Merge branch 'main' into session-manager

llama_stack/cli/stack/run.py

@@ -15,7 +15,7 @@ from llama_stack.log import get_logger
 
 REPO_ROOT = Path(__file__).parent.parent.parent.parent
 
-logger = get_logger(name=__name__, category="server")
+logger = get_logger(name=__name__, category="cli")
 
 
 class StackRun(Subcommand):

llama_stack/core/datatypes.py

@@ -318,6 +318,41 @@ class QuotaConfig(BaseModel):
     period: QuotaPeriod = Field(default=QuotaPeriod.DAY, description="Quota period to set")
 
 
+class CORSConfig(BaseModel):
+    allow_origins: list[str] = Field(default_factory=list)
+    allow_origin_regex: str | None = Field(default=None)
+    allow_methods: list[str] = Field(default=["OPTIONS"])
+    allow_headers: list[str] = Field(default_factory=list)
+    allow_credentials: bool = Field(default=False)
+    expose_headers: list[str] = Field(default_factory=list)
+    max_age: int = Field(default=600, ge=0)
+
+    @model_validator(mode="after")
+    def validate_credentials_config(self) -> Self:
+        if self.allow_credentials and (self.allow_origins == ["*"] or "*" in self.allow_origins):
+            raise ValueError("Cannot use wildcard origins with credentials enabled")
+        return self
+
+
+def process_cors_config(cors_config: bool | CORSConfig | None) -> CORSConfig | None:
+    if cors_config is False or cors_config is None:
+        return None
+
+    if cors_config is True:
+        # dev mode: allow localhost on any port
+        return CORSConfig(
+            allow_origins=[],
+            allow_origin_regex=r"https?://localhost:\d+",
+            allow_methods=["GET", "POST", "PUT", "DELETE", "OPTIONS"],
+            allow_headers=["Content-Type", "Authorization", "X-Requested-With"],
+        )
+
+    if isinstance(cors_config, CORSConfig):
+        return cors_config
+
+    raise ValueError(f"Expected bool or CORSConfig, got {type(cors_config).__name__}")
+
+
 class ServerConfig(BaseModel):
     port: int = Field(
         default=8321,
@@ -349,6 +384,12 @@ class ServerConfig(BaseModel):
         default=None,
         description="Per client quota request configuration",
     )
+    cors: bool | CORSConfig | None = Field(
+        default=None,
+        description="CORS configuration for cross-origin requests. Can be:\n"
+        "- true: Enable localhost CORS for development\n"
+        "- {allow_origins: [...], allow_methods: [...], ...}: Full configuration",
+    )
 
 
 class StackRunConfig(BaseModel):

llama_stack/core/library_client.py

@@ -146,39 +146,26 @@ class LlamaStackAsLibraryClient(LlamaStackClient):
     ):
         super().__init__()
         self.async_client = AsyncLlamaStackAsLibraryClient(
-            config_path_or_distro_name, custom_provider_registry, provider_data
+            config_path_or_distro_name, custom_provider_registry, provider_data, skip_logger_removal
         )
         self.pool_executor = ThreadPoolExecutor(max_workers=4)
-        self.skip_logger_removal = skip_logger_removal
         self.provider_data = provider_data
 
         self.loop = asyncio.new_event_loop()
 
-    def initialize(self):
-        if in_notebook():
-            import nest_asyncio
-
-            nest_asyncio.apply()
-            if not self.skip_logger_removal:
-                self._remove_root_logger_handlers()
-
         # use a new event loop to avoid interfering with the main event loop
         loop = asyncio.new_event_loop()
         asyncio.set_event_loop(loop)
         try:
-            return loop.run_until_complete(self.async_client.initialize())
+            loop.run_until_complete(self.async_client.initialize())
         finally:
             asyncio.set_event_loop(None)
 
