Error when using `AUTO` with recursive depth issue

[pramitchoudhary]

Running into the below error for the attached file.

result = parse(".../Lalayan_vs_Garland.pdf", parser_type="AUTO", page_nums=1, pages_per_split=1)

Error,

2025-06-06 12:34:54.736 | DEBUG    | lexoid.core.utils:router:587 - Using STATIC_PARSE for PDF without images.
2025-06-06 12:34:54.737 | DEBUG    | lexoid.api:parse_chunk:61 - Auto-detected parser type: ParserType.STATIC_PARSE
2025-06-06 12:34:54.737 | DEBUG    | lexoid.api:parse_chunk:67 - Using static parser
---------------------------------------------------------------------------
RecursionError                            Traceback (most recent call last)
File [~/.jupyter_venv/lib/python3.11/site-packages/pdfplumber/utils/pdfinternals.py:78](http://localhost:8888/lab/tree/projects/experiments/notebooks/LLMS/~/.jupyter_venv/lib/python3.11/site-packages/pdfplumber/utils/pdfinternals.py#line=77), in resolve_all(x)
     77 try:
---> 78     return resolve_all(resolved)
     79 except RecursionError as e:
...
MalformedPDFException: maximum recursion depth exceeded while calling a Python object

• Works fine in LLM_PARSE mode
• Tried with a different version of the same doc, that works.
• Not sure if the attached file got corrupted, but it did manage to read the other version fine.

Lalayan_vs_Garland.pdf

[dilithjay]

This seems to be an issue with PDFPlumber: jsvine/pdfplumber#638

I tried the solution mentioned here: jsvine/pdfplumber#638 (comment) and it worked but it requires editing the PDFPlumber source code. So, for now I think all we can do is to catch the exception.

[pramitchoudhary]

As we were discussing,

• Let's catch the exception and switch to LLM_PARSE in AUTO mode for now.
• For explicit STATIC_PARSE, we catch the exception and report.
  Or how do u feel about adding fail-safe (default: True) param n for both scenarios if an exception happens, we switch to LLM_PARSE and then later decide on alternatives for STATIC_PARSE, thoughts?

[pramitchoudhary]

@dilithjay we can close this, or is something pending?
#104