fix pii errors

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sentinelguard"
-version = "0.0.3"
+version = "0.0.4"
 description = "A comprehensive, production-ready LLM security and guardrails framework"
 readme = "README.md"
 license = {text = "MIT"}
@@ -36,7 +36,7 @@ dependencies = [
     "presidio-analyzer>=2.2.0",
     "presidio-anonymizer>=2.2.0",
     "spacy>=3.6.0",
-    "en-core-web-lg>=3.6.0",
+    "en-core-web-lg>=3.7.1",
     # Model-based detection (HuggingFace)
     "transformers>=4.30.0",
     "torch>=2.0.0",

sentinelguard/__init__.py

@@ -36,7 +36,7 @@
     guard = SentinelGuard(config=config)
 """
 
-__version__ = "0.0.3"
+__version__ = "0.0.4"
 __author__ = "SentinelGuard Contributors"
 
 from sentinelguard.core.guard import SentinelGuard

sentinelguard/scanners/output/bias.py

@@ -16,8 +16,6 @@
 import re
 from typing import Any, ClassVar, Dict, List, Optional, Tuple
 
-from transformers import pipeline
-
 from sentinelguard.core.scanner import OutputScanner, RiskLevel, ScanResult, register_scanner
 
 logger = logging.getLogger(__name__)
@@ -112,8 +110,9 @@ def __init__(
     def _load_model(self) -> None:
         if self._model is None:
             try:
+                from transformers import pipeline as hf_pipeline
                 logger.info("Loading bias detection model: %s", _BIAS_MODEL_ID)
-                self._model = pipeline("text-classification", model=_BIAS_MODEL_ID)
+                self._model = hf_pipeline("text-classification", model=_BIAS_MODEL_ID)
             except Exception as exc:
                 logger.warning("Failed to load bias model, falling back to regex only: %s", exc)
                 self._model = False  # sentinel: tried and failed

sentinelguard/scanners/prompt/jailbreak.py

@@ -19,8 +19,6 @@
 import re
 from typing import Any, ClassVar, Dict, List, Optional, Tuple
 
-from transformers import pipeline
-
 from sentinelguard.core.scanner import PromptScanner, RiskLevel, ScanResult, register_scanner
 
 logger = logging.getLogger(__name__)
@@ -179,8 +177,9 @@ def __init__(
     def _load_model(self) -> None:
         if self._model is None:
             try:
+                from transformers import pipeline as hf_pipeline
                 logger.info("Loading jailbreak detection model: %s", _JAILBREAK_MODEL_ID)
-                self._model = pipeline("text-classification", model=_JAILBREAK_MODEL_ID)
+                self._model = hf_pipeline("text-classification", model=_JAILBREAK_MODEL_ID)
             except Exception as exc:
                 logger.warning("Failed to load jailbreak model, falling back to patterns only: %s", exc)
                 self._model = False

sentinelguard/scanners/prompt/prompt_injection.py

@@ -12,8 +12,6 @@
 import re
 from typing import Any, ClassVar, List, Optional
 
-from transformers import pipeline
-
 from sentinelguard.core.scanner import PromptScanner, RiskLevel, ScanResult, register_scanner
 
 logger = logging.getLogger(__name__)
@@ -99,8 +97,9 @@ def __init__(
     def _load_model(self) -> None:
         if self._model is None:
             try:
+                from transformers import pipeline as hf_pipeline
                 logger.info("Loading prompt injection model: %s", _INJECTION_MODEL_ID)
-                self._model = pipeline("text-classification", model=_INJECTION_MODEL_ID)
+                self._model = hf_pipeline("text-classification", model=_INJECTION_MODEL_ID)
             except Exception as exc:
                 logger.warning("Failed to load injection model, falling back to patterns+heuristics: %s", exc)
                 self._model = False

sentinelguard/scanners/prompt/toxicity.py

@@ -11,8 +11,6 @@
 import re
 from typing import Any, ClassVar, Dict, List, Optional
 
-from transformers import pipeline
-
 from sentinelguard.core.scanner import PromptScanner, RiskLevel, ScanResult, register_scanner
 
 logger = logging.getLogger(__name__)
@@ -86,8 +84,9 @@ def __init__(
     def _load_model(self) -> None:
         if self._model is None:
             try:
+                from transformers import pipeline as hf_pipeline
                 logger.info("Loading toxicity model: %s", _TOXICITY_MODEL_ID)
-                self._model = pipeline("text-classification", model=_TOXICITY_MODEL_ID, top_k=None)
+                self._model = hf_pipeline("text-classification", model=_TOXICITY_MODEL_ID, top_k=None)
             except Exception as exc:
                 logger.warning("Failed to load toxicity model, falling back to patterns: %s", exc)
                 self._model = False