🚨 JSON Parsing & PDF Processing Issue – Rate Limit & Resume Extraction Errors

:pushpin: Description of the Issue:
Users are experiencing two distinct but related issues when processing PDF resumes in PickAxe, regardless of their formatting.


Error 1: JSON Parsing Error

  • Error Message: "Unexpected token 'd', data: {'id'... is not valid JSON"
  • Occurs When:
    • Uploading a PDF resume (whether exported from LinkedIn or reformatted).
    • AI requests confirmation to proceed with resume parsing.
    • User responds with “Yes please”, and the error appears.
  • Impact:
    • PickAxe fails to process the extracted text, preventing further AI execution.

:small_blue_diamond: Screenshot: [Attached: Pickaxe error.jpg]


Error 2: Rate Limit Error

  • Error Message: "Something went wrong! The owner of this tool has been notified of this error. Please try again later. (Rate limit error)"
  • Occurs When:
    • After retrying the resume processing step following the JSON error.
    • AI attempts to optimize the resume and generate a cover letter.
  • Impact:
    • Prevents users from continuing the workflow.
    • Multiple retries result in the same rate limit block.

:small_blue_diamond: Screenshot: [Attached: Pickaxe error 2.jpg]


Steps to Reproduce the Issue:

  1. Upload a PDF resume (exported from LinkedIn, Word, or Google Docs).
  2. PickAxe requests user confirmation before proceeding.
  3. User confirms with “Yes please”.
  4. PickAxe returns the JSON parsing error.
  5. User retries, leading to the Rate Limit error.

:pushpin: Additional Testing Insights

:small_blue_diamond: We tested whether the two-column format was causing the issue by:
:white_check_mark: Manually converting the resume into a single-column format using Google Docs and Word.
:white_check_mark: Saving it as a single-column PDF (removing any table structures).
:white_check_mark: Testing a plain text version of the PDF (removing all formatting).

:bulb: RESULT: The error still occurs, even in single-column PDFs.
:pushpin: Conclusion: The issue is not just related to two-column layouts, but rather how PickAxe processes PDFs in general.


:pushpin: Expected Behavior:

:white_check_mark: PickAxe should correctly extract contact information, job titles, company names, and employment dates from any properly formatted PDF.
:white_check_mark: AI should structure the resume into an ATS-friendly format without JSON parsing issues.
:white_check_mark: Users should be able to proceed with resume optimization without hitting a rate limit block.


:pushpin: Actual Behavior:

:x: PickAxe fails to read text properly from PDFs, breaking the workflow.
:x: Rate limit blocks further processing after the initial failure.
:x: The issue persists across different versions of PDFs, whether two-column or single-column.


:small_blue_diamond: Troubleshooting Steps Taken:

:white_check_mark: Tested different versions of the PDF (LinkedIn direct download, Google Docs, and Word).
:white_check_mark: Manually extracted text, confirming that AI can read the content outside PickAxe.
:white_check_mark: Tested single-column vs. two-column resumes, but the error persists.
:white_check_mark: Attempted different file input formats, but PickAxe still fails with PDFs.


:pushpin: Possible Cause:

  • PickAxe may have difficulty handling PDFs in general, not just two-column layouts.
  • PDF text extraction errors (misplaced job titles, missing fields, or encoding issues) might be causing JSON response failures.
  • The Rate Limit Error is likely triggered by multiple failed attempts, compounding the problem.

:pushpin: Suggested Fixes for PickAxe Team:

:small_blue_diamond: Improve how PickAxe handles PDFs, regardless of column structure.
:small_blue_diamond: Ensure that AI can properly extract structured text from PDFs.
:small_blue_diamond: Enhance JSON parsing to handle unexpected formatting.
:small_blue_diamond: Allow retries without triggering a rate limit block.


:pushpin: Final Thoughts & Expanded Observations

After extensive testing, we believe the core issue is how PickAxe reads PDFs in general, not just how it handles multi-column formats. While our initial assumption was that two-column layouts were causing the problem, the fact that the error persists even after converting the resume to a plain text, single-column format suggests that the PDF extraction process itself is flawed.

This means that PickAxe may not be handling PDFs consistently, leading to:
:one: Corrupt or incomplete data being passed to the AI, which results in the JSON parsing failure.
:two: Misalignment of key resume elements (e.g., contact details, dates, job titles) that causes unexpected behavior.
:three: Rate limiting due to repeated failures, preventing users from trying alternative solutions.

This issue significantly impacts usability for anyone relying on PickAxe to process resumes or documents in PDF format. Addressing this problem will improve stability for all users, not just those working with LinkedIn resumes.



Commonly Used Words for Searchability
:white_check_mark: “PickAxe PDF error”
:white_check_mark: “JSON parsing issue”
:white_check_mark: “Resume processing failure”
:white_check_mark: “Rate limit error”
:white_check_mark: “Two-column PDF issue”

1 Like

Thank you a lot for this full and detailed report about the issue.
We’ll start working on finding the reason why it happens to your Pickaxes.

I’ve just sent you a DM message for some details. Please take a look.

3 Likes

Thank you for taking a look at this detailed issue. I’m hoping it will answer the question as to why my pickaxes generally don’t intuitively recognise pdf uploads. I have to tell my users to say ‘upload done’ or ‘done’- otherwise it’s crickets on the page. Pdfs are my one biggest issue.

1 Like

The way the system works, the end-user always has to send a message in order fort he AI to respond. Even something as simple as “ok”.

Hi @admin_mike,
I have said, please go ahead with the resume and it cannot read the PDF as of Feb 3rd. @dev_vlad If you have questions, please let me know so I can assist you in getting this issue resolved. Thank you so much!