-    def _remove_root_logger_handlers(self):
+    def initialize(self):
         """
-        Remove all handlers from the root logger. Needed to avoid polluting the console with logs.
+        Deprecated method for backward compatibility.
         """
-        root_logger = logging.getLogger()
-
-        for handler in root_logger.handlers[:]:
-            root_logger.removeHandler(handler)
-            logger.info(f"Removed handler {handler.__class__.__name__} from root logger")
+        pass
 
     def request(self, *args, **kwargs):
         loop = self.loop
@@ -216,6 +203,7 @@ class AsyncLlamaStackAsLibraryClient(AsyncLlamaStackClient):
         config_path_or_distro_name: str,
         custom_provider_registry: ProviderRegistry | None = None,
         provider_data: dict[str, Any] | None = None,
+        skip_logger_removal: bool = False,
     ):
         super().__init__()
         # when using the library client, we should not log to console since many
@@ -223,6 +211,13 @@ class AsyncLlamaStackAsLibraryClient(AsyncLlamaStackClient):
         current_sinks = os.environ.get("TELEMETRY_SINKS", "sqlite").split(",")
         os.environ["TELEMETRY_SINKS"] = ",".join(sink for sink in current_sinks if sink != "console")
 
+        if in_notebook():
+            import nest_asyncio
+
+            nest_asyncio.apply()
+            if not skip_logger_removal:
+                self._remove_root_logger_handlers()
+
         if config_path_or_distro_name.endswith(".yaml"):
             config_path = Path(config_path_or_distro_name)
             if not config_path.exists():
@@ -239,7 +234,24 @@ class AsyncLlamaStackAsLibraryClient(AsyncLlamaStackClient):
         self.provider_data = provider_data
         self.route_impls: RouteImpls | None = None  # Initialize to None to prevent AttributeError
 
+    def _remove_root_logger_handlers(self):
+        """
+        Remove all handlers from the root logger. Needed to avoid polluting the console with logs.
+        """
+        root_logger = logging.getLogger()
+
+        for handler in root_logger.handlers[:]:
+            root_logger.removeHandler(handler)
+            logger.info(f"Removed handler {handler.__class__.__name__} from root logger")
+
     async def initialize(self) -> bool:
+        """
+        Initialize the async client.
+
+        Returns:
+            bool: True if initialization was successful
+        """
+
         try:
             self.route_impls = None
             self.impls = await construct_stack(self.config, self.custom_provider_registry)

llama_stack/core/routers/datasets.py

@@ -12,7 +12,7 @@ from llama_stack.apis.datasets import DatasetPurpose, DataSource
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import RoutingTable
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routers")
 
 
 class DatasetIORouter(DatasetIO):

llama_stack/core/routers/eval_scoring.py

@@ -16,7 +16,7 @@ from llama_stack.apis.scoring import (
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import RoutingTable
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routers")
 
 
 class ScoringRouter(Scoring):

llama_stack/core/routers/inference.py

@@ -65,7 +65,7 @@ from llama_stack.providers.datatypes import HealthResponse, HealthStatus, Routin
 from llama_stack.providers.utils.inference.inference_store import InferenceStore
 from llama_stack.providers.utils.telemetry.tracing import get_current_span
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="core::routers")
 
 
 class InferenceRouter(Inference):

llama_stack/core/routers/safety.py

@@ -13,7 +13,7 @@ from llama_stack.apis.shields import Shield
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import RoutingTable
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routers")
 
 
 class SafetyRouter(Safety):

llama_stack/core/routers/tool_runtime.py

@@ -22,7 +22,7 @@ from llama_stack.log import get_logger
 
 from ..routing_tables.toolgroups import ToolGroupsRoutingTable
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routers")
 
 
 class ToolRuntimeRouter(ToolRuntime):

llama_stack/core/routers/vector_io.py

@@ -30,7 +30,7 @@ from llama_stack.apis.vector_io import (
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import HealthResponse, HealthStatus, RoutingTable
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routers")
 
 
 class VectorIORouter(VectorIO):

llama_stack/core/routing_tables/benchmarks.py

@@ -14,7 +14,7 @@ from llama_stack.log import get_logger
 
 from .common import CommonRoutingTableImpl
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 class BenchmarksRoutingTable(CommonRoutingTableImpl, Benchmarks):

llama_stack/core/routing_tables/common.py

@@ -23,7 +23,7 @@ from llama_stack.core.store import DistributionRegistry
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import Api, RoutingTable
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 def get_impl_api(p: Any) -> Api:

