fix(client): support protobuf 6 runtimes

.github/workflows/ci.yml

@@ -96,8 +96,16 @@ jobs:
       run: cargo test --doc
 
   python-test:
-    name: Python Tests
+    name: Python Tests (protobuf ${{ matrix.protobuf_major }})
     runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - protobuf_major: 5
+            protobuf_spec: "protobuf==5.27.2"
+          - protobuf_major: 6
+            protobuf_spec: "protobuf>=6.30.2,<7.0.0"
 
     steps:
     - name: Checkout repository
@@ -115,7 +123,7 @@ jobs:
       working-directory: modelexpress_client/python
       run: |
         uv venv .venv --python 3.12
-        uv pip install -e ".[dev]"
+        uv pip install -e ".[dev]" '${{ matrix.protobuf_spec }}'
 
     - name: Run tests
       working-directory: modelexpress_client/python

modelexpress_client/python/README.md

@@ -15,8 +15,11 @@ pip install modelexpress
 # Editable install from source
 pip install -e .
 
-# With dev dependencies (pytest, grpcio-tools)
+# With test dependencies
 pip install -e ".[dev]"
+
+# Additionally install the pinned protobuf code generator when changing p2p.proto
+pip install -e ".[codegen]"
 ```
 
 NIXL is expected to be supplied by the runtime environment (TRT-LLM,
@@ -27,6 +30,7 @@ or `pip install nixl-cu13` separately, matching your host CUDA toolkit.
 ### Requirements
 
 - Python >= 3.10
+- protobuf >= 5.27.2 and < 7
 - NVIDIA GPUs with RDMA/InfiniBand support
 - [NIXL](https://github.com/ai-dynamo/nixl) (NVIDIA Interconnect eXchange Library)
 - A running [ModelExpress server](https://github.com/ai-dynamo/modelexpress/tree/main/modelexpress_server) (Rust gRPC service backed by Redis)

modelexpress_client/python/pyproject.toml

@@ -28,7 +28,7 @@ dependencies = [
     "google-crc32c>=1.5.0",
     "huggingface_hub>=0.20.0",
     "numpy>=1.24.0",
-    "protobuf>=5.27.0,<6.0.0", # protobuf 5.x compatibility
+    "protobuf>=5.27.2,<7.0.0",
     "pydantic>=2.0.0",
     "torch>=2.6.0",
     "runai-model-streamer[s3,gcs,azure]; sys_platform == 'linux'",
@@ -37,13 +37,15 @@ dependencies = [
 
 [project.optional-dependencies]
 dev = [
-    # Use grpcio-tools<=1.66.2 for protobuf 5.x, >=1.67.0 for protobuf 6.x
-    "grpcio-tools>=1.60.0,<=1.66.2",
     "opentelemetry-api>=1.41.1",
     "opentelemetry-sdk>=1.41.1",
     "pytest>=7.0.0",
     "pytest-asyncio>=0.21.0",
 ]
+codegen = [
+    # Keep generated Python stubs compatible with both protobuf 5 and 6.
+    "grpcio-tools==1.66.2",
+]
 otel = [
     "opentelemetry-api>=1.20.0",
 ]

modelexpress_client/python/tests/test_types.py

@@ -19,8 +19,8 @@
 class TestProtobufCompatibility:
     """Guard against generated protobuf code drifting from the installed runtime."""
 
-    def test_p2p_pb2_gencode_matches_runtime_major_version(self):
-        """Regenerate p2p_pb2.py if this fails (see pyproject.toml [dev] deps)."""
+    def test_p2p_pb2_gencode_supports_runtime_major_version(self):
+        """Keep protobuf 5 gencode usable on supported 5.x and 6.x runtimes."""
         import modelexpress.p2p_pb2 as pb2
 
         with open(pb2.__file__) as f:
@@ -29,12 +29,9 @@ def test_p2p_pb2_gencode_matches_runtime_major_version(self):
         assert m, "Could not parse gencode version from p2p_pb2.py"
         gencode_major = int(m.group(1))
         runtime_major = int(_pb_version.split(".")[0])
-        assert gencode_major == runtime_major, (
-            f"p2p_pb2.py was generated with protobuf {gencode_major}.x "
-            f"but runtime is {runtime_major}.x - regenerate with: "
-            f"python -m grpc_tools.protoc -I../../modelexpress_common/proto "
-            f"--python_out=modelexpress --grpc_python_out=modelexpress "
-            f"../../modelexpress_common/proto/p2p.proto"
+        assert gencode_major == 5, "checked-in bindings must remain protobuf 5 gencode"
+        assert runtime_major in {5, 6}, (
+            f"unsupported protobuf runtime major: {runtime_major}"
         )
 
     def test_publish_request_is_readable_by_0_4_schema(self):