BUG: Response is aborted when reasoning budget is exceeded

[majick]

Describe the bug
If the inference back end hits the reasoning budget, the response generation is aborted.

To Reproduce
Steps to reproduce the behavior:

1. Configure the model with a small reasoning budget (such as --reasoning-budget 64)
2. Submit a prompt
3. Watch the reasoning occur
4. Response is aborted and the prompt returned to the input field.

Expected behavior
Response continues after reasoning if either a natural EOS or forced EOS occurs.

Screenshots
I mean... just look with your eyes.

Desktop (please complete the following information):

• OS: Darwin localhost 24.6.0 Darwin Kernel Version 24.6.0: Tue Apr 21 20:13:48 PDT 2026; root:xnu-11417.140.69.710.16~1/RELEASE_ARM64_T8112 arm64 arm Darwin
• Browser: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/26.5 Safari/605.1.15
• Version: Sidekick 1.0.0-rc.18 (38)

Additional context

• Network inference: llama-swap -> llama.cpp main -> Gemma-4-26B-A4B Q6_K (bf16 mmproj)
• llama-swap returns 200
• llama.cpp returns 200

[majick]

llama-server \
    -t 16 -C 0xF \
    -c 128000 -cmoe --reasoning-budget 64 \
    -b 2048 -ub 2048 \
    --cache-ram 4096  -ctk bf16 -ctv bf16 -kvu -np 3 \
    --no-mmap -fa on --host 0.0.0.0  --port ${PORT} \
    --chat-template-file /home/user/tempmodels/gemma-4-thinking.jinja \
    -mm /home/user/tempmodels/gemma-4-26B-A4B-it-mmproj.bf16.gguf \
    -m  /home/user/tempmodels/gemma-4-26B-A4B-it-heretic.Q6_K.gguf

Example.