ci(test): OPS-9049 replace flake8 with ruff and upgrade the ruff hook to 0.17.3

.ai/python-guidelines.md

@@ -356,7 +356,7 @@ ruff format <touched_paths>
 ruff check --fix <touched_paths>
 ```
 
-Or use pre-commit (runs isort, black, flake8, ruff, and more):
+Or use pre-commit (runs isort, black, ruff, and more):
 
 ```bash
 pre-commit run --files <touched_files>
@@ -369,7 +369,6 @@ The repo's `.pre-commit-config.yaml` runs these Python hooks:
 
 - **isort** -- import sorting (`profile = "black"`, configured in `pyproject.toml`)
 - **black** -- code formatting
-- **flake8** -- style checks (`max-line-length=88`)
 - **ruff** -- fast linting with auto-fix
 - **codespell** -- spelling checks
 - **trailing-whitespace**, **end-of-file-fixer**, **check-yaml**, **check-json**, **check-toml**

.pre-commit-config.yaml

@@ -26,12 +26,6 @@ repos:
   hooks:
   - id: black
     types_or: [python, cython]
-- repo: https://github.com/PyCQA/flake8
-  rev: 7.3.0  # 5.0.4 crashes on Python 3.12+ (ast.Str removed)
-  hooks:
-  - id: flake8
-    args: [--max-line-length=88, --select=C,E,F,W,B,B950, --extend-ignore = E203,E501]
-    types_or: [python, cython]
 - repo: https://github.com/pre-commit/mirrors-clang-format
   rev: v16.0.5
   hooks:
@@ -71,9 +65,9 @@ repos:
 
 # Fast linting
 - repo: https://github.com/astral-sh/ruff-pre-commit
-  rev: v0.5.2
+  rev: v0.17.3
   hooks:
-  - id: ruff
+  - id: ruff-check
     args: [--fix, --exit-non-zero-on-fix]
 
 # NOTE: pyright may be able to find other classes of errors not covered above,

docs/fern/scripts/tests/test_gen_python_api.py

@@ -20,7 +20,7 @@
 ``conftest.py`` next to this file owns the ``sys.path`` shim that makes the
 top-level imports of ``api_discovery`` / ``api_rendering`` /
 ``gen_python_api`` resolve, so this module keeps every import at the top
-and stays clean under the repo's flake8 config.
+and stays clean under the repo's ruff config.
 
 Invocation from an isolated environment (bypasses the repo's Python
 resolution, which is unrelated to this generator)::

lib/bindings/kvbm/python/kvbm/__init__.py

@@ -1,7 +1,7 @@
 # SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-# flake8: noqa
+# ruff: noqa
 import logging
 
 logger = logging.getLogger(__name__)

lib/bindings/python/codegen/src/gen_python_prometheus_names.rs

@@ -75,7 +75,7 @@ const MAX_LINE_LEN: usize = 88;
 
 /// Emit one `NAME = "value"` assignment, wrapping in parentheses exactly the way black
 /// does when the single-line form would exceed `MAX_LINE_LEN`. Comments are left alone —
-/// black does not reflow them, and flake8 has `E501` in `extend-ignore`.
+/// black does not reflow them, and the ruff hook does not select `E501`.
 fn push_assignment(body_pad: &str, name: &str, value: &str, lines: &mut Vec<String>) {
     let single = format!("{}{} = \"{}\"", body_pad, name, value);
     if single.chars().count() <= MAX_LINE_LEN {

lib/bindings/python/src/dynamo/llm/__init__.py

@@ -1,14 +1,11 @@
 # SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-# flake8: noqa
-
-import logging
 from collections.abc import AsyncIterator
 from typing import Any, Protocol
 
 from dynamo._core import AisPerfConfig as AisPerfConfig
-from dynamo._core import EngineType
+from dynamo._core import EngineType as EngineType
 from dynamo._core import EntrypointArgs as EntrypointArgs
 from dynamo._core import FpmDirectPublisher as FpmDirectPublisher
 from dynamo._core import FpmEventRelay as FpmEventRelay
@@ -48,9 +45,9 @@
 from dynamo._core import compute_block_hash_for_seq as compute_block_hash_for_seq
 from dynamo._core import fetch_model as fetch_model
 from dynamo._core import lora_name_to_id as lora_name_to_id
-from dynamo._core import make_engine
+from dynamo._core import make_engine as make_engine
 from dynamo._core import register_model as register_model
-from dynamo._core import run_input
+from dynamo._core import run_input as run_input
 from dynamo._core import run_kv_indexer as run_kv_indexer
 from dynamo._core import run_select_service as run_select_service
 from dynamo._core import run_slot_tracker as run_slot_tracker
@@ -65,7 +62,7 @@
 from ._unary import LLMUnaryClient as LLMUnaryClient
 from ._unary import UnaryChatModel as UnaryChatModel
 from ._unary import with_engine_data as with_engine_data
-from .exceptions import HttpError
+from .exceptions import HttpError as HttpError
 from .exceptions import RouterQueueLimitExceeded as RouterQueueLimitExceeded
 from .exceptions import WorkerOverloaded as WorkerOverloaded
 

lib/bindings/python/src/dynamo/llm/exceptions.py

@@ -1,8 +1,6 @@
 # SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-# flake8: noqa
-
 import logging
 
 from dynamo._core import Cancelled as Cancelled

pyproject.toml

@@ -422,6 +422,13 @@ markers = [
 line-length = 88
 indent-width = 4
 
+[tool.ruff.lint]
+# E4, E7, E9, and F are the rules that ruff 0.5.2 enabled by default. Newer ruff enables many more.
+# W605 flags invalid escape sequences, and PLW0602 flags a `global` name that is never assigned.
+# TODO: enable flake8-bugbear (B) after its findings on main are fixed. Ruff has no B950;
+# E501 (line too long) is the closest ruff rule.
+select = ["E4", "E7", "E9", "F", "W605", "PLW0602"]
+
 [tool.ruff.lint.extend-per-file-ignores]
 "icp/tests/**/test_*.py" = ["F811", "F401"]
 "*_inc.py" = ["F821"]