model-serving-minefield

Community registry of LLM serving-path traps that produce confidently wrong measurements: templates, tool parsers, reasoning fields, quant kernel paths, CUDA toolchains, KV allocation, eval harnesses, versioning. Symptom-first, with the check that catches each. Maintained by Blackwellboy on GitHub, where it has earned 64 stars from the community.

It's actively developed around benchmarking, chat-template, llama-cpp, and is a solid reference for anyone building with these tools.