botinabox 1.2.1 → 1.3.1

package/README.md

@@ -12,7 +12,11 @@ A modular TypeScript framework for building multi-agent bots with LLM orchestrat
 - **Event-driven hooks** -- Priority-ordered, filter-based event bus for decoupled inter-layer communication.
 - **Budget controls** -- Per-agent and global cost tracking with warning thresholds and hard stops.
 - **Scheduling** -- Database-backed cron and one-time schedules that fire hook events.
-- **Google connectors** -- Gmail and Calendar integration via OAuth2.
+- **Connectors** -- Generic `Connector<T>` interface for external service integrations. Ships with Google Gmail and Calendar implementations (OAuth2 and service account auth).
+- **Domain tables** -- `defineDomainTables()` and `defineDomainEntityContexts()` for standard multi-agent app schemas (org, project, client, invoice, repository, and more).
+- **Auto-update** -- `autoUpdate()` checks npm for newer versions and installs them at startup.
+- **Cursor persistence** -- `SecretStore.loadCursor()` / `saveCursor()` for persisting sync state across restarts.
+- **Utilities** -- `truncateAtWord()` for word-boundary text truncation, `parseClaudeStream()`, `buildProcessEnv()`, and more.
 - **Security** -- Input sanitization, field length enforcement, audit logging, and HMAC webhook verification.
 
 ## Install

package/dist/channels/slack/index.d.ts

@@ -80,6 +80,61 @@ declare function extractVoiceTranscript(file: SlackFile): string | null;
  * with `[Voice message]`.
  */
 declare function parseSlackEvent(event: SlackEvent): InboundMessage;
+/**
+ * Enrich a voice message with local transcription when Slack's built-in
+ * transcription is unavailable.
+ *
+ * Downloads the audio file from Slack using the bot token, converts to WAV,
+ * and transcribes locally via whisper-node. Returns the original message
+ * unchanged if transcription fails or is not needed.
+ *
+ * @param msg - The parsed inbound message (from parseSlackEvent)
+ * @param botToken - Slack bot token for authenticated file download
+ * @returns The message with body replaced by transcript, or original on failure
+ */
+declare function enrichVoiceMessage(msg: InboundMessage, botToken: string): Promise<InboundMessage>;
+
+/**
+ * Local voice transcription via whisper-node (whisper.cpp bindings).
+ *
+ * whisper-node is an optional dependency — if not installed, transcription
+ * degrades gracefully (returns null). Requires ffmpeg on the system PATH
+ * for audio format conversion.
+ *
+ * Setup:
+ *   npm install whisper-node
+ *   npx whisper-node download    # download a model (e.g. base.en)
+ *   brew install ffmpeg           # or equivalent for your platform
+ */
+interface TranscribeOptions {
+    /** Whisper model name (default: "base.en"). Run `npx whisper-node download` to get models. */
+    modelName?: string;
+    /** Language code (default: "auto"). Use "en" for English-only models. */
+    language?: string;
+}
+interface TranscribeResult {
+    /** The full transcribed text */
+    text: string;
+    /** Individual segments with timestamps */
+    segments: Array<{
+        start: string;
+        end: string;
+        speech: string;
+    }>;
+}
+/**
+ * Transcribe an audio buffer using whisper-node.
+ *
+ * @param audioBuffer - Raw audio data (any format ffmpeg can decode)
+ * @param filename - Original filename (used for temp file extension)
+ * @param opts - Transcription options
+ * @returns Transcribed text, or null if transcription fails or whisper-node is not installed
+ */
+declare function transcribeAudio(audioBuffer: Buffer, filename: string, opts?: TranscribeOptions): Promise<string | null>;
+/**
+ * Download an audio file from a URL with bearer token authentication.
+ */
+declare function downloadAudio(url: string, token: string): Promise<Buffer | null>;
 
 /**
  * Slack outbound formatting.
@@ -104,4 +159,4 @@ interface SlackConfig {
     signingSecret?: string;
 }
 