llama_stack/core/routing_tables/datasets.py

@@ -26,7 +26,7 @@ from llama_stack.log import get_logger
 
 from .common import CommonRoutingTableImpl
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 class DatasetsRoutingTable(CommonRoutingTableImpl, Datasets):

llama_stack/core/routing_tables/models.py

@@ -17,7 +17,7 @@ from llama_stack.log import get_logger
 
 from .common import CommonRoutingTableImpl, lookup_model
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 class ModelsRoutingTable(CommonRoutingTableImpl, Models):

llama_stack/core/routing_tables/scoring_functions.py

@@ -19,7 +19,7 @@ from llama_stack.log import get_logger
 
 from .common import CommonRoutingTableImpl
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 class ScoringFunctionsRoutingTable(CommonRoutingTableImpl, ScoringFunctions):

llama_stack/core/routing_tables/shields.py

@@ -15,7 +15,7 @@ from llama_stack.log import get_logger
 
 from .common import CommonRoutingTableImpl
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 class ShieldsRoutingTable(CommonRoutingTableImpl, Shields):

llama_stack/core/routing_tables/toolgroups.py

@@ -14,7 +14,7 @@ from llama_stack.log import get_logger
 
 from .common import CommonRoutingTableImpl
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 def parse_toolgroup_from_toolgroup_name_pair(toolgroup_name_with_maybe_tool_name: str) -> str | None:

llama_stack/core/routing_tables/vector_dbs.py

@@ -30,7 +30,7 @@ from llama_stack.log import get_logger
 
 from .common import CommonRoutingTableImpl, lookup_model
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="core::routing_tables")
 
 
 class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):

llama_stack/core/server/auth.py

@@ -15,7 +15,7 @@ from llama_stack.core.server.auth_providers import create_auth_provider
 from llama_stack.core.server.routes import find_matching_route, initialize_route_impls
 from llama_stack.log import get_logger
 
-logger = get_logger(name=__name__, category="auth")
+logger = get_logger(name=__name__, category="core::auth")
 
 
 class AuthenticationMiddleware:

llama_stack/core/server/auth_providers.py

@@ -23,7 +23,7 @@ from llama_stack.core.datatypes import (
 )
 from llama_stack.log import get_logger
 
-logger = get_logger(name=__name__, category="auth")
+logger = get_logger(name=__name__, category="core::auth")
 
 
 class AuthResponse(BaseModel):

llama_stack/core/server/quota.py

@@ -15,7 +15,7 @@ from llama_stack.providers.utils.kvstore.api import KVStore
 from llama_stack.providers.utils.kvstore.config import KVStoreConfig, SqliteKVStoreConfig
 from llama_stack.providers.utils.kvstore.kvstore import kvstore_impl
 
-logger = get_logger(name=__name__, category="quota")
+logger = get_logger(name=__name__, category="core::server")
 
 
 class QuotaMiddleware:

llama_stack/core/server/server.py

@@ -28,6 +28,7 @@ from aiohttp import hdrs
 from fastapi import Body, FastAPI, HTTPException, Request, Response
 from fastapi import Path as FastapiPath
 from fastapi.exceptions import RequestValidationError
+from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse, StreamingResponse
 from openai import BadRequestError
 from pydantic import BaseModel, ValidationError
@@ -40,6 +41,7 @@ from llama_stack.core.datatypes import (
     AuthenticationRequiredError,
     LoggingConfig,
     StackRunConfig,
+    process_cors_config,
 )
 from llama_stack.core.distribution import builtin_automatically_routed_apis
 from llama_stack.core.external import ExternalApiSpec, load_external_apis
@@ -82,7 +84,7 @@ from .quota import QuotaMiddleware
 
