Fix CI across probe-agent, training-monitor, and inference-monitor

Probe Agent:
- Create libnvidia-ml.so.1 symlink to CUDA stubs in build and
  unit-test CI jobs so test binaries can load without a real driver
- Add LD_LIBRARY_PATH for stubs dir in build step

Training Monitor (58 ruff errors):
- Fix UP035/UP037: modernize type annotations (collections.abc imports,
  remove string quotes from annotations)
- Fix F841: remove unused variables
- Fix F401: remove unused imports
- Fix E402: add noqa for post-importorskip imports in tests
- Fix E501: break long lines
- Fix SIM102/SIM105/SIM108: simplify conditionals and exception handling
- Fix B007/B905: prefix unused loop vars, add strict=False to zip
- Fix N803/N806: rename uppercase variables

Inference Monitor (38 ruff errors):
- Fix F401: remove unused imports
- Fix UP041: replace asyncio.TimeoutError with TimeoutError
- Fix SIM105: use contextlib.suppress instead of try/except/pass
- Fix E501: break long lines
- Fix F841: prefix unused variables with underscore
- Fix N803: rename uppercase parameter
- Fix SIM108: use ternary expressions

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: DannyWall92

.github/workflows/ci-probe-agent.yml

@@ -40,6 +40,7 @@ jobs:
             libgrpc++-dev libgtest-dev libgmock-dev pkg-config ca-certificates \
             libspdlog-dev nlohmann-json3-dev libssl-dev
           rm -rf /var/lib/apt/lists/*
+          ln -s /usr/local/cuda/lib64/stubs/libnvidia-ml.so /usr/local/cuda/lib64/stubs/libnvidia-ml.so.1
 
       - name: Configure CMake
         working-directory: probe-agent
@@ -51,6 +52,8 @@ jobs:
 
       - name: Build
         working-directory: probe-agent
+        env:
+          LD_LIBRARY_PATH: /usr/local/cuda/lib64/stubs
         run: cmake --build build --parallel $(nproc)
 
       - name: Upload build artifacts
@@ -75,6 +78,7 @@ jobs:
             cmake ninja-build g++ libgtest-dev libgmock-dev libprotobuf-dev \
             libgrpc++-dev pkg-config libspdlog-dev nlohmann-json3-dev libssl-dev
           rm -rf /var/lib/apt/lists/*
+          ln -s /usr/local/cuda/lib64/stubs/libnvidia-ml.so /usr/local/cuda/lib64/stubs/libnvidia-ml.so.1
 
       - name: Build and test
         working-directory: probe-agent

inference-monitor/src/sentinel_inference/analyzers/entropy_analyzer.py

@@ -12,9 +12,6 @@
 
 from __future__ import annotations
 
-import dataclasses
-from typing import Any
-
 import numpy as np
 import structlog
 

inference-monitor/src/sentinel_inference/analyzers/kl_divergence.py

@@ -11,7 +11,6 @@
 import dataclasses
 import time
 from collections import OrderedDict
-from typing import Any
 
 import numpy as np
 import structlog

inference-monitor/src/sentinel_inference/analyzers/logit_analyzer.py

@@ -66,9 +66,11 @@ class EWMATracker:
         "_sigma",
     )
 
-    def __init__(self, lambda_: float = 0.1, L: float = 3.5, burn_in: int = 1000) -> None:
+    def __init__(
+        self, lambda_: float = 0.1, control_limit: float = 3.5, burn_in: int = 1000,
+    ) -> None:
         self._lambda = lambda_
-        self._L = L
+        self._L = control_limit
         self._burn_in = burn_in
         self._count: int = 0
         self._ewma: float = 0.0

inference-monitor/src/sentinel_inference/analyzers/spectral_analyzer.py

@@ -18,9 +18,7 @@
 
 from __future__ import annotations
 
-import dataclasses
 from collections import deque
-from typing import Any
 
 import numpy as np
 import structlog

inference-monitor/src/sentinel_inference/analyzers/statistical_tests.py

@@ -7,9 +7,7 @@
 
 from __future__ import annotations
 
-import dataclasses
 from collections import deque
-from typing import Any
 
 import numpy as np
 import structlog
@@ -110,7 +108,9 @@ def analyze(self, logits: np.ndarray) -> list[AnomalyEvent]:
                     ucl=self._config.ks_p_value_threshold,
                     lcl=0.0,
                     sample_count=len(self._baseline),
-                    severity="critical" if ks_p < self._config.ks_p_value_threshold / 10 else "warning",
+                    severity=(
+                        "critical" if ks_p < self._config.ks_p_value_threshold / 10 else "warning"
+                    ),
                     details={"ks_statistic": ks_stat, "p_value": ks_p},
                 )
             )

