langchain[patch]: make BooleanOutputParser check words not substrings (#20064)

- **Description**: fixes BooleanOutputParser detecting sub-words ("NOW
this is likely (YES)" -> `True`, not `AmbiguousError`)
- **Issue(s)**: fixes #11408 (follow-up to #17810)
- **Dependencies**: None
- **GitHub handle**: @casperdcl

<!-- if unreviewd after a few days, @-mention one of baskaryan, efriis,
eyurtsev, hwchase17 -->

- [x] **Add tests and docs**: If you're adding a new integration, please
include
1. a test for the integration, preferably unit tests that do not rely on
network access,
2. an example notebook showing its use. It lives in
`docs/docs/integrations` directory.
- [ ] **Lint and test**: Run `make format`, `make lint` and `make test`
from the root of the package(s) you've modified. See contribution
guidelines for more: https://python.langchain.com/docs/contributing/

---------

Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>

libs/langchain/langchain/output_parsers/boolean.py

@@ -1,3 +1,5 @@
+import re
+
 from langchain_core.output_parsers import BaseOutputParser
 
 
@@ -17,22 +19,27 @@ class BooleanOutputParser(BaseOutputParser[bool]):
 
         Returns:
             boolean
-
         """
-        cleaned_upper_text = text.strip().upper()
+        regexp = rf"\b({self.true_val}|{self.false_val})\b"
-        if (
+
-            self.true_val.upper() in cleaned_upper_text
+        truthy = {
-            and self.false_val.upper() in cleaned_upper_text
+            val.upper()
-        ):
+            for val in re.findall(regexp, text, flags=re.IGNORECASE | re.MULTILINE)
+        }
+        if self.true_val.upper() in truthy:
+            if self.false_val.upper() in truthy:
                 raise ValueError(
-                f"Ambiguous response. Both {self.true_val} and {self.false_val} in "
+                    f"Ambiguous response. Both {self.true_val} and {self.false_val} "
-                f"received: {text}."
+                    f"in received: {text}."
                 )
-        elif self.true_val.upper() in cleaned_upper_text:
             return True
-        elif self.false_val.upper() in cleaned_upper_text:
+        elif self.false_val.upper() in truthy:
+            if self.true_val.upper() in truthy:
+                raise ValueError(
+                    f"Ambiguous response. Both {self.true_val} and {self.false_val} "
+                    f"in received: {text}."
+                )
             return False
-        else:
         raise ValueError(
             f"BooleanOutputParser expected output value to include either "
             f"{self.true_val} or {self.false_val}. Received {text}."

libs/langchain/tests/unit_tests/output_parsers/test_boolean_parser.py

@@ -1,3 +1,5 @@
+import pytest
+
 from langchain.output_parsers.boolean import BooleanOutputParser
 
 
@@ -24,16 +26,16 @@ def test_boolean_output_parser_parse() -> None:
     result = parser.parse("Not relevant (NO)")
     assert result is False
 
-    # Test ambiguous input
+    # Test valid input
-    try:
+    result = parser.parse("NOW this is relevant (YES)")
-        parser.parse("yes and no")
+    assert result is True
-        assert False, "Should have raised ValueError"
-    except ValueError:
-        pass
 
-    # Test invalid input
+    # Test ambiguous input
-    try:
+    with pytest.raises(ValueError):
-        parser.parse("INVALID")
+        parser.parse("YES NO")
-        assert False, "Should have raised ValueError"
+
-    except ValueError:
+    with pytest.raises(ValueError):
-        pass
+        parser.parse("NO YES")
+    # Bad input
+    with pytest.raises(ValueError):
+        parser.parse("BOOM")