From b70933984e10e8ca786ce85c0e5243c927a75756 Mon Sep 17 00:00:00 2001
From: Nishchay Mahor <nishchaymahor@gmail.com>
Date: Sat, 4 Jul 2026 23:57:51 -0700
Subject: [PATCH] fix: handle SSML phoneme attribute order in ssml_to_deepgram

ssml_to_deepgram only matched <phoneme> tags with alphabet before ph, so
<phoneme ph="..." alphabet="ipa">word</phoneme> silently lost its
pronunciation and fell through to the generic tag strip, leaving the bare
word. SSML attribute order is not significant, so match the attributes as a
group and pull ph out of it. Add a regression test.
---
 src/deepgram/helpers/text_builder.py | 12 +++++++++---
 tests/custom/test_text_builder.py    | 12 ++++++++++--
 2 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/src/deepgram/helpers/text_builder.py b/src/deepgram/helpers/text_builder.py
index b859b217..3d766414 100644
--- a/src/deepgram/helpers/text_builder.py
+++ b/src/deepgram/helpers/text_builder.py
@@ -239,12 +239,18 @@ def ssml_to_deepgram(ssml_text: str) -> str:
     # Parse XML fragments manually to handle mixed content
     # Use regex to find and replace SSML elements
 
-    # Handle <phoneme> tags
-    phoneme_pattern = r'<phoneme\s+alphabet=["\']ipa["\']\s+ph=["\'](.*?)["\']\s*>(.*?)</phoneme>'
+    # Handle <phoneme> tags. Attribute order is not significant in SSML, so match
+    # the attributes as a group and pull `ph` out of it rather than requiring
+    # `alphabet` before `ph` (which silently dropped the pronunciation otherwise).
+    phoneme_pattern = r"<phoneme\s+([^>]*?)\s*>(.*?)</phoneme>"
 
     def replace_phoneme(match):
-        ipa = match.group(1)
+        attributes = match.group(1)
         word = match.group(2)
+        ph_match = re.search(r'ph=["\'](.*?)["\']', attributes)
+        if ph_match is None:
+            return word
+        ipa = ph_match.group(1)
         return json.dumps({"word": word, "pronounce": ipa}, ensure_ascii=False)
 
     ssml_text = re.sub(phoneme_pattern, replace_phoneme, ssml_text)
diff --git a/tests/custom/test_text_builder.py b/tests/custom/test_text_builder.py
index 77a7ed1b..45ca7009 100644
--- a/tests/custom/test_text_builder.py
+++ b/tests/custom/test_text_builder.py
@@ -234,10 +234,18 @@ def test_basic_phoneme(self):
         """Test converting basic phoneme tag"""
         ssml = '<phoneme alphabet="ipa" ph="ˌæzəˈθaɪəpriːn">azathioprine</phoneme>'
         result = ssml_to_deepgram(ssml)
-        
+
         assert '"word": "azathioprine"' in result
         assert '"pronounce": "ˌæzəˈθaɪəpriːn"' in result
-    
+
+    def test_phoneme_attribute_order_independent(self):
+        """ph before alphabet must work too (SSML attribute order is not significant)"""
+        ssml = '<phoneme ph="ˌæzəˈθaɪəpriːn" alphabet="ipa">azathioprine</phoneme>'
+        result = ssml_to_deepgram(ssml)
+
+        assert '"word": "azathioprine"' in result
+        assert '"pronounce": "ˌæzəˈθaɪəpriːn"' in result
+
     def test_basic_break(self):
         """Test converting break tag (milliseconds)"""
         ssml = '<break time="500ms"/>'