 REPO_ROOT = Path(__file__).parent.parent.parent.parent
 
-logger = get_logger(name=__name__, category="server")
+logger = get_logger(name=__name__, category="core::server")
 
 
 def warn_with_traceback(message, category, filename, lineno, file=None, line=None):
@@ -413,7 +415,7 @@ def main(args: argparse.Namespace | None = None):
         config_contents = yaml.safe_load(fp)
         if isinstance(config_contents, dict) and (cfg := config_contents.get("logging_config")):
             logger_config = LoggingConfig(**cfg)
-        logger = get_logger(name=__name__, category="server", config=logger_config)
+        logger = get_logger(name=__name__, category="core::server", config=logger_config)
         if args.env:
             for env_pair in args.env:
                 try:
@@ -483,6 +485,12 @@ def main(args: argparse.Namespace | None = None):
             window_seconds=window_seconds,
         )
 
+    if config.server.cors:
+        logger.info("Enabling CORS")
+        cors_config = process_cors_config(config.server.cors)
+        if cors_config:
+            app.add_middleware(CORSMiddleware, **cors_config.model_dump())
+
     if Api.telemetry in impls:
         setup_logger(impls[Api.telemetry])
     else:

llama_stack/core/store/registry.py

@@ -16,7 +16,7 @@ from llama_stack.log import get_logger
 from llama_stack.providers.utils.kvstore import KVStore, kvstore_impl
 from llama_stack.providers.utils.kvstore.config import KVStoreConfig, SqliteKVStoreConfig
 
-logger = get_logger(__name__, category="core")
+logger = get_logger(__name__, category="core::registry")
 
 
 class DistributionRegistry(Protocol):

llama_stack/core/utils/config_resolution.py

@@ -10,7 +10,7 @@ from pathlib import Path
 from llama_stack.core.utils.config_dirs import DISTRIBS_BASE_DIR
 from llama_stack.log import get_logger
 
-logger = get_logger(name=__name__, category="config_resolution")
+logger = get_logger(name=__name__, category="core")
 
 
 DISTRO_DIR = Path(__file__).parent.parent.parent.parent / "llama_stack" / "distributions"

llama_stack/models/llama/llama3/multimodal/model.py

@@ -36,7 +36,7 @@ from .utils import get_negative_inf_value, to_2tuple
 
 MP_SCALE = 8
 
-logger = get_logger(name=__name__, category="models")
+logger = get_logger(name=__name__, category="models::llama")
 
 
 def reduce_from_tensor_model_parallel_region(input_):

llama_stack/models/llama/llama3/tool_utils.py

@@ -11,7 +11,7 @@ from llama_stack.log import get_logger
 
 from ..datatypes import BuiltinTool, RecursiveType, ToolCall, ToolPromptFormat
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="models::llama")
 
 BUILTIN_TOOL_PATTERN = r'\b(?P<tool_name>\w+)\.call\(query="(?P<query>[^"]*)"\)'
 CUSTOM_TOOL_CALL_PATTERN = re.compile(r"<function=(?P<function_name>[^}]+)>(?P<args>{.*?})")

llama_stack/models/llama/llama4/quantization/loader.py

@@ -18,7 +18,7 @@ from ...datatypes import QuantizationMode
 from ..model import Transformer, TransformerBlock
 from ..moe import MoE
 
-log = get_logger(name=__name__, category="models")
+log = get_logger(name=__name__, category="models::llama")
 
 
 def swiglu_wrapper_no_reduce(

llama_stack/models/llama/quantize_impls.py

@@ -9,7 +9,7 @@ import collections
 
 from llama_stack.log import get_logger
 
-log = get_logger(name=__name__, category="llama")
+log = get_logger(name=__name__, category="models::llama")
 
 try:
     import fbgemm_gpu.experimental.gen_ai  # noqa: F401

llama_stack/providers/inline/agents/meta_reference/agent_instance.py

@@ -84,7 +84,7 @@ MEMORY_QUERY_TOOL = "knowledge_search"
 WEB_SEARCH_TOOL = "web_search"
 RAG_TOOL_GROUP = "builtin::rag"
 
-logger = get_logger(name=__name__, category="agents")
+logger = get_logger(name=__name__, category="agents::meta_reference")
 
 
 class ChatAgent(ShieldRunnerMixin):

llama_stack/providers/inline/agents/meta_reference/agents.py

@@ -51,7 +51,7 @@ from .config import MetaReferenceAgentsImplConfig
 from .persistence import AgentInfo
 from .responses.openai_responses import OpenAIResponsesImpl
 
