From 3f8f3de28e9e33cbf0889111a0362116de7a8928 Mon Sep 17 00:00:00 2001
From: Davide Menini <48685774+dmenini@users.noreply.github.com>
Date: Fri, 1 Sep 2023 02:11:52 +0200
Subject: [PATCH] fix (parsers/json): do not escape double quotes if already
 escaped (#9916)

This PR fixes an issues I found when upgrading to a more recent version
of Langchain. I was using 0.0.142 before, and this issue popped up
already when the `_custom_parser` was added to `output_parsers/json`.

Anyway, the issue is that the parser tries to escape quotes when they
are double-escaped (e.g. `\\"`), leading to OutputParserException.
This is particularly undesired in my app, because I have an Agent that
uses a single input Tool, which expects as input a JSON string with the
structure:
```python
{
    "foo": string,
    "bar": string
}
```
The LLM (GPT3.5) response is (almost) always something like
`"action_input": "{\\"foo\\": \\"bar\\", \\"bar\\": \\"foo\\"}"` and
since the upgrade this is not correctly parsed.

---------

Co-authored-by: taamedag <Davide.Menini@swisscom.com>
---
 .../langchain/output_parsers/json.py          |  2 +-
 .../unit_tests/output_parsers/test_json.py    | 52 +++++++++++++++++++
 2 files changed, 53 insertions(+), 1 deletion(-)

diff --git a/libs/langchain/langchain/output_parsers/json.py b/libs/langchain/langchain/output_parsers/json.py
index f0f653eca2..7465aba2fe 100644
--- a/libs/langchain/langchain/output_parsers/json.py
+++ b/libs/langchain/langchain/output_parsers/json.py
@@ -13,7 +13,7 @@ def _replace_new_line(match: re.Match[str]) -> str:
     value = re.sub(r"\n", r"\\n", value)
     value = re.sub(r"\r", r"\\r", value)
     value = re.sub(r"\t", r"\\t", value)
-    value = re.sub('"', r"\"", value)
+    value = re.sub(r'(?<!\\)"', r"\"", value)
 
     return match.group(1) + value + match.group(3)
 
diff --git a/libs/langchain/tests/unit_tests/output_parsers/test_json.py b/libs/langchain/tests/unit_tests/output_parsers/test_json.py
index afaa4bf64f..21bc600fd0 100644
--- a/libs/langchain/tests/unit_tests/output_parsers/test_json.py
+++ b/libs/langchain/tests/unit_tests/output_parsers/test_json.py
@@ -67,6 +67,34 @@ JSON_WITH_MARKDOWN_CODE_BLOCK_AND_NEWLINES = """```json
 }
 ```"""
 
+JSON_WITH_UNESCAPED_QUOTES_IN_NESTED_JSON = """```json
+{
+    "action": "Final Answer",
+    "action_input": "{"foo": "bar", "bar": "foo"}"
+}
+```"""
+
+JSON_WITH_ESCAPED_QUOTES_IN_NESTED_JSON = """```json
+{
+    "action": "Final Answer",
+    "action_input": "{\"foo\": \"bar\", \"bar\": \"foo\"}"
+}
+```"""
+
+JSON_WITH_PYTHON_DICT = """```json
+{
+    "action": "Final Answer",
+    "action_input": {"foo": "bar", "bar": "foo"}
+}
+```"""
+
+JSON_WITH_ESCAPED_DOUBLE_QUOTES_IN_NESTED_JSON = """```json
+{
+    "action": "Final Answer",
+    "action_input": "{\\"foo\\": \\"bar\\", \\"bar\\": \\"foo\\"}"
+}
+```"""
+
 NO_TICKS = """{
     "foo": "bar"
 }"""
@@ -131,3 +159,27 @@ def test_parse_json_with_code_blocks() -> None:
         "action": "Final Answer",
         "action_input": '```bar\n<div id="1" class="value">\n\ttext\n</div>```',
     }
+
+
+TEST_CASES_ESCAPED_QUOTES = [
+    JSON_WITH_UNESCAPED_QUOTES_IN_NESTED_JSON,
+    JSON_WITH_ESCAPED_QUOTES_IN_NESTED_JSON,
+    JSON_WITH_ESCAPED_DOUBLE_QUOTES_IN_NESTED_JSON,
+]
+
+
+@pytest.mark.parametrize("json_string", TEST_CASES_ESCAPED_QUOTES)
+def test_parse_nested_json_with_escaped_quotes(json_string: str) -> None:
+    parsed = parse_json_markdown(json_string)
+    assert parsed == {
+        "action": "Final Answer",
+        "action_input": '{"foo": "bar", "bar": "foo"}',
+    }
+
+
+def test_parse_json_with_python_dict() -> None:
+    parsed = parse_json_markdown(JSON_WITH_PYTHON_DICT)
+    assert parsed == {
+        "action": "Final Answer",
+        "action_input": {"foo": "bar", "bar": "foo"},
+    }