--speak_messages TTS updated cross platform

interpreter/terminal_interface/terminal_interface.py

@@ -14,6 +14,8 @@
 import re
 import subprocess
 import time
+import pyttsx3
+import threading
 
 from ..core.utils.scan_code import scan_code
 from ..core.utils.system_debug_info import system_info
@@ -43,7 +45,6 @@
  # If they don't have readline, that's fine
  pass
 
-
 def terminal_interface(interpreter, message):
  # Auto run and offline (this.. this isnt right) don't display messages.
  # Probably worth abstracting this to something like "debug_cli" at some point.
@@ -197,20 +198,19 @@ def terminal_interface(interpreter, message):
  interpreter.computer.os.notify(sanitized_message)
 
  # Speak message aloud
- if platform.system() == "Darwin" and interpreter.speak_messages:
- if voice_subprocess:
- voice_subprocess.terminate()
- voice_subprocess = subprocess.Popen(
- [
- "osascript",
- "-e",
- f'say "{sanitized_message}" using "Fred"',
- ]
- )
- else:
- pass
- # User isn't on a Mac, so we can't do this. You should tell them something about that when they first set this up.
- # Or use a universal TTS library.
+ def speak_text(text):
+ engine = pyttsx3.init()
+ engine.setProperty('rate', 150)
+ engine.setProperty('volume', 0.9)
+ engine.say(text)
+ engine.runAndWait()
+
+ def speak_async(text):
+ thread = threading.Thread(target=speak_text, args=(text,))
+ thread.start()
+
+ if interpreter.speak_messages:
+ speak_async(sanitized_message)
 
  # Assistant code blocks
  elif chunk["role"] == "assistant" and chunk["type"] == "code":

pyproject.toml

@@ -32,6 +32,7 @@ tiktoken = "^0.6.0"
 aifs = "^0.0.15"
 platformdirs = "^4.2.0"
 pydantic = "^2.6.4"
+pyttsx3 = "^2.90"
 
 # Optional [os] dependencies
 opencv-python = { version = "^4.8.1.78", optional = true }