vellum 0.2.11 → 0.2.13

package/src/calls/elevenlabs-config.ts

@@ -0,0 +1,29 @@
+import { getConfig } from '../config/loader.js';
+import { getSecureKey } from '../security/secure-keys.js';
+import { getLogger } from '../util/logger.js';
+
+const log = getLogger('elevenlabs-config');
+
+export interface ElevenLabsConfig {
+  apiKey: string;
+  apiBaseUrl: string;
+  agentId: string;
+  registerCallTimeoutMs: number;
+}
+
+export function getElevenLabsConfig(): ElevenLabsConfig {
+  const config = getConfig();
+  const voice = config.calls.voice;
+
+  const apiKey = getSecureKey('credential:elevenlabs:api_key') ?? '';
+  if (!apiKey) {
+    log.warn('No ElevenLabs API key found in secure key store (credential:elevenlabs:api_key)');
+  }
+
+  return {
+    apiKey,
+    apiBaseUrl: voice.elevenlabs.apiBaseUrl,
+    agentId: voice.elevenlabs.agentId,
+    registerCallTimeoutMs: voice.elevenlabs.registerCallTimeoutMs,
+  };
+}

package/src/calls/twilio-routes.ts

@@ -25,6 +25,9 @@ import { getTwilioConfig } from './twilio-config.js';
 import { loadConfig } from '../config/loader.js';
 import { getTwilioRelayUrl } from '../inbound/public-ingress-urls.js';
 import { fireCallCompletionNotifier } from './call-state.js';
+import { resolveVoiceQualityProfile } from './voice-quality.js';
+import { getElevenLabsConfig } from './elevenlabs-config.js';
+import { ElevenLabsClient } from './elevenlabs-client.js';
 
 const log = getLogger('twilio-routes');
 
@@ -39,17 +42,22 @@ function escapeXml(str: string): string {
     .replace(/'/g, ''');
 }
 
-function generateTwiML(callSessionId: string, relayUrl: string, welcomeGreeting: string): string {
+export function generateTwiML(
+  callSessionId: string,
+  relayUrl: string,
+  welcomeGreeting: string,
+  profile: { language: string; transcriptionProvider: string; ttsProvider: string; voice: string },
+): string {
   return `<?xml version="1.0" encoding="UTF-8"?>
 <Response>
   <Connect>
     <ConversationRelay
       url="${escapeXml(relayUrl)}?callSessionId=${escapeXml(callSessionId)}"
       welcomeGreeting="${escapeXml(welcomeGreeting)}"
-      voice="Google.en-US-Journey-O"
-      language="en-US"
-      transcriptionProvider="Deepgram"
-      ttsProvider="Google"
+      voice="${escapeXml(profile.voice)}"
+      language="${escapeXml(profile.language)}"
+      transcriptionProvider="${escapeXml(profile.transcriptionProvider)}"
+      ttsProvider="${escapeXml(profile.ttsProvider)}"
       interruptible="true"
       dtmfDetection="true"
     />
@@ -131,6 +139,14 @@ export async function handleVoiceWebhook(req: Request): Promise<Response> {
     log.info({ callSessionId, callSid }, 'Stored CallSid from voice webhook');
   }
 
+  const profile = resolveVoiceQualityProfile(loadConfig());
+
+  log.info({ callSessionId, mode: profile.mode, ttsProvider: profile.ttsProvider, voice: profile.voice }, 'Voice quality profile resolved');
+
+  if (profile.validationErrors.length > 0) {
+    log.warn({ callSessionId, errors: profile.validationErrors }, 'Voice quality profile has validation warnings');
+  }
+
   const twilioConfig = getTwilioConfig();
   let relayUrl: string;
   try {
@@ -141,7 +157,40 @@ export async function handleVoiceWebhook(req: Request): Promise<Response> {
   }
   const welcomeGreeting = process.env.CALL_WELCOME_GREETING ?? 'Hello, how can I help you today?';
 
