bug: Core dump in parsing bad PDF?

[cab938]

Version: 0.7.4

Describe the Bug

I attach multiple PDFs to a conversation and submit them with a text prompt using Qwen3-next. The Jan UI closes, and the error on the CLI is:

[2025-12-05][20:08:18][tauri_plugin_rag::commands][INFO] Parsing document: MY_FILENAME_AND_PATH.pdf (type: pdf)

thread 'tokio-runtime-worker' (3375451) panicked at /home/runner/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/pdf-extract-0.7.12/src/lib.rs:203:30:
X1
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Aborted (core dumped)

Steps to Reproduce

1. As above, sadly I can't share the PDF file as it is sensitive

Screenshots / Logs

Operating System

• MacOS
• Windows
• Linux

[cab938]

Also, it doesn't kill the llamacpp backend, so this defacto is a memory leak since I must go kill that process manually to restart.

[Minh141120]

Thanks for flagging this, @cab938, and for sharing the detailed logs. This is super helpful. We'll investigate the PDF parsing crash and the lingering backend process, and we'll keep you posted as soon as we have updates or a fix.

[github-roushan]

Can i take this @Minh141120. Thank you

[Minh141120]

Absolutely @github-roushan. Feel free to take this on. Even if an issue shows an assignee, contributions from anyone in the community are always welcome.

We'll also be investigating the issue on our side and will keep everyone updated in this thread.

[qnixsynapse]

pdf-extract crate is crashing probably failing to read pdf which in turn crashes entire application, so cleanups are not being performed, probably due to missing exception handling in implementation.