-logger = get_logger(name=__name__, category="agents")
+logger = get_logger(name=__name__, category="agents::meta_reference")
 
 
 class MetaReferenceAgentsImpl(Agents):

llama_stack/providers/inline/agents/meta_reference/persistence.py

@@ -17,7 +17,7 @@ from llama_stack.core.request_headers import get_authenticated_user
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.kvstore import KVStore
 
-log = get_logger(name=__name__, category="agents")
+log = get_logger(name=__name__, category="agents::meta_reference")
 
 
 class AgentSessionInfo(Session):

llama_stack/providers/inline/agents/meta_reference/responses/openai_responses.py

@@ -41,7 +41,7 @@ from .utils import (
     convert_response_text_to_chat_response_format,
 )
 
-logger = get_logger(name=__name__, category="responses")
+logger = get_logger(name=__name__, category="openai::responses")
 
 
 class OpenAIResponsePreviousResponseWithInputItems(BaseModel):

llama_stack/providers/inline/agents/meta_reference/responses/streaming.py

@@ -47,7 +47,7 @@ from llama_stack.log import get_logger
 from .types import ChatCompletionContext, ChatCompletionResult
 from .utils import convert_chat_choice_to_response_message, is_function_tool_call
 
-logger = get_logger(name=__name__, category="responses")
+logger = get_logger(name=__name__, category="agents::meta_reference")
 
 
 class StreamingResponseOrchestrator:

llama_stack/providers/inline/agents/meta_reference/responses/tool_executor.py

@@ -38,7 +38,7 @@ from llama_stack.log import get_logger
 
 from .types import ChatCompletionContext, ToolExecutionResult
 
-logger = get_logger(name=__name__, category="responses")
+logger = get_logger(name=__name__, category="agents::meta_reference")
 
 
 class ToolExecutor:

llama_stack/providers/inline/agents/meta_reference/safety.py

@@ -11,7 +11,7 @@ from llama_stack.apis.safety import Safety, SafetyViolation, ViolationLevel
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.telemetry import tracing
 
-log = get_logger(name=__name__, category="agents")
+log = get_logger(name=__name__, category="agents::meta_reference")
 
 
 class SafetyException(Exception):  # noqa: N818

llama_stack/providers/remote/inference/fireworks/fireworks.py

@@ -65,7 +65,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 from .config import FireworksImplConfig
 from .models import MODEL_ENTRIES
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="inference::fireworks")
 
 
 class FireworksInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProviderData):

llama_stack/providers/remote/inference/llama_openai_compat/llama.py

@@ -10,7 +10,7 @@ from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 
 from .models import MODEL_ENTRIES
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="inference::llama_openai_compat")
 
 
 class LlamaCompatInferenceAdapter(OpenAIMixin, LiteLLMOpenAIMixin):

llama_stack/providers/remote/inference/nvidia/NVIDIA.md

@@ -41,6 +41,11 @@ client.initialize()
 
 ### Create Completion
 
+> Note on Completion API
+>
+> The hosted NVIDIA Llama NIMs (e.g., `meta-llama/Llama-3.1-8B-Instruct`) with ```NVIDIA_BASE_URL="https://integrate.api.nvidia.com"``` does not support the ```completion``` method, while the locally deployed NIM does.
+
+
 ```python
 response = client.inference.completion(
     model_id="meta-llama/Llama-3.1-8B-Instruct",
@@ -76,6 +81,73 @@ response = client.inference.chat_completion(
 print(f"Response: {response.completion_message.content}")
 ```
 