inference-monitor/src/sentinel_inference/config.py

@@ -120,7 +120,8 @@ class StatisticalTestsConfig(BaseModel):
         default=1,
         ge=0,
         le=4,
-        description="Index into Anderson-Darling significance levels (0=15%, 1=10%, 2=5%, 3=2.5%, 4=1%).",
+        description="Index into Anderson-Darling significance levels "
+        "(0=15%, 1=10%, 2=5%, 3=2.5%, 4=1%).",
     )
     baseline_size: int = Field(
         default=5000,
@@ -252,15 +253,15 @@ class MonitorConfig(BaseModel):
     metrics: MetricsConfig = Field(default_factory=MetricsConfig)
 
     @model_validator(mode="after")
-    def _set_defaults(self) -> "MonitorConfig":
+    def _set_defaults(self) -> MonitorConfig:
         if not self.node_id:
             self.node_id = os.environ.get("HOSTNAME", "unknown")
         if not self.replica_id:
             self.replica_id = os.environ.get("POD_NAME", self.node_id)
         return self
 
     @classmethod
-    def from_yaml(cls, path: str | Path) -> "MonitorConfig":
+    def from_yaml(cls, path: str | Path) -> MonitorConfig:
         """Load configuration from a YAML file, with env-var overrides."""
         import yaml  # type: ignore[import-untyped]
 
@@ -273,7 +274,7 @@ def from_yaml(cls, path: str | Path) -> "MonitorConfig":
         return cls.model_validate(raw)
 
     @classmethod
-    def from_env(cls) -> "MonitorConfig":
+    def from_env(cls) -> MonitorConfig:
         """Build config from environment variables prefixed with SENTINEL_."""
         overrides: dict[str, str] = {}
         for key, val in os.environ.items():

inference-monitor/src/sentinel_inference/grpc_client.py

@@ -9,6 +9,7 @@
 from __future__ import annotations
 
 import asyncio
+import contextlib
 import hashlib
 import time
 import uuid
@@ -201,24 +202,18 @@ async def stop(self) -> None:
         self._running = False
         if self._flush_task is not None:
             self._flush_task.cancel()
-            try:
+            with contextlib.suppress(asyncio.CancelledError):
                 await self._flush_task
-            except asyncio.CancelledError:
-                pass
         if self._ack_task is not None:
             self._ack_task.cancel()
-            try:
+            with contextlib.suppress(asyncio.CancelledError):
                 await self._ack_task
-            except asyncio.CancelledError:
-                pass
         # Final flush.
         await self._flush()
         # Close the bidi stream if open.
         if self._stream is not None:
-            try:
+            with contextlib.suppress(Exception):
                 await self._stream.done_writing()
-            except Exception:
-                pass
             self._stream = None
         if self._channel is not None:
             await self._channel.close()
@@ -386,8 +381,10 @@ async def _do_send(self, report: AnomalyReport) -> None:
     async def _do_send_proto(self, report: AnomalyReport) -> None:
         """Send using generated protobuf stubs."""
         from google.protobuf.timestamp_pb2 import Timestamp  # type: ignore[import-untyped]
-        from sentinel.v1 import anomaly_pb2  # type: ignore[import-untyped]
-        from sentinel.v1 import common_pb2  # type: ignore[import-untyped]
+        from sentinel.v1 import (
+            anomaly_pb2,  # type: ignore[import-untyped]
+            common_pb2,  # type: ignore[import-untyped]
+        )
 
         # Build the AnomalyBatch protobuf message.
         batch = anomaly_pb2.AnomalyBatch()
@@ -473,7 +470,6 @@ async def _do_send_manual(self, report: AnomalyReport) -> None:
         """
         import json
 
-        import grpc  # type: ignore[import-untyped]
 
         payload = json.dumps({
             "source_hostname": report.source_hostname,

inference-monitor/src/sentinel_inference/interceptors/triton_interceptor.py

@@ -11,7 +11,6 @@
 import struct
 import uuid
 from multiprocessing import shared_memory
-from typing import Any
 
 import numpy as np
 import structlog

inference-monitor/src/sentinel_inference/interceptors/trtllm_interceptor.py

@@ -108,5 +108,5 @@ async def capture_output(self) -> TensorCapture | None:
         """Retrieve the next captured tensor from the internal queue."""
         try:
             return await asyncio.wait_for(self._queue.get(), timeout=0.1)
-        except asyncio.TimeoutError:
+        except TimeoutError:
             return None