-  const twiml = generateTwiML(callSessionId, relayUrl, welcomeGreeting);
+  if (profile.mode === 'elevenlabs_agent') {
+    try {
+      const elevenLabsConfig = getElevenLabsConfig();
+      const client = new ElevenLabsClient({
+        apiBaseUrl: elevenLabsConfig.apiBaseUrl,
+        apiKey: elevenLabsConfig.apiKey,
+        timeoutMs: elevenLabsConfig.registerCallTimeoutMs,
+      });
+
+      const result = await client.registerCall({
+        agent_id: elevenLabsConfig.agentId,
+        from_number: formBody.get('From') || session.fromNumber,
+        to_number: formBody.get('To') || session.toNumber,
+        direction: 'outbound',
+      });
+
+      log.info({ callSessionId }, 'ElevenLabs register-call succeeded');
+      return new Response(result.twiml, {
+        status: 200,
+        headers: { 'Content-Type': 'text/xml' },
+      });
+    } catch (err) {
+      log.error({ err, callSessionId }, 'ElevenLabs register-call failed');
+      if (profile.fallbackToStandardOnError) {
+        log.warn({ callSessionId }, 'Falling back to twilio_standard mode');
+        const standardProfile = resolveVoiceQualityProfile({ ...loadConfig(), calls: { ...loadConfig().calls, voice: { ...loadConfig().calls.voice, mode: 'twilio_standard' } } });
+        const twiml = generateTwiML(callSessionId, relayUrl, welcomeGreeting, standardProfile);
+        return new Response(twiml, { status: 200, headers: { 'Content-Type': 'text/xml' } });
+      }
+      return new Response('ElevenLabs service unavailable', { status: 502 });
+    }
+  }
+
+  const twiml = generateTwiML(callSessionId, relayUrl, welcomeGreeting, profile);
 
   log.info({ callSessionId }, 'Returning ConversationRelay TwiML');
 

package/src/calls/voice-quality.ts

@@ -0,0 +1,92 @@
+import { loadConfig } from '../config/loader.js';
+
+export interface VoiceQualityProfile {
+  mode: 'twilio_standard' | 'twilio_elevenlabs_tts' | 'elevenlabs_agent';
+  language: string;
+  transcriptionProvider: string;
+  ttsProvider: string;
+  voice: string;
+  agentId?: string;
+  fallbackToStandardOnError: boolean;
+  validationErrors: string[];
+}
+
+/**
+ * Build a Twilio-compatible ElevenLabs voice string.
+ * Format: voiceId or voiceId-modelId-stability_similarity_style
+ */
+export function buildElevenLabsVoiceSpec(config: {
+  voiceId: string;
+  voiceModelId: string;
+  stability: number;
+  similarityBoost: number;
+  style: number;
+}): string {
+  if (!config.voiceId) return '';
+  return `${config.voiceId}-${config.voiceModelId}-${config.stability}_${config.similarityBoost}_${config.style}`;
+}
+
+/**
+ * Resolve the effective voice quality profile from config.
+ * Returns a profile with all resolved values ready for use by TwiML generation
+ * and call orchestration.
+ */
+export function resolveVoiceQualityProfile(config?: ReturnType<typeof loadConfig>): VoiceQualityProfile {
+  const cfg = config ?? loadConfig();
+  const voice = cfg.calls.voice;
+  const errors: string[] = [];
+
+  // Default/standard profile
+  const standardProfile: VoiceQualityProfile = {
+    mode: 'twilio_standard',
+    language: voice.language,
+    transcriptionProvider: voice.transcriptionProvider,
+    ttsProvider: 'Google',
+    voice: 'Google.en-US-Journey-O',
+    fallbackToStandardOnError: voice.fallbackToStandardOnError,
+    validationErrors: [],
+  };
+
+  if (voice.mode === 'twilio_standard') {
+    return standardProfile;
+  }
+
+  if (voice.mode === 'twilio_elevenlabs_tts') {
+    if (!voice.elevenlabs.voiceId && !voice.fallbackToStandardOnError) {
+      errors.push('calls.voice.elevenlabs.voiceId is required for twilio_elevenlabs_tts mode when fallback is disabled');
+    }
+    if (!voice.elevenlabs.voiceId && voice.fallbackToStandardOnError) {
+      return { ...standardProfile, validationErrors: ['calls.voice.elevenlabs.voiceId is empty; falling back to twilio_standard'] };
+    }
+    return {
+      mode: 'twilio_elevenlabs_tts',
+      language: voice.language,
+      transcriptionProvider: voice.transcriptionProvider,
+      ttsProvider: 'ElevenLabs',
+      voice: buildElevenLabsVoiceSpec(voice.elevenlabs),
+      fallbackToStandardOnError: voice.fallbackToStandardOnError,
+      validationErrors: errors,
+    };
+  }
+
+  if (voice.mode === 'elevenlabs_agent') {
+    if (!voice.elevenlabs.agentId && !voice.fallbackToStandardOnError) {
+      errors.push('calls.voice.elevenlabs.agentId is required for elevenlabs_agent mode when fallback is disabled');
+    }
+    if (!voice.elevenlabs.agentId && voice.fallbackToStandardOnError) {
+      return { ...standardProfile, validationErrors: ['calls.voice.elevenlabs.agentId is empty; falling back to twilio_standard'] };
+    }
+    return {
+      mode: 'elevenlabs_agent',
+      language: voice.language,
+      transcriptionProvider: voice.transcriptionProvider,
+      ttsProvider: 'ElevenLabs',
+      voice: buildElevenLabsVoiceSpec(voice.elevenlabs),
+      agentId: voice.elevenlabs.agentId,
+      fallbackToStandardOnError: voice.fallbackToStandardOnError,
+      validationErrors: errors,
+    };
+  }
+
+  return standardProfile;
+}

