move imports to method

Author: slekkala1

llama_stack/providers/inline/safety/code_scanner/code_scanner.py

@@ -6,9 +6,10 @@
 
 import logging
 import uuid
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
-from codeshield.cs import CodeShield, CodeShieldScanResult
+if TYPE_CHECKING:
+    from codeshield.cs import CodeShieldScanResult
 
 from llama_stack.apis.inference import Message
 from llama_stack.apis.safety import (
@@ -59,6 +60,8 @@ async def run_shield(
         if not shield:
             raise ValueError(f"Shield {shield_id} not found")
 
+        from codeshield.cs import CodeShield
+
         text = "\n".join([interleaved_content_as_str(m.content) for m in messages])
         log.info(f"Running CodeScannerShield on {text[50:]}")
         result = await CodeShield.scan_code(text)
@@ -72,7 +75,7 @@ async def run_shield(
             )
         return RunShieldResponse(violation=violation)
 
-    def get_moderation_object_results(self, scan_result: CodeShieldScanResult) -> ModerationObjectResults:
+    def get_moderation_object_results(self, scan_result: "CodeShieldScanResult") -> ModerationObjectResults:
         categories = {}
         category_scores = {}
         category_applied_input_types = {}
@@ -102,6 +105,8 @@ async def run_moderation(self, input: str | list[str], model: str) -> Moderation
         inputs = input if isinstance(input, list) else [input]
         results = []
 
+        from codeshield.cs import CodeShield
+
         for text_input in inputs:
             log.info(f"Running CodeScannerShield moderation on input: {text_input[:100]}...")
             scan_result = await CodeShield.scan_code(text_input)

llama_stack/providers/inline/safety/llama_guard/llama_guard.py

@@ -455,7 +455,7 @@ def create_moderation_object(self, model: str, unsafe_code: str | None = None) -
 
     def is_content_safe(self, response: str, unsafe_code: str | None = None) -> bool:
         """Check if content is safe based on response and unsafe code."""
-        if response.strip() == SAFE_RESPONSE:
+        if response.strip().lower().startswith(SAFE_RESPONSE):
             return True
 
         if unsafe_code:

pyproject.toml

@@ -25,8 +25,8 @@ classifiers = [
 ]
 dependencies = [
     "aiohttp",
-    "fastapi>=0.115.0,<1.0",                  # server
-    "fire",                                   # for MCP in LLS client
+    "fastapi>=0.115.0,<1.0", # server
+    "fire", # for MCP in LLS client
     "httpx",
     "huggingface-hub>=0.34.0,<1.0",
     "jinja2>=3.1.6",
@@ -44,12 +44,12 @@ dependencies = [
     "tiktoken",
     "pillow",
     "h11>=0.16.0",
-    "python-multipart>=0.0.20",               # For fastapi Form
-    "uvicorn>=0.34.0",                        # server
-    "opentelemetry-sdk>=1.30.0",              # server
+    "python-multipart>=0.0.20", # For fastapi Form
+    "uvicorn>=0.34.0", # server
+    "opentelemetry-sdk>=1.30.0", # server
     "opentelemetry-exporter-otlp-proto-http>=1.30.0", # server
-    "aiosqlite>=0.21.0",                      # server - for metadata store
-    "asyncpg",                                # for metadata store
+    "aiosqlite>=0.21.0", # server - for metadata store
+    "asyncpg", # for metadata store
 ]
 
 [project.optional-dependencies]
@@ -163,6 +163,7 @@ explicit = true
 [tool.uv.sources]
 torch = [{ index = "pytorch-cpu" }]
 torchvision = [{ index = "pytorch-cpu" }]
+llama-stack-client = { path = "../llama-stack-client-python" }
 
 [tool.ruff]
 line-length = 120