Merge 4eab312c35 into 05d8f11085

gateway/platforms/helpers.py

@@ -157,8 +157,8 @@ class TextBatchAggregator:
 # Pre-compiled regexes for performance
 _RE_BOLD = re.compile(r"\*\*(.+?)\*\*", re.DOTALL)
 _RE_ITALIC_STAR = re.compile(r"\*(.+?)\*", re.DOTALL)
-_RE_BOLD_UNDER = re.compile(r"__(.+?)__", re.DOTALL)
-_RE_ITALIC_UNDER = re.compile(r"_(.+?)_", re.DOTALL)
+_RE_BOLD_UNDER = re.compile(r"(?<![a-zA-Z0-9])__(?=[^_\s])(.+?)(?<=[^_])__(?![a-zA-Z0-9])", re.DOTALL)
+_RE_ITALIC_UNDER = re.compile(r"(?<![a-zA-Z0-9])_(?=[^_\s])(.+?)(?<=[^_])_(?![a-zA-Z0-9])", re.DOTALL)
 _RE_CODE_BLOCK = re.compile(r"```[a-zA-Z0-9_+-]*\n?")
 _RE_INLINE_CODE = re.compile(r"`(.+?)`")
 _RE_HEADING = re.compile(r"^#{1,6}\s+", re.MULTILINE)

tests/gateway/test_helpers_markdown_underscore.py

@@ -0,0 +1,60 @@
+"""Tests for underscore markdown stripping in gateway helpers.
+
+The blanket _(.+?)_ and __(.+?)__ patterns incorrectly consumed
+snake_case identifiers like send_as_bot and user_id.  The fix adds
+lookbehind/lookahead boundaries so underscores adjacent to
+alphanumeric characters are not treated as markdown formatting.
+
+Mirrors fixes already tested in hermes_cli/test_cli_markdown_rendering.py.
+"""
+import unittest
+
+from gateway.platforms.helpers import strip_markdown
+
+
+class TestGatewayUnderscoreRegex(unittest.TestCase):
+    """Verify markdown stripping preserves snake_case identifiers."""
+
+    def test_snake_case_preserved(self):
+        text = "Set send_as_bot to true and check user_id"
+        result = strip_markdown(text)
+        self.assertIn("send_as_bot", result)
+        self.assertIn("user_id", result)
+
+    def test_bold_underscore_stripped(self):
+        result = strip_markdown("Here is __bold text__ for you")
+        self.assertIn("bold text", result)
+        self.assertNotIn("__bold", result)
+
+    def test_italic_underscore_stripped(self):
+        result = strip_markdown("Here is _italic_ text")
+        self.assertIn("italic", result)
+        self.assertNotIn("_italic_", result)
+
+    def test_double_underscore_in_identifier_preserved(self):
+        """Double underscores embedded in alphanumeric context should survive."""
+        # e.g. x__y where underscores are between alphanumeric chars
+        result = strip_markdown("Check my_var__name for details")
+        self.assertIn("my_var__name", result)
+
+    def test_config_keys_preserved(self):
+        result = strip_markdown("Set max_tokens to 4096 and api_base_url to localhost")
+        self.assertIn("max_tokens", result)
+        self.assertIn("api_base_url", result)
+
+    def test_asterisk_bold_unaffected(self):
+        result = strip_markdown("This is **bold** text")
+        self.assertIn("bold", result)
+        self.assertNotIn("**", result)
+
+    def test_mixed_formatting_and_identifiers(self):
+        result = strip_markdown("Use **bold** and set send_as_bot to _true_")
+        self.assertIn("send_as_bot", result)
+        self.assertNotIn("**", result)
+
+    def test_multiple_snake_case_in_one_line(self):
+        text = "Configure thread_id, session_key, and platform_name"
+        result = strip_markdown(text)
+        self.assertIn("thread_id", result)
+        self.assertIn("session_key", result)
+        self.assertIn("platform_name", result)