github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #14482] 0.17.2 can't read PDFs on macOS #55906

New Issue
Closed
opened 2026-04-29 09:55:32 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-29 09:55:32 -05:00
Owner
Copy Link

Originally created by @xmddmx on GitHub (Feb 27, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14482

What is the issue?

Tried to read the attached PDF file using several models, and all claim the PDF is damaged.

gemma3:27b: Invalid File Format The system tried to open the file as a PDF but failed because: It lacks the required end-of-file marker (%%EOF), which is a standard component of PDFs.

qwen3.5:35b: 1. File Status: Unreadable
Error Message: failed to create PDF reader: not a PDF file: missing %%EOF

gpt-oss:20b : It looks like the file you attached isn’t being recognized as a valid PDF—there’s no proper PDF header/footer (the missing %%EOF marker). Because of that I can’t extract any text or metadata from it.

Onkyo.pdf

Relevant log output

nothing relevant showing up in server.log or app.log

OS

macOS

GPU

Apple

CPU

Apple

Ollama version

0.17.2

Originally created by @xmddmx on GitHub (Feb 27, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14482 ### What is the issue? Tried to read the attached PDF file using several models, and all claim the PDF is damaged. gemma3:27b: Invalid File Format The system tried to open the file as a PDF but failed because: It lacks the required end-of-file marker (%%EOF), which is a standard component of PDFs. qwen3.5:35b: 1. File Status: Unreadable Error Message: failed to create PDF reader: not a PDF file: missing %%EOF gpt-oss:20b : It looks like the file you attached isn’t being recognized as a valid PDF—there’s no proper PDF header/footer (the missing %%EOF marker). Because of that I can’t extract any text or metadata from it. [Onkyo.pdf](https://github.com/user-attachments/files/25590784/Onkyo.pdf) ### Relevant log output ```shell nothing relevant showing up in server.log or app.log ``` ### OS macOS ### GPU Apple ### CPU Apple ### Ollama version 0.17.2
GiteaMirror added the bug label 2026-04-29 09:55:32 -05:00
GiteaMirror commented 2026-04-29 09:55:35 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 27, 2026):

Models don't understand PDF files, what happens when you upload a PDF to a model is that the client extracts the text from the file and sends that to the model. In this case, the client has a PDF extractor that is throwing an error because the PDF is badly formed, and the error message is being sent to the model.

Looking at the contents of the file, we see that there is what looks like a piece of HTML appended to the file after the %%EOF marker:

trailer
<</Size 9719/Root 4935 0 R/Info 9717 0 R/ID[<0527547B0E58A8B1F9640678CBC18361><867A241D7C81DB44888F7C5E6EFF6A8C>]>>
startxref
4102427
%%EOF
<!DOCTYPE html PUBLIC "-//IETF//DTD HTML 2.0//EN">
<HTML>
   <HEAD>
      <TITLE>
         A Small Hello
      </TITLE>
   </HEAD>
<BODY>
   <H1>Hi, I'm super! Thanks for asking!</H1>
   <P>Experiencing Errors, Please Hold On!</P>
</BODY>
</HTML>

So while the PDF extractor in the ollama client could be more resilient, this PDF is malformed. You can fix this by removing this trailing junk, either by loading it into a PDF editor and saving it, or just truncating the file by removing the last 261 bytes:

cp Onkyo.pdf Onkyo-fixed.pdf 
truncate -s -261 Onkyo-fixed.pdf
<!-- gh-comment-id:3972781363 --> @rick-github commented on GitHub (Feb 27, 2026): Models don't understand PDF files, what happens when you upload a PDF to a model is that the client extracts the text from the file and sends that to the model. In this case, the client has a PDF extractor that is throwing an error because the PDF is badly formed, and the error message is being sent to the model. Looking at the contents of the file, we see that there is what looks like a piece of HTML appended to the file after the `%%EOF` marker: ``` trailer <</Size 9719/Root 4935 0 R/Info 9717 0 R/ID[<0527547B0E58A8B1F9640678CBC18361><867A241D7C81DB44888F7C5E6EFF6A8C>]>> startxref 4102427 %%EOF <!DOCTYPE html PUBLIC "-//IETF//DTD HTML 2.0//EN"> <HTML> <HEAD> <TITLE> A Small Hello </TITLE> </HEAD> <BODY> <H1>Hi, I'm super! Thanks for asking!</H1> <P>Experiencing Errors, Please Hold On!</P> </BODY> </HTML> ``` So while the PDF extractor in the ollama client could be more resilient, this PDF is malformed. You can fix this by removing this trailing junk, either by loading it into a PDF editor and saving it, or just truncating the file by removing the last 261 bytes: ```console cp Onkyo.pdf Onkyo-fixed.pdf truncate -s -261 Onkyo-fixed.pdf ```
GiteaMirror commented 2026-04-29 09:55:36 -05:00
Author
Owner
Copy Link

@xmddmx commented on GitHub (Feb 27, 2026):

Oh! I had no idea that was a nonstandard PDF file. Consider this a non issue.

<!-- gh-comment-id:3973336367 --> @xmddmx commented on GitHub (Feb 27, 2026): Oh! I had no idea that was a nonstandard PDF file. Consider this a non issue.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#55906
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?