Merge branch 'main' into e2e-llmd

Author: rootfs

config/testing/config.e2e.yaml

@@ -69,11 +69,11 @@ classifier:
     use_cpu: true
     category_mapping_path: "models/lora_intent_classifier_bert-base-uncased_model/category_mapping.json"
   pii_model:
-    model_id: "models/pii_classifier_modernbert-base_presidio_token_model"  # TODO: Use local model for now before the code can download the entire model from huggingface
-    use_modernbert: true
+    model_id: "models/lora_pii_detector_bert-base-uncased_model"
+    use_modernbert: false  # BERT-based LoRA model (this field is ignored - always auto-detects)
     threshold: 0.7
     use_cpu: true
-    pii_mapping_path: "models/pii_classifier_modernbert-base_presidio_token_model/pii_type_mapping.json"
+    pii_mapping_path: "models/lora_pii_detector_bert-base-uncased_model/pii_type_mapping.json"
 categories:
   - name: business
     description: "Business and management related queries"
@@ -359,6 +359,24 @@ decisions:
           enabled: true
           pii_types_allowed: ["EMAIL_ADDRESS", "PERSON", "GPE", "PHONE_NUMBER", "US_SSN", "CREDIT_CARD"]
 
+  # Default catch-all decision for unmatched requests (E2E PII test fix)
+  # This ensures PII detection is always enabled, even when no specific decision matches
+  - name: "default_decision"
+    description: "Default catch-all decision - blocks all PII for safety"
+    priority: 1  # Lowest priority - only matches if nothing else does
+    rules:
+      operator: "OR"
+      conditions:
+        - type: "always"  # Always matches as fallback
+    modelRefs:
+      - model: "Model-B"
+        use_reasoning: false
+    plugins:
+      - type: "pii"
+        configuration:
+          enabled: true
+          pii_types_allowed: []  # Block ALL PII - empty list means nothing allowed
+
 default_model: "Model-A"
 
 # API Configuration

deploy/helm/semantic-router/values.yaml

@@ -167,6 +167,9 @@ initContainer:
       repo: LLM-Semantic-Router/jailbreak_classifier_modernbert-base_model
     - name: pii_classifier_modernbert-base_presidio_token_model
       repo: LLM-Semantic-Router/pii_classifier_modernbert-base_presidio_token_model
+    # LoRA PII detector (for auto-detection feature)
+    - name: lora_pii_detector_bert-base-uncased_model
+      repo: LLM-Semantic-Router/lora_pii_detector_bert-base-uncased_model
 
 
 # Autoscaling configuration

deploy/kubernetes/aibrix/semantic-router-values/values.yaml

@@ -437,8 +437,10 @@ config:
       use_cpu: true
       category_mapping_path: "models/category_classifier_modernbert-base_model/category_mapping.json"
     pii_model:
-      model_id: "models/pii_classifier_modernbert-base_presidio_token_model"
-      use_modernbert: true
+      # Support both traditional (modernbert) and LoRA-based PII detection
+      # When model_type is "auto", the system will auto-detect LoRA configuration
+      model_id: "models/lora_pii_detector_bert-base-uncased_model"
+      use_modernbert: false  # Use LoRA PII model with auto-detection
       threshold: 0.7
       use_cpu: true
       pii_mapping_path: "models/pii_classifier_modernbert-base_presidio_token_model/pii_type_mapping.json"