DegenerateStreamGuard too strict for certain completions

[salixh5]

Prerequisites

• I searched open issues and discussions and did not find an existing report of this bug.
• This is not a security vulnerability. (Vulnerabilities go to GitHub Security Advisories — see SECURITY.md.)
• I am running the latest code from the dev branch (the default branch you get on clone, where fixes land first) and the bug still reproduces there. Please git pull the latest dev before filing.

Install Method

Docker (docker compose up)

Operating System

Linux

Steps to Reproduce

1. Have an advanced model (e.g. GPT 5.6 Sol) generate example questions for a test or something similar.
2. The output might have repetitions of TeX markup like \frac in it in rapid succession, thus triggering the DegenerateStreamGuard.
3. Output gets thrown away and error Stopped generation: gpt-5.6-sol started repeating tokens. Try a different model or lower temperature. is shown.

Expected Behaviour

I want to see the full reply. The guard condition might make sense for some local models, but not when used with SOTA models etc. -- would be nice to be able to disable it.

Actual Behaviour

Just an error, no way to see the full reply.

Logs / Screenshots

Model / Backend (if relevant)

OpenAI Subscription with GPT 5.6 Sol

Are you willing to submit a fix?

No — I am only filing the report

Additional Information

No response