package/src/cli/main-screen.tsx

@@ -1,19 +1,9 @@
-import { type ReactElement } from "react";
-import { basename } from "node:path";
-import { Box, render as inkRender, Text } from "ink";
+import { createRequire } from "node:module";
+import { dirname, join } from "node:path";
 import { getSocketPath, getWorkspaceDir } from "../util/platform.js";
-import { APP_VERSION } from "../version.js";
 
 const LEFT_PANEL_WIDTH = 36;
-
-const VELLY_ART = [
-  "    ,___,",
-  "   ( O O )",
-  "    /)V(\\",
-  "   //   \\\\",
-  '  /"     "\\',
-  "  ^       ^",
-];
+const RIGHT_LINE_COUNT = 11;
 
 export interface MainScreenLayout {
   height: number;
@@ -21,116 +11,24 @@ export interface MainScreenLayout {
   statusCol: number;
 }
 
-function DefaultMainScreen(): ReactElement {
+export function renderMainScreen(): MainScreenLayout {
   const socketPath = getSocketPath();
   const workspace = getWorkspaceDir();
-  const dirName = basename(workspace);
-
-  const tips = [
-    "Send a message to start chatting",
-    "Use /help to see available commands",
-  ];
-
-  const leftLines = [
-    " ",
-    "    Meet your Assistant!",
-    " ",
-    ...VELLY_ART.map((l) => `  ${l}`),
-    " ",
-    `  ${socketPath}`,
-    `  ~/${dirName}`,
-  ];
-
-  const rightLines = [
-    " ",
-    "Tips for getting started",
-    ...tips,
-    " ",
-    "Daemon",
-    "connecting...",
-    "Version",
-    APP_VERSION,
-    "Status",
-    "checking...",
-  ];
+  const assistantId = workspace.split("/").pop() ?? "vellum";
 
-  const maxLines = Math.max(leftLines.length, rightLines.length);
-
-  return (
-    <Box flexDirection="column" width={72}>
-      <Text dimColor>{"── Vellum " + "─".repeat(62)}</Text>
-      <Box flexDirection="row">
-        <Box flexDirection="column" width={LEFT_PANEL_WIDTH}>
-          {Array.from({ length: maxLines }, (_, i) => {
-            const line = leftLines[i] ?? " ";
-            if (i === 1) {
-              return (
-                <Text key={i} bold>
-                  {line}
-                </Text>
-              );
-            }
-            if (i > 2 && i <= 2 + VELLY_ART.length) {
-              return (
-                <Text key={i} color="magenta">
-                  {line}
-                </Text>
-              );
-            }
-            if (i > 2 + VELLY_ART.length) {
-              return (
-                <Text key={i} dimColor>
-                  {line}
-                </Text>
-              );
-            }
-            return <Text key={i}>{line}</Text>;
-          })}
-        </Box>
-        <Box flexDirection="column">
-          {Array.from({ length: maxLines }, (_, i) => {
-            const line = rightLines[i] ?? " ";
-            const isHeading = i === 1 || i === 6;
-            const isDim = i === 5 || i === 7 || i === 9;
-            if (isHeading) {
-              return (
-                <Text key={i} color="magenta">
-                  {line}
-                </Text>
-              );
-            }
-            if (isDim) {
-              return (
-                <Text key={i} dimColor>
-                  {line}
-                </Text>
-              );
-            }
-            return <Text key={i}>{line}</Text>;
-          })}
-        </Box>
-      </Box>
-      <Text dimColor>{"─".repeat(72)}</Text>
-      <Text> </Text>
-      <Text dimColor> ? for shortcuts</Text>
-      <Text> </Text>
-    </Box>
-  );
-}
-
-export function renderMainScreen(): MainScreenLayout {
-  const leftLineCount = 3 + VELLY_ART.length + 3;
-  const rightLineCount = 11;
-  const maxLines = Math.max(leftLineCount, rightLineCount);
+  const require = createRequire(import.meta.url);
+  const cliPkgPath = require.resolve("@vellumai/cli/package.json");
+  const cliRoot = dirname(cliPkgPath);
+  // Dynamic require to bypass NodeNext strict module resolution for the
+  // CLI package which ships raw TypeScript with bundler-style imports.
+  const { render } = require(join(cliRoot, "src", "components", "DefaultMainScreen.tsx")) as {
+    render: (runtimeUrl: string, assistantId: string, species: string) => number;
+  };
 
