Why the RetryOutputParser (and inner parser) does not appear in LangSmith Traces?

[jameswpm]

The retry behavior is as expected when I use the RetryOutputParser with an inner parser that throws an error. Still, the LangSmith traces don't show either the execution of the Retry parser explicitly (as it occurs with other parsers like the PydanticOutputParser) or the inner output parser that threw the error.

The last one looks very important in showing what caused the error in the traces.

Is this expected or a possible problem?

Code to reproduce:

from langchain.output_parsers import RetryOutputParser
from langchain_core.output_parsers.pydantic import PydanticOutputParser
from langchain_core.pydantic_v1 import BaseModel
from langchain_openai import ChatOpenAI
from langchain_core.runnables import RunnableLambda, RunnableParallel
from langchain_core.exceptions import OutputParserException

import os

os.environ["LANGCHAIN_TRACING_V2"] = "true"
os.environ["LANGCHAIN_PROJECT"] = "Test"
os.environ["LANGCHAIN_ENDPOINT"] = "https://api.smith.langchain.com"
os.environ["LANGCHAIN_API_KEY"] = "lsv2_**********"

from langchain_core.prompts import (
    PromptTemplate,
)

class CustomParser(PydanticOutputParser):         

    def parse(self, output: str) -> dict:
        raise OutputParserException("Failed to parse")
    
    @property
    def _type(self) -> str:
        return "custom_parser_throw_exception"
    
class TestModel(BaseModel):
    a: int
    b: str

parser = CustomParser(pydantic_object=TestModel)

model = ChatOpenAI(temperature=0)

retry_parser = RetryOutputParser.from_llm(parser=parser, llm=model, max_retries=3)

prompt = PromptTemplate(
    template="Answer the user query.\n{format_instructions}\n{query}\n",
    input_variables=["query"],
    partial_variables={"format_instructions": parser.get_format_instructions()},
)

completion_chain = prompt | model

main_chain = RunnableParallel(
    completion=completion_chain, prompt_value=prompt
) | RunnableLambda(lambda x: retry_parser.parse_with_prompt(**x))


print(main_chain.invoke({"query": "who is leo di caprios gf?"}))

Link to the trace in LangSmith: https://smith.langchain.com/public/eb7ec0ac-d165-4883-b66f-c0991adea315/r