llm: add FastAPI shim, gateway LLM endpoints, tests, and docs

models/qwen3/README.md

@@ -0,0 +1,9 @@
+Place the Qwen3 GGUF model file for the local llm profile in this directory.
+
+Expected default filename:
+
+- `Qwen3-1.7B-Instruct-Q4_K_M.gguf`
+
+You can use a different filename, but then set `MODEL_PATH` in `.env` to match the mounted path inside the container.
+
+The model is intentionally not auto-downloaded at startup. Operators should provision it explicitly so container startup is predictable.