+### Tool Calling Example ###
+```python
+from llama_stack.models.llama.datatypes import ToolDefinition, ToolParamDefinition
+
+tool_definition = ToolDefinition(
+    tool_name="get_weather",
+    description="Get current weather information for a location",
+    parameters={
+        "location": ToolParamDefinition(
+            param_type="string",
+            description="The city and state, e.g. San Francisco, CA",
+            required=True,
+        ),
+        "unit": ToolParamDefinition(
+            param_type="string",
+            description="Temperature unit (celsius or fahrenheit)",
+            required=False,
+            default="celsius",
+        ),
+    },
+)
+
+tool_response = client.inference.chat_completion(
+    model_id="meta-llama/Llama-3.1-8B-Instruct",
+    messages=[{"role": "user", "content": "What's the weather like in San Francisco?"}],
+    tools=[tool_definition],
+)
+
+print(f"Tool Response: {tool_response.completion_message.content}")
+if tool_response.completion_message.tool_calls:
+    for tool_call in tool_response.completion_message.tool_calls:
+        print(f"Tool Called: {tool_call.tool_name}")
+        print(f"Arguments: {tool_call.arguments}")
+```
+
+### Structured Output Example
+```python
+from llama_stack.apis.inference import JsonSchemaResponseFormat, ResponseFormatType
+
+person_schema = {
+    "type": "object",
+    "properties": {
+        "name": {"type": "string"},
+        "age": {"type": "integer"},
+        "occupation": {"type": "string"},
+    },
+    "required": ["name", "age", "occupation"],
+}
+
+response_format = JsonSchemaResponseFormat(
+    type=ResponseFormatType.json_schema, json_schema=person_schema
+)
+
+structured_response = client.inference.chat_completion(
+    model_id="meta-llama/Llama-3.1-8B-Instruct",
+    messages=[
+        {
+            "role": "user",
+            "content": "Create a profile for a fictional person named Alice who is 30 years old and is a software engineer. ",
+        }
+    ],
+    response_format=response_format,
+)
+
+print(f"Structured Response: {structured_response.completion_message.content}")
+```
+
 ### Create Embeddings
 > Note on OpenAI embeddings compatibility
 >

llama_stack/providers/remote/inference/nvidia/nvidia.py

@@ -7,7 +7,7 @@
 import warnings
 from collections.abc import AsyncIterator
 
-from openai import NOT_GIVEN, APIConnectionError, BadRequestError
+from openai import NOT_GIVEN, APIConnectionError
 
 from llama_stack.apis.common.content_types import (
     InterleavedContent,
@@ -57,7 +57,7 @@ from .openai_utils import (
 )
 from .utils import _is_nvidia_hosted
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="inference::nvidia")
 
 
 class NVIDIAInferenceAdapter(OpenAIMixin, Inference, ModelRegistryHelper):
@@ -197,15 +197,11 @@ class NVIDIAInferenceAdapter(OpenAIMixin, Inference, ModelRegistryHelper):
             }
             extra_body["input_type"] = task_type_options[task_type]
 
-        try:
-            response = await self.client.embeddings.create(
-                model=provider_model_id,
-                input=input,
-                extra_body=extra_body,
-            )
-        except BadRequestError as e:
-            raise ValueError(f"Failed to get embeddings: {e}") from e
-
+        response = await self.client.embeddings.create(
+            model=provider_model_id,
+            input=input,
+            extra_body=extra_body,
+        )
         #
         # OpenAI: CreateEmbeddingResponse(data=[Embedding(embedding=list[float], ...)], ...)
         #  ->

llama_stack/providers/remote/inference/nvidia/utils.py

@@ -10,7 +10,7 @@ from llama_stack.log import get_logger
 
 from . import NVIDIAConfig
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="inference::nvidia")
 
 
 def _is_nvidia_hosted(config: NVIDIAConfig) -> bool:

llama_stack/providers/remote/inference/ollama/ollama.py

@@ -85,7 +85,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 
 from .models import MODEL_ENTRIES
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="inference::ollama")
 
 
 class OllamaInferenceAdapter(

llama_stack/providers/remote/inference/openai/openai.py

@@ -11,7 +11,7 @@ from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 from .config import OpenAIConfig
 from .models import MODEL_ENTRIES
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="inference::openai")
 
 
 #

llama_stack/providers/remote/inference/tgi/tgi.py

@@ -58,7 +58,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 
 from .config import InferenceAPIImplConfig, InferenceEndpointImplConfig, TGIImplConfig
 
