abstractvoice 0.6.1__tar.gz → 0.6.2__tar.gz

{abstractvoice-0.6.1 → abstractvoice-0.6.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: abstractvoice
-Version: 0.6.1
+Version: 0.6.2
 Summary: A modular Python library for voice interactions with AI systems
 Author-email: Laurent-Philippe Albou <contact@abstractcore.ai>
 License-Expression: MIT

{abstractvoice-0.6.1 → abstractvoice-0.6.2}/abstractvoice/__init__.py

@@ -29,5 +29,5 @@ warnings.filterwarnings(
 # Import the main class for public API
 from .voice_manager import VoiceManager
 
-__version__ = "0.6.1"
+__version__ = "0.6.2"
 __all__ = ["VoiceManager"]

{abstractvoice-0.6.1 → abstractvoice-0.6.2}/abstractvoice/examples/cli_repl.py

@@ -19,6 +19,7 @@ import threading
 import time
 import requests
 from abstractvoice import VoiceManager
+from abstractvoice.text_sanitize import sanitize_markdown_for_speech
 
 
 # ANSI color codes
@@ -1563,6 +1564,10 @@ class VoiceREPL(cmd.Cmd):
         if not self.voice_manager:
             return
 
+        # LLM output often contains Markdown. Strip the most common formatting
+        # tokens so TTS stays natural (do not change what is printed).
+        speak_text = sanitize_markdown_for_speech(text)
+
         is_clone = bool(self.current_tts_voice)
         if not is_clone:
             # Offline-first: Piper voices must be explicitly cached. Provide a clear
@@ -1583,7 +1588,7 @@ class VoiceREPL(cmd.Cmd):
         try:
             if is_clone:
                 ind.start()
-            self.voice_manager.speak(text, voice=self.current_tts_voice)
+            self.voice_manager.speak(speak_text, voice=self.current_tts_voice)
 
             if not is_clone:
                 return

abstractvoice-0.6.2/abstractvoice/text_sanitize.py

@@ -0,0 +1,33 @@
+"""Lightweight text sanitizers for better speech UX.
+
+Design goals:
+- No heavy dependencies (regex only).
+- Deterministic and conservative (only strip the most common Markdown syntax).
+"""
+
+from __future__ import annotations
+
+import re
+
+
+_MD_HEADER_RE = re.compile(r"(?m)^[ \t]*#{1,5}(?!#)[ \t]*(\S.*)$")
+_MD_BOLD_RE = re.compile(r"\*\*([^*\n]+?)\*\*")
+# Avoid matching "**bold**" as italic. The bold pass runs first, but keep this safe.
+_MD_ITALIC_RE = re.compile(r"(?<!\*)\*([^*\n]+?)\*(?!\*)")
+
+
+def sanitize_markdown_for_speech(text: str) -> str:
+    """Remove common Markdown syntax that sounds bad in TTS.
+
+    Intentionally minimal:
+    - Strip ATX headers at the start of a line: "#", "##", ... "#####"
+    - Strip emphasis markers: "**bold**" and "*italic*"
+    """
+    if not text:
+        return ""
+
+    out = str(text)
+    out = _MD_HEADER_RE.sub(r"\1", out)
+    out = _MD_BOLD_RE.sub(r"\1", out)
+    out = _MD_ITALIC_RE.sub(r"\1", out)
+    return out

{abstractvoice-0.6.1 → abstractvoice-0.6.2}/abstractvoice.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: abstractvoice
-Version: 0.6.1
+Version: 0.6.2
 Summary: A modular Python library for voice interactions with AI systems
 Author-email: Laurent-Philippe Albou <contact@abstractcore.ai>
 License-Expression: MIT

{abstractvoice-0.6.1 → abstractvoice-0.6.2}/abstractvoice.egg-info/SOURCES.txt

@@ -8,6 +8,7 @@ abstractvoice/dependency_check.py
 abstractvoice/prefetch.py
 abstractvoice/recognition.py
 abstractvoice/stop_phrase.py
+abstractvoice/text_sanitize.py
 abstractvoice/voice_manager.py
 abstractvoice.egg-info/PKG-INFO
 abstractvoice.egg-info/SOURCES.txt
@@ -71,6 +72,7 @@ tests/test_repl_interrupt_cloned_voice_does_not_resume_old_audio.py
 tests/test_stop_phrase_continuous_detector.py
 tests/test_stop_phrase_matching_is_tolerant.py
 tests/test_stop_speaking_restores_recognizer_state.py
+tests/test_text_sanitize_markdown.py
 tests/test_voice_clone_store_delete_rename.py
 tests/test_voice_cloner_engine_dispatch.py
 tests/test_voice_cloning_integration_hal9000.py

abstractvoice-0.6.2/tests/test_text_sanitize_markdown.py

@@ -0,0 +1,29 @@
+from __future__ import annotations
+
+from abstractvoice.text_sanitize import sanitize_markdown_for_speech
+
+
+def test_sanitize_removes_bold_single_word() -> None:
+    assert sanitize_markdown_for_speech("Hello **world**!") == "Hello world!"
+
+
+def test_sanitize_removes_bold_multiple_words() -> None:
+    assert sanitize_markdown_for_speech("**sentence with multiple words**") == "sentence with multiple words"
+
+
+def test_sanitize_removes_italic_single_word() -> None:
+    assert sanitize_markdown_for_speech("Use *italics* here.") == "Use italics here."
+
+
+def test_sanitize_removes_italic_multiple_words() -> None:
+    assert sanitize_markdown_for_speech("Use *a sentence with words* here.") == "Use a sentence with words here."
+
+
+def test_sanitize_strips_markdown_headers_keep_text() -> None:
+    txt = "# Title\n## Subtitle\n##### Small header\nNot a header\n###"
+    assert sanitize_markdown_for_speech(txt) == "Title\nSubtitle\nSmall header\nNot a header\n###"
+
+
+def test_sanitize_does_not_touch_bullet_marker_without_closing_star() -> None:
+    assert sanitize_markdown_for_speech("* item") == "* item"
+