feat: /voice use best grop whisper provider free and fast

README.md

@@ -95,7 +95,8 @@ cp VersperClaw ~/.local/bin
     "ENABLE_TOOL_SEARCH": "true",
     "TAVILY_API_KEY": "",
   # "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "0", default 1 
-    "teammateMode": "tmux" # or in-process
+    "teammateMode": "tmux", # or in-process
+    "groqApiKey": "gsk_" # https://console.groq.com
   }
 
   # set auto-dream config
@@ -154,6 +155,8 @@ mitmproxy -p 8888
 
 ```bash
 # /voice
+# node-edge-tts
+# local whisper-stt
 uv venv
 uv pip install faster-whisper edge-tts voxcpm
 

src/commands/voice/voice.ts

@@ -88,10 +88,10 @@ export const call: LocalCommandCall = async () => {
     }
   }
 
-  // Enable voice with local provider
+  // Enable voice with Groq provider (cloud Whisper API, no Python needed)
   const result = updateSettingsForSource('userSettings', {
     voiceEnabled: true,
-    voiceProvider: 'local' as const,
+    voiceProvider: 'groq' as const,
     voiceAutoTTS: true,
   })
   if (result.error) {
@@ -124,6 +124,6 @@ export const call: LocalCommandCall = async () => {
   }
   return {
     type: 'text' as const,
-    value: `Voice mode enabled (local whisper). Hold ${key} to record.${langNote}`,
+    value: `Voice mode enabled (Groq Whisper). Hold ${key} to record.${langNote}`,
   }
 }

src/hooks/useAutoTTS.ts

@@ -1,5 +1,6 @@
 import { useEffect, useRef } from 'react'
-import { speakWithEdgeTTS, playAudioFile } from '../services/voice/edgeTTS.js'
+import { playAudioFile } from '../services/voice/edgeTTS.js'
+import { edgeTts } from '../friend/tts.js'
 import { getInitialSettings } from '../utils/settings/settings.js'
 import type { RenderableMessage } from '../types/message.js'
 
@@ -67,10 +68,7 @@ export function useAutoTTS(messages: RenderableMessage[], isLoading?: boolean):
       const text = content.text
 
       const run = async () => {
-        const result = await speakWithEdgeTTS(text, {
-          voice,
-          pythonPath: settings.voiceTTSCommand || undefined,
-        })
+        const result = await edgeTts({ text, voice })
         if (result.success && result.audioPath) {
           await playAudioFile(result.audioPath)
         }

src/hooks/useVoice.ts

@@ -22,6 +22,7 @@ import {
   type VoiceStreamConnection,
 } from '../services/voiceStreamSTT.js'
 import { connectDoubaoStream } from '../services/doubaoSTT.js'
+import { connectGroqStream } from '../services/voice/groqSTT.js'
 import { connectLocalWhisperStream, preloadWhisperModel } from '../services/voice/whisperSTT.js'
 import { logForDebugging } from '../utils/debug.js'
 import { toError } from '../utils/errors.js'
@@ -148,6 +149,10 @@ function isLocalProvider(): boolean {
   return getInitialSettings().voiceProvider === 'local'
 }
 
+function isGroqProvider(): boolean {
+  return getInitialSettings().voiceProvider === 'groq'
+}
+
 // Lazy-loaded voice module. We defer importing voice.ts (and its native
 // audio-capture-napi dependency) until voice input is actually activated.
 // On macOS, loading the native audio module can trigger a TCC microphone
@@ -401,8 +406,8 @@ export function useVoice({
           !silentDropRetriedRef.current &&
           fullAudioRef.current.length > 0
         ) {
-// Local whisper doesn't support silent-drop replay (different backend)
-        if (isLocalProvider()) {
+// Local whisper & groq don't support silent-drop replay (batch backends)
+        if (isLocalProvider() || isGroqProvider()) {
           callbacks.onClose()
           return
         }
@@ -595,7 +600,7 @@ export function useVoice({
   // stop when it loses focus. This enables a "multi-clauding army"
   // workflow where voice input follows window focus.
   useEffect(() => {
-    if (!enabled || !focusMode || isDoubaoProvider() || isLocalProvider()) {
+    if (!enabled || !focusMode || isDoubaoProvider() || isLocalProvider() || isGroqProvider()) {
       // Focus mode was disabled while a focus-driven recording was active —
       // stop the recording so it doesn't linger until the silence timer fires.
       if (focusTriggeredRef.current && stateRef.current === 'recording') {
@@ -810,6 +815,9 @@ export function useVoice({
           connectLocalWhisperStream(cbs, { language: stt.code })
       } else if (isDoubaoProvider()) {
         connectFn = (cbs, opts) => connectDoubaoStream(cbs, opts)
+      } else if (isGroqProvider()) {
+        connectFn = (cbs, opts) =>
+          connectGroqStream(cbs, { language: opts.language })
       } else {
         connectFn = (cbs, opts) => connectVoiceStream(cbs, opts)
       }
@@ -1026,6 +1034,13 @@ export function useVoice({
             onErrorRef.current?.(
               'Local voice mode failed. Ensure Python with faster-whisper is installed.',
             )
+          } else if (isGroqProvider()) {
+            logForDebugging(
+              '[voice] Groq STT failed to initialize',
+            )
+            onErrorRef.current?.(
+              'Groq STT failed. Check your GROQ_API_KEY.',
+            )
           } else {
             logForDebugging(
               '[voice] Failed to connect to voice_stream (no OAuth token?)',
@@ -1051,8 +1066,8 @@ export function useVoice({
       })
     }
 
-    // Doubao and local backends don't use keyterms — skip the async fetch
-    if (isDoubaoProvider() || isLocalProvider()) {
+    // Doubao, local, and groq backends don't use keyterms — skip the async fetch
+    if (isDoubaoProvider() || isLocalProvider() || isGroqProvider()) {
       attemptConnect([])
     } else {
       void getVoiceKeyterms().then(attemptConnect)
@@ -1070,7 +1085,7 @@ export function useVoice({
   // delay of ~500ms on macOS).
   const handleKeyEvent = useCallback(
     (fallbackMs = REPEAT_FALLBACK_MS): void => {
-      const sttAvailable = isLocalProvider()
+      const sttAvailable = isLocalProvider() || isGroqProvider()
         ? true
         : isDoubaoProvider()
           ? isDoubaoAvailableSync()

src/utils/settings/types.ts

@@ -868,10 +868,11 @@ export const SettingsSchema = lazySchema(() =>
               .optional()
               .describe('Enable voice mode (hold-to-talk dictation)'),
             voiceProvider: z
-              .enum(['local', 'doubao', 'anthropic'])
+              .enum(['local', 'doubao', 'anthropic', 'groq'])
               .optional()
               .describe(
-                'Voice STT backend: "local" (whisper/whisper.cpp / faster-whisper) or "doubao" (Doubao ASR).'
+                'Voice STT backend: "local" (whisper/whisper.cpp / faster-whisper),'
+                  + ' "doubao" (Doubao ASR), or "groq" (Groq Whisper API).'
                   + ' "anthropic" is a client-side compatibility alias.',
               ),
             voiceAutoTTS: z