Perplexica disconnects from LM Studio before LLM can finish processing

[kylearoberts]

Describe the bug
On long LLM processing jobs Perplexica disconnects from LM Studio before the processing is complete.

Running LM Studio with a 7900 XTX using qwen3.5-9b

To Reproduce
Steps to reproduce the behavior:

1. Create a prompt that has a large context to create a long prompt processing time (~5 minutes).
2. LM Studio reports: Client disconnected. Stopping generation...

Expected behavior
Perplexica should not disconnect from LM Studio before the processing is complete.

[Malis128]

It seems to be specifically the Qwen3.5 family of models causing this. Have this issue with Qwen3.5 9b, 27b, and 35b a3, but no others

[kylearoberts]

I got this to happen with Nemotron 3 Nano doing deep research and balanced research as well.