[Bug] sd-server HTTP error 500 with sdcpp API endpoints

[afischer-sh]

Git commit

When running sd-server, the sdcpp API does not work and always gives HTTP/500 errors.
Example: GET /sdcpp/v1/capabilities -> 500 Internal Server Error

OpenAI API works (e.g. GET /v1/models)
Stable Diffusion WebUI API works (e.g. GET /sdapi/v1/sd-models)

No errors in sd-server output (running with -v).

Operating System & Version

Ubuntu 24.04

GGML backends

CUDA

Command-line arguments used

./sd-server --listen-ip 0.0.0.0 --diffusion-model ... --vae ... --llm ... --verbose

Steps to reproduce

Open :1234/sdcpp/v1/capabilities

What you expected to happen

HTTP/200

What actually happened

HTTP/500

Logs / error messages / stack trace

No response

Additional context / environment details

No response

[TechnoJo4]

Got the same issue, deepseekv4 traced it back to

stable-diffusion.cpp/examples/common/common.h

Line 103 in 3d6064b

std::string lora_model_dir = ".";

stable-diffusion.cpp/examples/server/runtime.cpp

Line 257 in 3d6064b

for (auto& entry : fs::recursive_directory_iterator(lora_dir)) {

Setting --lora-model-dir, or in my case, setting working_dir in my docker compose, both fixed the issue.

[afischer-sh]

Got the same issue, deepseekv4 traced it back to

stable-diffusion.cpp/examples/common/common.h

Line 103 in 3d6064b

std::string lora_model_dir = ".";
stable-diffusion.cpp/examples/server/runtime.cpp

Line 257 in 3d6064b

for (auto& entry : fs::recursive_directory_iterator(lora_dir)) {
Setting --lora-model-dir, or in my case, setting working_dir in my docker compose, both fixed the issue.

Thank you! That workaround did the job.