From 1c120e9615167db46f9eaf815a2d42348ee130f1 Mon Sep 17 00:00:00 2001
From: Erick Friis <erick@langchain.dev>
Date: Mon, 16 Dec 2024 10:25:16 -0800
Subject: [PATCH] core: xml output parser tags docstring (#28745)

---
 libs/core/langchain_core/output_parsers/xml.py | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/libs/core/langchain_core/output_parsers/xml.py b/libs/core/langchain_core/output_parsers/xml.py
index 3df9750f8e0..4fef349c498 100644
--- a/libs/core/langchain_core/output_parsers/xml.py
+++ b/libs/core/langchain_core/output_parsers/xml.py
@@ -142,6 +142,20 @@ class XMLOutputParser(BaseTransformOutputParser):
     """Parse an output using xml format."""
 
     tags: Optional[list[str]] = None
+    """Tags to tell the LLM to expect in the XML output.
+
+    Note this may not be perfect depending on the LLM implementation.
+
+    For example, with tags=["foo", "bar", "baz"]:
+            1. A well-formatted XML instance:
+                "<foo>\n   <bar>\n      <baz></baz>\n   </bar>\n</foo>"
+
+            2. A badly-formatted XML instance (missing closing tag for 'bar'):
+                "<foo>\n   <bar>\n   </foo>"
+
+            3. A badly-formatted XML instance (unexpected 'tag' element):
+                "<foo>\n   <tag>\n   </tag>\n</foo>"
+    """
     encoding_matcher: re.Pattern = re.compile(
         r"<([^>]*encoding[^>]*)>\n(.*)", re.MULTILINE | re.DOTALL
     )