fix: disable strict X-Request-ID UUID validation in starlette_context to prevent HF proxy 400 rejection

Dockerfile

@@ -9,6 +9,11 @@ RUN pip install "uvicorn<0.30.0"
 # Install llama-cpp-python with a server (Using pre-compiled CPU wheels to skip the long C++ build!)
 RUN pip install "llama-cpp-python[server]" huggingface-hub --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
 
+# CRITICAL FIX: Hugging Face proxy injects a non-UUID 'X-Request-ID' header (e.g. 'qKsNoh').
+# starlette_context's RequestIdPlugin strictly validates this as UUIDv4 by default and throws a 400 Bad Request with an empty body.
+# We must patch llama_cpp to disable this validation.
+RUN python3 -c "import llama_cpp.server.app as app; path = app.__file__; open(path, 'w').write(open(path).read().replace('RequestIdPlugin()', 'RequestIdPlugin(validate=False)'))"
+
 # Download a fast, compressed model (Qwen 1.5B is perfect for free CPUs)
 RUN python3 -c "from huggingface_hub import hf_hub_download; hf_hub_download(repo_id='Qwen/Qwen2.5-1.5B-Instruct-GGUF', filename='qwen2.5-1.5b-instruct-q4_k_m.gguf', local_dir='/model')"