-export { type BoltClient, SlackAdapter, type SlackConfig, type SlackEvent, type SlackFile, createSlackAdapter as default, extractVoiceTranscript, formatForSlack, parseSlackEvent };
+export { type BoltClient, SlackAdapter, type SlackConfig, type SlackEvent, type SlackFile, type TranscribeOptions, type TranscribeResult, createSlackAdapter as default, downloadAudio, enrichVoiceMessage, extractVoiceTranscript, formatForSlack, parseSlackEvent, transcribeAudio };

package/dist/channels/slack/index.js

@@ -1,7 +1,10 @@
 import {
+  downloadAudio,
+  enrichVoiceMessage,
   extractVoiceTranscript,
-  parseSlackEvent
-} from "../../chunk-2LGXQPEA.js";
+  parseSlackEvent,
+  transcribeAudio
+} from "../../chunk-GS2JFL6I.js";
 
 // src/channels/slack/outbound.ts
 function formatForSlack(text) {
@@ -64,8 +67,12 @@ var SlackAdapter = class {
   /** Simulate receiving an inbound message (for testing/webhooks). */
   async receive(event) {
     if (this.onMessage) {
-      const { parseSlackEvent: parseSlackEvent2 } = await import("../../inbound-CGIXRXGC.js");
-      const msg = parseSlackEvent2(event);
+      const { parseSlackEvent: parseSlackEvent2 } = await import("../../inbound-AFBUPSPG.js");
+      const { enrichVoiceMessage: enrichVoiceMessage2 } = await import("../../inbound-AFBUPSPG.js");
+      let msg = parseSlackEvent2(event);
+      if (msg.body.includes("[Voice message") && this.config?.botToken) {
+        msg = await enrichVoiceMessage2(msg, this.config.botToken);
+      }
       await this.onMessage(msg);
     }
   }
@@ -76,7 +83,10 @@ function createSlackAdapter(client) {
 export {
   SlackAdapter,
   createSlackAdapter as default,
+  downloadAudio,
+  enrichVoiceMessage,
   extractVoiceTranscript,
   formatForSlack,
-  parseSlackEvent
+  parseSlackEvent,
+  transcribeAudio
 };

