safety

api_update_plan.md

@@ -219,9 +219,9 @@ Before finalizing documentation, verify:
 - Minimal code changes - primarily docstring additions only
 
 ### File Processing Order (Complete Absolute Paths)
-1. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/agents/agents.py` - Core agent system (start here, most complete)
+[x] 1. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/agents/agents.py` - Core agent system (start here, most complete)
-2. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/inference/inference.py` - Core LLM functionality 
+[x] 2. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/inference/inference.py` - Core LLM functionality 
-3. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/safety/safety.py` - Safety and moderation
+[x] 3. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/safety/safety.py` - Safety and moderation
 4. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/models/models.py` - Model metadata and management
 5. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/tools/tools.py` - Tool system APIs
 6. `/Users/saip/Documents/GitHub/llama-stack/llama_stack/apis/tools/rag_tool.py` - RAG tool runtime

docs/_static/llama-stack-spec.html

@@ -6801,10 +6801,12 @@
                 "type": "object",
                 "properties": {
                     "violation_level": {
-                        "$ref": "#/components/schemas/ViolationLevel"
+                        "$ref": "#/components/schemas/ViolationLevel",
+                        "description": "Severity level of the violation"
                     },
                     "user_message": {
-                        "type": "string"
+                        "type": "string",
+                        "description": "(Optional) Message to convey to the user about the violation"
                     },
                     "metadata": {
                         "type": "object",
@@ -6829,7 +6831,8 @@
                                     "type": "object"
                                 }
                             ]
-                        }
+                        },
+                        "description": "Additional metadata including specific violation codes for debugging and telemetry"
                     }
                 },
                 "additionalProperties": false,
@@ -6837,7 +6840,8 @@
                     "violation_level",
                     "metadata"
                 ],
-                "title": "SafetyViolation"
+                "title": "SafetyViolation",
+                "description": "Details of a safety violation detected by content moderation."
             },
             "ShieldCallStep": {
                 "type": "object",
@@ -7140,7 +7144,8 @@
                     "warn",
                     "error"
                 ],
-                "title": "ViolationLevel"
+                "title": "ViolationLevel",
+                "description": "Severity level of a safety violation."
             },
             "AgentTurnResponseEvent": {
                 "type": "object",
@@ -15758,11 +15763,13 @@
                 "type": "object",
                 "properties": {
                     "violation": {
-                        "$ref": "#/components/schemas/SafetyViolation"
+                        "$ref": "#/components/schemas/SafetyViolation",
+                        "description": "(Optional) Safety violation detected by the shield, if any"
                     }
                 },
                 "additionalProperties": false,
-                "title": "RunShieldResponse"
+                "title": "RunShieldResponse",
+                "description": "Response from running a safety shield."
             },
             "SaveSpansToDatasetRequest": {
                 "type": "object",

docs/_static/llama-stack-spec.yaml

@@ -4888,8 +4888,11 @@ components:
       properties:
         violation_level:
           $ref: '#/components/schemas/ViolationLevel'
+          description: Severity level of the violation
         user_message:
           type: string
+          description: >-
+            (Optional) Message to convey to the user about the violation
         metadata:
           type: object
           additionalProperties:
@@ -4900,11 +4903,16 @@ components:
               - type: string
               - type: array
               - type: object
+          description: >-
+            Additional metadata including specific violation codes for debugging and
+            telemetry
       additionalProperties: false
       required:
         - violation_level
         - metadata
       title: SafetyViolation
+      description: >-
+        Details of a safety violation detected by content moderation.
     ShieldCallStep:
       type: object
       properties:
@@ -5122,6 +5130,7 @@ components:
         - warn
         - error
       title: ViolationLevel
+      description: Severity level of a safety violation.
     AgentTurnResponseEvent:
       type: object
       properties:
@@ -11128,8 +11137,11 @@ components:
       properties:
         violation:
           $ref: '#/components/schemas/SafetyViolation'
+          description: >-
+            (Optional) Safety violation detected by the shield, if any
       additionalProperties: false
       title: RunShieldResponse
+      description: Response from running a safety shield.
     SaveSpansToDatasetRequest:
       type: object
       properties:

llama_stack/apis/safety/safety.py

@@ -17,6 +17,12 @@ from llama_stack.schema_utils import json_schema_type, webmethod
 
 @json_schema_type
 class ViolationLevel(Enum):
+    """Severity level of a safety violation.
+
+    :cvar INFO: Informational level violation that does not require action
+    :cvar WARN: Warning level violation that suggests caution but allows continuation
+    :cvar ERROR: Error level violation that requires blocking or intervention
+    """
     INFO = "info"
     WARN = "warn"
     ERROR = "error"
@@ -24,6 +30,12 @@ class ViolationLevel(Enum):
 
 @json_schema_type
 class SafetyViolation(BaseModel):
+    """Details of a safety violation detected by content moderation.
+
+    :param violation_level: Severity level of the violation
+    :param user_message: (Optional) Message to convey to the user about the violation
+    :param metadata: Additional metadata including specific violation codes for debugging and telemetry
+    """
     violation_level: ViolationLevel
 
     # what message should you convey to the user
@@ -36,6 +48,10 @@ class SafetyViolation(BaseModel):
 
 @json_schema_type
 class RunShieldResponse(BaseModel):
+    """Response from running a safety shield.
+
+    :param violation: (Optional) Safety violation detected by the shield, if any
+    """
     violation: SafetyViolation | None = None