fix(helpers): preserve SSML breaks with additional attributes

src/deepgram/helpers/text_builder.py

@@ -297,11 +297,17 @@ def replace_phoneme(match):
     ssml_text = re.sub(phoneme_pattern, replace_phoneme, ssml_text)
 
     # Handle <break> tags
-    break_pattern = r'<break\s+time=["\'](\d+(?:\.\d+)?)(ms|s)["\']\s*/>'
+    break_pattern = r"<break\s+([^<>]*?)\s*/>"
 
     def replace_break(match):
-        value = float(match.group(1))
-        unit = match.group(2)
+        time_match = re.search(
+            r"(?:^|\s)time\s*=\s*([\"'])(\d+(?:\.\d+)?)(ms|s)\1(?=\s|$)",
+            match.group(1),
+        )
+        if time_match is None:
+            return match.group(0)
+        value = float(time_match.group(2))
+        unit = time_match.group(3)
 
         duration_ms = value * 1000 if unit == "s" else value
         if not duration_ms.is_integer():

tests/custom/test_text_builder.py

@@ -554,3 +554,19 @@ def test_standalone_function_workflow(self):
         assert '"pronounce": "ˌæzəˈθaɪəpriːn"' in text
         assert '"word": "dupilumab"' in text
         assert '"pronounce": "duːˈpɪljuːmæb"' in text
+
+
+@pytest.mark.parametrize(
+    "markup",
+    [
+        '<break time = "500ms"/>',
+        "<break time='0.5s' strength='medium'/>",
+        '<break strength="medium" time="500ms"/>',
+        '<break\nstrength="medium"\ntime = "0.5s" />',
+    ],
+)
+def test_ssml_break_time_accepts_attribute_spacing_and_order(markup):
+    ssml = f"<speak>Hello{markup}world</speak>"
+
+    assert ssml_to_deepgram(ssml) == r"Hello\{pause:500ms\}world"
+    assert TextBuilder().from_ssml(ssml).build() == r"Hello\{pause:500ms\}world"