);
+      if (!active) return;
+      try {
+        await active.client.close();
+      } catch (error) {
+        emitError(active, error);
+      } finally {
+        if (sessions.get(key) === active) sessions.delete(key);
+        active.client.removeAllListeners();
+      }
+    },
+    async closeAll() {
+      await Promise.all(
+        Array.from(sessions.values(), async (active) => {
+          try {
+            await active.client.close();
+          } catch {
+          } finally {
+            active.client.removeAllListeners();
+          }
+        })
+      );
+      sessions.clear();
+    }
+  };
+}
+
 // src/backend/backend.ts
 function initBackend(emitter) {
   const listeners = [];
@@ -11803,6 +12070,7 @@ function initBackend(emitter) {
   const _inference = createInferenceModule(ctx, { skills, search, toolbox });
   const memory = createMemoryModule(ctx, { inference: _inference });
   const imageGeneration = createImageGenerationModule(ctx, { ilink: ilink.state });
+  const asr = createAsrManager(ctx);
   const session = createSessionModule(ctx, {
     bridgeConnect,
     askUser,
@@ -12190,11 +12458,22 @@ function initBackend(emitter) {
     },
     async getMemorySystemFilePath({ projectId }) {
       return memory.getFilePath(projectId);
+    },
+    // Realtime ASR
+    async startRealtimeAsr({ clientId, asrSessionId }) {
+      return asr.start(clientId, asrSessionId);
+    },
+    async sendRealtimeAsrFrame({ clientId, asrSessionId, audioB64 }) {
+      asr.frame(clientId, asrSessionId, audioB64);
+    },
+    async stopRealtimeAsr({ clientId, asrSessionId }) {
+      await asr.stop(clientId, asrSessionId);
     }
   };
   webui.setHandlers(backendHandlers2);
   async function closeBackend2() {
     automation.stopAllCrons();
+    await asr.closeAll();
     webui.stop();
     await ilink.stopIlink({ logout: false });
     session.clearSession();
@@ -12210,7 +12489,7 @@ function initBackend(emitter) {
 }
 
 // package.json
-var version = "0.1.52";
+var version = "0.1.54";
 
 // src/server/main.ts
 var port;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@zythum02/fello-server",
-  "version": "0.1.52",
+  "version": "0.1.54",
   "description": "Fello Server — standalone headless AI agent server",
   "type": "module",
   "bin": {
@@ -36,6 +36,7 @@
     "ripgrep": "^0.3.1",
     "sharp": "^0.35.3",
     "undici": "^6.28.0",
+    "unified-realtime-asr": "0.2.1",
     "web-tree-sitter": "^0.26.11",
     "ws": "^8.21.2",
     "yaml": "^2.9.0",

package/out/renderer/assets/clipboard-xI9lExaB.js

@@ -1,15 +0,0 @@
-import { c as createLucideIcon } from "./index-BJPggXxf.js";
-const __iconNode = [
-  ["rect", { width: "8", height: "4", x: "8", y: "2", rx: "1", ry: "1", key: "tgr4d6" }],
-  [
-    "path",
-    {
-      d: "M16 4h2a2 2 0 0 1 2 2v14a2 2 0 0 1-2 2H6a2 2 0 0 1-2-2V6a2 2 0 0 1 2-2h2",
-      key: "116196"
-    }
-  ]
-];
-const Clipboard = createLucideIcon("clipboard", __iconNode);
-export {
-  Clipboard as C
-};