[GH-ISSUE #4696] [BUG]: Markdown File Upload Causes Critical Application Crash and 503 Service Error #2976

Closed
opened 2026-02-22 18:32:06 -05:00 by yindo · 4 comments
Owner

Originally created by @mshahid538 on GitHub (Nov 27, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4696

How are you running AnythingLLM?

Not listed

What happened?

What happened:
Upon attempting to upload a standard Markdown (.md) file via the "Website Builder" feature in the AnythingLLM workspace, the entire application immediately crashed. The front-end displayed a JSON parsing error: "unexpected token '<'", suggesting the API response was an HTML or malformed payload instead of valid JSON. This failure resulted in the application becoming completely inaccessible, showing a persistent "503 Service Temporarily Offline nginx" error. We had to restart the entire application/deployment to restore functionality.

What did you expect to happen:
I expected the Markdown file to upload successfully, be processed as a source document, and be indexed for use in the AnythingLLM workspace without any interruption or crash to the core service.

Are there known steps to reproduce?

Yes, the bug is reliably reproducible with the following steps on a self-hosted instance:

  1. Deployment Context: Use AnythingLLM v1.19 deployed in a self-hosted environment (our instance is running at ai.weown.app).
  2. Navigate: Go to an existing or new Workspace.
  3. Initiate Upload: Select the "Website Builder" upload method.
  4. Select File: Upload a typical Markdown file (.md).
  5. Observe Failure: The system immediately fails the upload process. The UI shows the "unexpected token '<'" JSON error, and the entire backend service enters the "503 Service Temporary Offline nginx" state, requiring a restart to recover.
Originally created by @mshahid538 on GitHub (Nov 27, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4696 ### How are you running AnythingLLM? Not listed ### What happened? **What happened:** Upon attempting to upload a standard Markdown (.md) file via the "**Website Builder**" feature in the AnythingLLM workspace, the entire application immediately crashed. The front-end displayed a JSON parsing error: `"unexpected token '<'"`, suggesting the API response was an HTML or malformed payload instead of valid JSON. This failure resulted in the application becoming completely inaccessible, showing a persistent _"503 Service Temporarily Offline nginx"_ error. We had to restart the entire application/deployment to restore functionality. **What did you expect to happen:** I expected the Markdown file to upload successfully, be processed as a source document, and be indexed for use in the AnythingLLM workspace without any interruption or crash to the core service. ### Are there known steps to reproduce? Yes, the bug is reliably reproducible with the following steps on a self-hosted instance: 1. Deployment Context: Use AnythingLLM v1.19 deployed in a self-hosted environment (our instance is running at ai.weown.app). 2. Navigate: Go to an existing or new Workspace. 3. Initiate Upload: Select the "Website Builder" upload method. 4. Select File: Upload a typical Markdown file (.md). 5. Observe Failure: The system immediately fails the upload process. The UI shows the "unexpected token '<'" JSON error, and the entire backend service enters the "503 Service Temporary Offline nginx" state, requiring a restart to recover.
yindo added the possible bug label 2026-02-22 18:32:06 -05:00
yindo closed this issue 2026-02-22 18:32:06 -05:00
Author
Owner

@mshahid538 commented on GitHub (Nov 27, 2025):

Adding the markdown file that I used to upload:

Brixies.co Layout.md

@mshahid538 commented on GitHub (Nov 27, 2025): Adding the markdown file that I used to upload: [Brixies.co Layout.md](https://github.com/user-attachments/files/23809326/Brixies.co.Layout.md)
Author
Owner

@timothycarambat commented on GitHub (Dec 2, 2025):

"Website Builder" feature in the AnythingLLM workspace,

We have no such feature. You seems to be using a fork of AnythingLLM or something else because we do not have any such functionality anywhere in our UI. Can you provide a screenshot and the docker tag specifically you are using?

v1.19

Check our Releases, we are on version 1.9.0 - so you are on the wrong repo

@timothycarambat commented on GitHub (Dec 2, 2025): > "Website Builder" feature in the AnythingLLM workspace, We have no such feature. You seems to be using a fork of AnythingLLM or something else because we do not have any such functionality anywhere in our UI. Can you provide a screenshot and the docker tag specifically you are using? > v1.19 Check our [Releases](https://github.com/Mintplex-Labs/anything-llm/releases), we are on version 1.9.0 - so you are on the wrong repo
Author
Owner

@mshahid538 commented on GitHub (Dec 3, 2025):

Yes, here are some screenshots for you to look over.

  1. Uploading a markdown file and getting JSON Parsing error.
Image Image
@mshahid538 commented on GitHub (Dec 3, 2025): Yes, here are some screenshots for you to look over. 1. Uploading a markdown file and getting JSON Parsing error. <img width="1895" height="1010" alt="Image" src="https://github.com/user-attachments/assets/d1af8621-c744-4cbf-b2fb-59933425b898" /> <img width="1891" height="1006" alt="Image" src="https://github.com/user-attachments/assets/862383a1-b00d-4534-8723-9a4b83db4c1a" />
Author
Owner

@mshahid538 commented on GitHub (Dec 4, 2025):

Hello Team,

Thank you for your immediate follow-up on this issue.

We sincerely apologize—there was a clerical error in our initial report regarding the version.

To be clear, our deployment is running the official, standard v1.9.0 release from the Docker image mintplexlabs/anythingllm:1.9.0, as documented in your official repository and our internal infrastructure standards.

We are still experiencing the instability with this confirmed version.

Issue Details Remain: The AnythingLLM Pod experiences an extreme memory utilization spike during the document embedding process (specifically for larger documents), leading directly to an OOMKilled event in our Kubernetes cluster and a brief 503 error before the Pod restarts.

We believe this is related to the resource requirements of the embedding model initialization.

Thank you for your continued support on this.

Image
@mshahid538 commented on GitHub (Dec 4, 2025): **Hello Team,** Thank you for your immediate follow-up on this issue. We sincerely apologize—there was a clerical error in our initial report regarding the version. To be clear, our deployment is running the official, standard v1.9.0 release from the Docker image mintplexlabs/anythingllm:1.9.0, as documented in your official repository and our internal infrastructure standards. We are still experiencing the instability with this confirmed version. Issue Details Remain: The AnythingLLM Pod experiences an extreme memory utilization spike during the document embedding process (specifically for larger documents), leading directly to an OOMKilled event in our Kubernetes cluster and a brief 503 error before the Pod restarts. We believe this is related to the resource requirements of the embedding model initialization. Thank you for your continued support on this. <img width="1420" height="510" alt="Image" src="https://github.com/user-attachments/assets/4ac1754a-2e94-4dc7-a09f-d5c2350dab1b" />
yindo changed title from [BUG]: Markdown File Upload Causes Critical Application Crash and 503 Service Error to [GH-ISSUE #4696] [BUG]: Markdown File Upload Causes Critical Application Crash and 503 Service Error 2026-06-05 14:49:35 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#2976