package/dist/chunk-GS2JFL6I.js

@@ -0,0 +1,144 @@
+// src/channels/slack/transcribe.ts
+import { execSync } from "child_process";
+import { writeFileSync, unlinkSync, mkdirSync } from "fs";
+import { join } from "path";
+import { randomUUID } from "crypto";
+import os from "os";
+import { createRequire } from "module";
+var TEMP_DIR = join(os.tmpdir(), "botinabox-audio");
+async function transcribeAudio(audioBuffer, filename, opts) {
+  let whisper;
+  try {
+    const require2 = createRequire(import.meta.url);
+    const mod = require2("whisper-node");
+    whisper = mod.whisper ?? mod.default ?? mod;
+  } catch {
+    console.warn("[botinabox] whisper-node not installed \u2014 voice transcription unavailable. Run: npm install whisper-node && npx whisper-node download");
+    return null;
+  }
+  try {
+    execSync("ffmpeg -version", { stdio: "ignore" });
+  } catch {
+    console.warn("[botinabox] ffmpeg not found \u2014 required for audio conversion. Install: brew install ffmpeg");
+    return null;
+  }
+  const id = randomUUID().slice(0, 8);
+  const ext = filename.split(".").pop() ?? "aac";
+  mkdirSync(TEMP_DIR, { recursive: true });
+  const inputPath = join(TEMP_DIR, `${id}.${ext}`);
+  const wavPath = join(TEMP_DIR, `${id}.wav`);
+  try {
+    writeFileSync(inputPath, audioBuffer);
+    execSync(
+      `ffmpeg -y -i "${inputPath}" -ar 16000 -ac 1 -c:a pcm_s16le "${wavPath}"`,
+      { stdio: "ignore", timeout: 3e4 }
+    );
+    const segments = await whisper(wavPath, {
+      modelName: opts?.modelName ?? "base.en",
+      whisperOptions: {
+        language: opts?.language ?? "auto"
+      }
+    });
+    if (!segments || segments.length === 0) return null;
+    return segments.map((s) => s.speech).join(" ").trim();
+  } catch (err) {
+    console.error("[botinabox] Transcription failed:", err);
+    return null;
+  } finally {
+    try {
+      unlinkSync(inputPath);
+    } catch {
+    }
+    try {
+      unlinkSync(wavPath);
+    } catch {
+    }
+  }
+}
+async function downloadAudio(url, token) {
+  try {
+    const resp = await fetch(url, {
+      headers: { Authorization: `Bearer ${token}` }
+    });
+    if (!resp.ok) {
+      console.error(`[botinabox] Audio download failed: ${resp.status} ${resp.statusText}`);
+      return null;
+    }
+    return Buffer.from(await resp.arrayBuffer());
+  } catch (err) {
+    console.error("[botinabox] Audio download error:", err);
+    return null;
+  }
+}
+
+// src/channels/slack/inbound.ts
+var AUDIO_TYPES = /* @__PURE__ */ new Set(["aac", "mp4", "m4a", "ogg", "webm", "mp3", "wav"]);
+function extractVoiceTranscript(file) {
+  const isAudio = file.subtype === "slack_audio" || AUDIO_TYPES.has(file.filetype ?? "");
+  if (!isAudio) return null;
+  const transcript = file.transcription?.preview?.content ?? (typeof file.preview === "string" ? file.preview : null);
+  return transcript ?? null;
+}
+function parseSlackEvent(event) {
+  const id = event.client_msg_id ?? event.ts ?? event.event_ts ?? `slack-${Date.now()}`;
+  const channel = event.channel ?? "unknown";
+  const from = event.user ?? "unknown";
+  const threadId = event.thread_ts !== void 0 ? event.thread_ts : void 0;
+  const receivedAt = event.ts ? new Date(parseFloat(event.ts) * 1e3).toISOString() : (/* @__PURE__ */ new Date()).toISOString();
+  let body = event.text ?? "";
+  if (event.subtype === "file_share" && event.files?.length) {
+    for (const file of event.files) {
+      const transcript = extractVoiceTranscript(file);
+      if (transcript) {
+        body = body ? `${body}
+
+[Voice message] ${transcript}` : `[Voice message] ${transcript}`;
+        break;
+      }
+    }
+  }
+  if (event.subtype === "file_share" && event.files?.length && !body) {
+    const hasAudio = event.files.some(
+      (f) => f.subtype === "slack_audio" || AUDIO_TYPES.has(f.filetype ?? "")
+    );
+    if (hasAudio) {
+      body = "[Voice message \u2014 no transcript available]";
+    }
+  }
+  return {
+    id,
+    channel,
+    from,
+    body,
+    threadId,
+    receivedAt,
+    raw: event
+  };
+}
+async function enrichVoiceMessage(msg, botToken) {
+  if (!msg.body.includes("[Voice message \u2014 no transcript available]")) return msg;
+  const raw = msg.raw;
+  const files = raw?.files;
+  if (!files?.length) return msg;
+  const audioFile = files.find(
+    (f) => f.subtype === "slack_audio" || AUDIO_TYPES.has(f.filetype ?? "")
+  );
+  if (!audioFile?.url_private) return msg;
+  const buffer = await downloadAudio(audioFile.url_private, botToken);
+  if (!buffer) return msg;
+  const filename = audioFile.name ?? `voice.${audioFile.filetype ?? "aac"}`;
+  const transcript = await transcribeAudio(buffer, filename);
+  if (!transcript) return msg;
+  return {
+    ...msg,
+    body: `[Voice message] ${transcript}`
+  };
+}
+
+export {
+  transcribeAudio,
+  downloadAudio,
+  extractVoiceTranscript,
+  parseSlackEvent,
+  enrichVoiceMessage
+};

package/dist/inbound-AFBUPSPG.js

@@ -0,0 +1,10 @@
+import {
+  enrichVoiceMessage,
+  extractVoiceTranscript,
+  parseSlackEvent
+} from "./chunk-GS2JFL6I.js";
+export {
+  enrichVoiceMessage,
+  extractVoiceTranscript,
+  parseSlackEvent
+};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "botinabox",
-  "version": "1.2.1",
+  "version": "1.3.1",
   "description": "Bot in a Box — framework for building multi-agent bots",
   "type": "module",
   "main": "./dist/index.js",
@@ -63,6 +63,9 @@
     "uuid": "^13.0.0",
     "yaml": "^2.7.0"
   },
+  "optionalDependencies": {
+    "whisper-node": "^1.1.1"
+  },
   "peerDependencies": {
     "@anthropic-ai/sdk": "^0.52.0",
     "googleapis": ">=140.0.0",