-log = get_logger(name=__name__, category="inference")
+log = get_logger(name=__name__, category="inference::tgi")
 
 
 def build_hf_repo_model_entries():

llama_stack/providers/remote/inference/together/together.py

@@ -61,7 +61,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 from .config import TogetherImplConfig
 from .models import MODEL_ENTRIES
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="inference::together")
 
 
 class TogetherInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProviderData):

llama_stack/providers/remote/inference/vllm/vllm.py

@@ -85,7 +85,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 
 from .config import VLLMInferenceAdapterConfig
 
-log = get_logger(name=__name__, category="inference")
+log = get_logger(name=__name__, category="inference::vllm")
 
 
 def build_hf_repo_model_entries():

llama_stack/providers/remote/post_training/nvidia/utils.py

@@ -15,7 +15,7 @@ from llama_stack.providers.remote.post_training.nvidia.config import SFTLoRADefa
 
 from .config import NvidiaPostTrainingConfig
 
-logger = get_logger(name=__name__, category="integration")
+logger = get_logger(name=__name__, category="post_training::nvidia")
 
 
 def warn_unsupported_params(config_dict: Any, supported_keys: set[str], config_name: str) -> None:

llama_stack/providers/remote/safety/bedrock/bedrock.py

@@ -21,7 +21,7 @@ from llama_stack.providers.utils.bedrock.client import create_bedrock_client
 
 from .config import BedrockSafetyConfig
 
-logger = get_logger(name=__name__, category="safety")
+logger = get_logger(name=__name__, category="safety::bedrock")
 
 
 class BedrockSafetyAdapter(Safety, ShieldsProtocolPrivate):

llama_stack/providers/remote/safety/nvidia/nvidia.py

@@ -9,7 +9,7 @@ from typing import Any
 import requests
 
 from llama_stack.apis.inference import Message
-from llama_stack.apis.safety import RunShieldResponse, Safety, SafetyViolation, ViolationLevel
+from llama_stack.apis.safety import ModerationObject, RunShieldResponse, Safety, SafetyViolation, ViolationLevel
 from llama_stack.apis.shields import Shield
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import ShieldsProtocolPrivate
@@ -17,7 +17,7 @@ from llama_stack.providers.utils.inference.openai_compat import convert_message_
 
 from .config import NVIDIASafetyConfig
 
-logger = get_logger(name=__name__, category="safety")
+logger = get_logger(name=__name__, category="safety::nvidia")
 
 
 class NVIDIASafetyAdapter(Safety, ShieldsProtocolPrivate):
@@ -67,6 +67,9 @@ class NVIDIASafetyAdapter(Safety, ShieldsProtocolPrivate):
         self.shield = NeMoGuardrails(self.config, shield.shield_id)
         return await self.shield.run(messages)
 
+    async def run_moderation(self, input: str | list[str], model: str) -> ModerationObject:
+        raise NotImplementedError("NVIDIA safety provider currently does not implement run_moderation")
+
 
 class NeMoGuardrails:
     """

llama_stack/providers/remote/safety/sambanova/sambanova.py

@@ -25,7 +25,7 @@ from llama_stack.providers.utils.inference.openai_compat import convert_message_
 
 from .config import SambaNovaSafetyConfig
 
-logger = get_logger(name=__name__, category="safety")
+logger = get_logger(name=__name__, category="safety::sambanova")
 
 CANNED_RESPONSE_TEXT = "I can't answer that. Can I help with something else?"
 

llama_stack/providers/remote/vector_io/chroma/chroma.py

@@ -33,7 +33,7 @@ from llama_stack.providers.utils.memory.vector_store import (
 
 from .config import ChromaVectorIOConfig as RemoteChromaVectorIOConfig
 
-log = get_logger(name=__name__, category="vector_io")
+log = get_logger(name=__name__, category="vector_io::chroma")
 
 ChromaClientType = chromadb.api.AsyncClientAPI | chromadb.api.ClientAPI
 

llama_stack/providers/remote/vector_io/milvus/milvus.py

@@ -36,7 +36,7 @@ from llama_stack.providers.utils.vector_io.vector_utils import sanitize_collecti
 
 from .config import MilvusVectorIOConfig as RemoteMilvusVectorIOConfig
 
-logger = get_logger(name=__name__, category="vector_io")
+logger = get_logger(name=__name__, category="vector_io::milvus")
 
