This is one of these games that would be super awesome... if they'd just work. No matter what I do, at some point in every run the chat just stops responding. The graphics and the music continue to loop but Babs just falls silent. And no, it's not my PC, I got 32 GB RAM and a GeForce RTX 5080.
If the dev manages to make it stable, I'll come back to it and happily pay (love the quiz), but in the current state: hard pass.
Viewing post in The Spanking Therapist - Governess Barbara comments
It works. This is the game that has sold the most, not only here but with patrons and subscribers, and if it was really unstable, I would have had tons complaints. However, making it run can be picky as it's two separate programs communicating with each other through a HTTP address, and this communication is fragile. If it doesn't work with your PC it's probably because of a third party program/configuration that may be disrupting communication between ollama and unreal. Dropbox, some messy antivirus, an update running in the background. The only way to figure it out is by running a PC monitor that show the exact error. And next release, Barbara therapist, will be fully integrated with unreal, to avoid this kind of issues and can run in more PCs.
Clean boot tested with third-party services/startups disabled. The issue still occurs.
Ollama process is running and port 11434 is reachable.
The game process remains responsive and keeps established TCP connections to 127.0.0.1:11434.
ollama ps shows the model loaded on 100% GPU.
GPU load is about 92%, VRAM usage about 8.7/16.3 GB, temperature about 61°C.
While the in-game chatbot is stuck, a separate ollama run request to the same model answers immediately.
So Ollama and the model are still responsive; the failure seems to be specific to the game request/response handling rather than system load or third-party software.
I reproduced the issue in a Windows Clean Boot.
The game remains responsive and keeps established TCP connections to Ollama on 127.0.0.1:11434.
Ollama and the model remain fully responsive.
I enabled OLLAMA_DEBUG_LOG_REQUESTS=1 and captured the exact failing /api/chat request.
The request JSON is valid. Replaying the exact same request manually with curl returns HTTP 200 and a complete response in about 7.8 seconds with done:true and done_reason:"stop".
So the request itself is valid and Ollama can answer it. The issue appears to be in the game's handling of the Ollama response or the request lifecycle, not in system load, Ollama availability, or the model.
If you want, I can e-mail you a detailed bug report.