fix(dictation): remove unwanted leading space when starting new transcription (#307)

Reset last_injected_text when recognition transitions to LISTENING state,
ensuring the first dictation segment in each session is not prefixed with
a space. Also simplify the space-prepending logic: always add a separator
space between consecutive segments (not just after punctuation), and strip
incoming text as a safety net against whisper tokenizer artifacts.

src/vocalinux/main.py

@@ -177,6 +177,7 @@ def main():
         sys.exit(1)
 
     # Now it's safe to import GTK-dependent modules
+    from .common_types import RecognitionState
     from .speech_recognition import recognition_manager
     from .text_injection import text_injector
     from .ui import tray_indicator
@@ -292,35 +293,19 @@ def main():
 
         # Create a wrapper function to track injected text for action handler
         def text_callback_wrapper(text: str):
-            """Wrapper to track injected text and handle it."""
-            # Check if we need to add a space before the new text
-            text_to_inject = text
+            """Wrapper to track injected text and handle spacing between segments."""
+            # Strip any leading/trailing whitespace from the incoming text as a
+            # safety net (whisper tokenizer sometimes prepends spaces to tokens)
+            text_to_inject = text.strip()
+            if not text_to_inject:
+                return
 
-            # Punctuation and characters that typically need a space after them
-            chars_needing_space = {
-                ".",
-                ",",
-                "!",
-                "?",
-                ";",
-                ":",
-                ")",
-                "]",
-                "}",
-                "-",
-                "_",
-                # Also quotes
-                '"',
-                "'",
-            }
-
-            # If last injected text exists and ends with a character that needs space,
-            # add a space before the new text
+            # Add a separating space between consecutive dictation segments,
+            # but never for the very first segment (avoids unwanted leading space
+            # when starting dictation in an empty text field).
             if action_handler.last_injected_text and action_handler.last_injected_text.strip():
-                last_char = action_handler.last_injected_text[-1]
-                if last_char in chars_needing_space:
-                    text_to_inject = " " + text_to_inject
-                    logger.debug(f"Added space before new text (last char: '{last_char}')")
+                text_to_inject = " " + text_to_inject
+                logger.debug("Added space separator before new segment")
 
             success = text_system.inject_text(text_to_inject)
             if success:
@@ -330,6 +315,12 @@ def main():
         speech_engine.register_text_callback(text_callback_wrapper)
         speech_engine.register_action_callback(action_handler.handle_action)
 
+        def on_state_change(state: RecognitionState):
+            if state == RecognitionState.LISTENING:
+                action_handler.set_last_injected_text("")
+
+        speech_engine.register_state_callback(on_state_change)
+
         # Initialize and start the system tray indicator
         indicator = tray_indicator.TrayIndicator(
             speech_engine=speech_engine,

tests/test_main.py

@@ -233,6 +233,7 @@ class TestMainModule(unittest.TestCase):
             mock_speech_instance.register_action_callback.assert_called_once_with(
                 mock_action_instance.handle_action
             )
+            mock_speech_instance.register_state_callback.assert_called_once()
 
             # Verify the tray indicator was started
             mock_tray_instance.run.assert_called_once()
@@ -648,5 +649,74 @@ class TestMainConfigPrecedence(unittest.TestCase):
                 self.assertEqual(call_kwargs["audio_device_index"], 2)
 
 
+class TestTextCallbackSpacing(unittest.TestCase):
+    """Test spacing logic in text_callback_wrapper."""
+
+    def _make_callback(self):
+        """Build text_callback_wrapper with mocked dependencies."""
+        from vocalinux.ui.action_handler import ActionHandler
+
+        text_system = MagicMock()
+        text_system.inject_text.return_value = True
+        action_handler = ActionHandler(text_system)
+
+        def text_callback_wrapper(text: str):
+            text_to_inject = text.strip()
+            if not text_to_inject:
+                return
+            if action_handler.last_injected_text and action_handler.last_injected_text.strip():
+                text_to_inject = " " + text_to_inject
+            success = text_system.inject_text(text_to_inject)
+            if success:
+                action_handler.set_last_injected_text(text)
+
+        return text_callback_wrapper, text_system, action_handler
+
+    def test_first_segment_has_no_leading_space(self):
+        cb, text_system, _ = self._make_callback()
+        cb("Hello world")
+        text_system.inject_text.assert_called_once_with("Hello world")
+
+    def test_subsequent_segment_gets_space_separator(self):
+        cb, text_system, _ = self._make_callback()
+        cb("Hello")
+        cb("world")
+        calls = [c.args[0] for c in text_system.inject_text.call_args_list]
+        self.assertEqual(calls, ["Hello", " world"])
+
+    def test_reset_clears_leading_space(self):
+        cb, text_system, ah = self._make_callback()
+        cb("first session")
+        ah.set_last_injected_text("")
+        text_system.inject_text.reset_mock()
+        cb("second session")
+        text_system.inject_text.assert_called_once_with("second session")
+
+    def test_whitespace_only_input_is_skipped(self):
+        cb, text_system, _ = self._make_callback()
+        cb("   ")
+        text_system.inject_text.assert_not_called()
+
+    def test_input_with_leading_space_is_stripped(self):
+        cb, text_system, _ = self._make_callback()
+        cb(" Hello world")
+        text_system.inject_text.assert_called_once_with("Hello world")
+
+    def test_multiple_segments_all_get_separators(self):
+        cb, text_system, _ = self._make_callback()
+        cb("one")
+        cb("two")
+        cb("three")
+        calls = [c.args[0] for c in text_system.inject_text.call_args_list]
+        self.assertEqual(calls, ["one", " two", " three"])
+
+    def test_space_after_punctuation_segment(self):
+        cb, text_system, _ = self._make_callback()
+        cb("Hello.")
+        cb("World")
+        calls = [c.args[0] for c in text_system.inject_text.call_args_list]
+        self.assertEqual(calls, ["Hello.", " World"])
+
+
 if __name__ == "__main__":
     unittest.main()