[Bugfix][Frontend] Cap fastapi < 0.137 to avoid prometheus-fastapi-instrumentator crash on serve startup

A fresh pip install pulls fastapi 0.137, whose new _IncludedRouter route type
(a BaseRoute with no .path) breaks prometheus-fastapi-instrumentator's route
iteration -> 'vllm serve' dies at startup with:
  AttributeError: '_IncludedRouter' object has no attribute 'path'

requirements/common.txt pinned fastapi[standard]>=0.115.0 with no upper bound.
Cap to <0.137; lift once the instrumentator handles _IncludedRouter
(https://github.com/trallnag/prometheus-fastapi-instrumentator/issues/370).

Signed-off-by: avesed <trevor.lijunlin@gmail.com>

docker/Dockerfile.cpu

@@ -168,6 +168,12 @@ RUN --mount=type=cache,target=/root/.cache/uv \
 ######################### TRITON-CPU BUILD IMAGE #########################
 FROM base AS vllm-triton-cpu-build
 
+# Support for cross-compilation with x86 ISA including AVX2 and AVX512: docker build --build-arg VLLM_CPU_X86="true" ...
+# Re-declared here because this stage is `FROM base` (not `vllm-build`), so it
+# does not inherit the ARG/ENV defined there. Without it, the guard below would
+# see an empty value and build triton-cpu on non-x86 targets (e.g. arm64).
+ARG VLLM_CPU_X86=0
+
 WORKDIR /vllm-workspace
 
 RUN mkdir dist
@@ -269,6 +275,11 @@ ENV HF_HUB_DOWNLOAD_TIMEOUT 60
 ######################### RELEASE IMAGE #########################
 FROM base AS vllm-openai
 
+# Re-declared here because this stage is `FROM base` (not `vllm-build`), so the
+# RUN below that gates the triton-cpu wheel install on $VLLM_CPU_X86 would
+# otherwise see an empty value and try to install it on non-x86 targets.
+ARG VLLM_CPU_X86=0
+
 WORKDIR /vllm-workspace
 
 RUN --mount=type=cache,target=/root/.cache/uv \

requirements/common.txt

@@ -11,7 +11,7 @@ transformers >= 4.56.0, != 5.0.*, != 5.1.*, != 5.2.*, != 5.3.*, != 5.4.*, != 5.5
 tokenizers >= 0.21.1  # Required for fast incremental detokenization.
 safetensors >= 0.6.2  # MXFP4/MXFP6 dtype support (F8_E8M0, F4) added in 0.6.0: https://github.com/huggingface/safetensors/pull/611
 protobuf >= 5.29.6, !=6.30.*, !=6.31.*, !=6.32.*, !=6.33.0.*, !=6.33.1.*, !=6.33.2.*, !=6.33.3.*, !=6.33.4.* # Required by LlamaTokenizer, gRPC. CVE-2026-0994
-fastapi[standard] >= 0.115.0 # Required by FastAPI's form models in the OpenAI API server's audio transcriptions endpoint.
+fastapi[standard] >= 0.115.0, < 0.137 # Required by FastAPI's form models in the OpenAI API server's audio transcriptions endpoint; <0.137 due to prometheus-fastapi-instrumentator#370.
 aiohttp >= 3.13.3
 openai >= 2.0.0  # For Responses API with reasoning content
 pydantic >= 2.12.0