-  const { unmount } = inkRender(<DefaultMainScreen />, {
-    exitOnCtrlC: false,
-  });
-  unmount();
+  const height = render(socketPath, assistantId, "vellum");
 
-  const statusCanvasLine = rightLineCount + 1;
+  const statusCanvasLine = RIGHT_LINE_COUNT + 1;
   const statusCol = LEFT_PANEL_WIDTH + 1;
-  const height = 1 + maxLines + 4;
 
   return { height, statusLine: statusCanvasLine, statusCol };
 }

package/src/config/bundled-skills/macos-automation/SKILL.md

@@ -0,0 +1,66 @@
+---
+name: "macOS Automation"
+description: "Automate native macOS apps and system interactions via osascript (AppleScript)"
+user-invocable: false
+disable-model-invocation: false
+metadata: {"vellum": {"emoji": "🍎", "os": ["darwin"]}}
+---
+
+Use this skill to interact with native macOS apps and system-level features via `osascript` (AppleScript) through `host_bash`. Always prefer osascript over browser automation or computer-use for anything involving a native macOS app.
+
+## Supported Apps
+
+**Communication:** Messages, Mail, Microsoft Outlook, FaceTime
+**Contacts & Calendar:** Contacts, Calendar, Reminders
+**Notes & Writing:** Notes, TextEdit, Pages, BBEdit, CotEditor
+**Files:** Finder, Path Finder
+**Browsers:** Safari, Google Chrome
+**Music & Media:** Music (iTunes), Spotify, VLC, Podcasts, TV
+**Productivity:** OmniFocus, Things 3, OmniOutliner, OmniPlan, OmniGraffle
+**Office:** Microsoft Word, Microsoft Excel, Numbers, Keynote
+**Developer tools:** Xcode, Terminal, iTerm2, Script Editor
+**System:** System Events (UI scripting for any app), System Settings
+**Automation:** Keyboard Maestro, Alfred, Automator
+**Creative:** Adobe Photoshop, Final Cut Pro
+
+For any unlisted app, check scriptability first:
+```bash
+osascript -e 'tell application "AppName" to get name'
+```
+
+## Examples
+
+```bash
+# Send an iMessage
+osascript -e 'tell application "Messages" to send "Hello!" to buddy "user@example.com"'
+
+# Look up a contact
+osascript -e 'tell application "Contacts" to get {name, phones} of every person whose name contains "Marina"'
+
+# Read upcoming calendar events
+osascript -e 'tell application "Calendar" to get summary of every event of calendar "Home" whose start date > (current date)'
+
+# Create a reminder
+osascript -e 'tell application "Reminders" to make new reminder with properties {name:"Buy milk", due date:((current date) + 1 * hours)}'
+
+# Send an email
+osascript -e 'tell application "Mail" to send (make new outgoing message with properties {subject:"Hi", content:"Hello", visible:true})'
+
+# Create a note
+osascript -e 'tell application "Notes" to make new note at folder "Notes" with properties {body:"My note"}'
+
+# Open a URL in Safari
+osascript -e 'tell application "Safari" to open location "https://example.com"'
+
+# Play/pause Music
+osascript -e 'tell application "Music" to playpause'
+
+# Display a system notification
+osascript -e 'display notification "Done!" with title "Vellum"'
+```
+
+## Tips
+
+- For multi-line scripts, write them to a `.applescript` file and run with `osascript path/to/script.applescript`
+- Use `System Events` for UI scripting apps that don't have their own AppleScript dictionary
+- AppleScript permissions are gated by macOS TCC — if a command fails with a permission error, use `request_system_permission` to prompt the user