 VERSION = "v3"
 VECTOR_DBS_PREFIX = f"vector_dbs:milvus:{VERSION}::"

llama_stack/providers/remote/vector_io/pgvector/pgvector.py

@@ -34,7 +34,7 @@ from llama_stack.providers.utils.memory.vector_store import (
 
 from .config import PGVectorVectorIOConfig
 
-log = get_logger(name=__name__, category="vector_io")
+log = get_logger(name=__name__, category="vector_io::pgvector")
 
 VERSION = "v3"
 VECTOR_DBS_PREFIX = f"vector_dbs:pgvector:{VERSION}::"

llama_stack/providers/remote/vector_io/qdrant/qdrant.py

@@ -36,7 +36,7 @@ from llama_stack.providers.utils.memory.vector_store import (
 
 from .config import QdrantVectorIOConfig as RemoteQdrantVectorIOConfig
 
-log = get_logger(name=__name__, category="vector_io")
+log = get_logger(name=__name__, category="vector_io::qdrant")
 CHUNK_ID_KEY = "_chunk_id"
 
 # KV store prefixes for vector databases

llama_stack/providers/remote/vector_io/weaviate/weaviate.py

@@ -34,7 +34,7 @@ from llama_stack.providers.utils.vector_io.vector_utils import sanitize_collecti
 
 from .config import WeaviateVectorIOConfig
 
-log = get_logger(name=__name__, category="vector_io")
+log = get_logger(name=__name__, category="vector_io::weaviate")
 
 VERSION = "v3"
 VECTOR_DBS_PREFIX = f"vector_dbs:weaviate:{VERSION}::"

llama_stack/providers/utils/inference/embedding_mixin.py

@@ -28,7 +28,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import interleaved_con
 EMBEDDING_MODELS = {}
 
 
-log = get_logger(name=__name__, category="inference")
+log = get_logger(name=__name__, category="providers::utils")
 
 
 class SentenceTransformerEmbeddingMixin:

llama_stack/providers/utils/inference/litellm_openai_mixin.py

@@ -54,7 +54,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
     interleaved_content_as_str,
 )
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="providers::utils")
 
 
 class LiteLLMOpenAIMixin(

llama_stack/providers/utils/inference/model_registry.py

@@ -17,7 +17,7 @@ from llama_stack.providers.utils.inference import (
     ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR,
 )
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="providers::utils")
 
 
 class RemoteInferenceProviderConfig(BaseModel):

llama_stack/providers/utils/inference/openai_compat.py

@@ -134,7 +134,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
     decode_assistant_message,
 )
 
-logger = get_logger(name=__name__, category="inference")
+logger = get_logger(name=__name__, category="providers::utils")
 
 
 class OpenAICompatCompletionChoiceDelta(BaseModel):

llama_stack/providers/utils/inference/openai_mixin.py

@@ -25,7 +25,7 @@ from llama_stack.apis.inference import (
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.inference.openai_compat import prepare_openai_completion_params
 
-logger = get_logger(name=__name__, category="core")
+logger = get_logger(name=__name__, category="providers::utils")
 
 
 class OpenAIMixin(ABC):

llama_stack/providers/utils/inference/prompt_adapter.py

@@ -58,7 +58,7 @@ from llama_stack.models.llama.sku_list import resolve_model
 from llama_stack.models.llama.sku_types import ModelFamily, is_multimodal
 from llama_stack.providers.utils.inference import supported_inference_models
 
-log = get_logger(name=__name__, category="inference")
+log = get_logger(name=__name__, category="providers::utils")
 
 
 class ChatCompletionRequestWithRawContent(ChatCompletionRequest):

llama_stack/providers/utils/kvstore/mongodb/mongodb.py

@@ -13,7 +13,7 @@ from llama_stack.providers.utils.kvstore import KVStore
 
 from ..config import MongoDBKVStoreConfig
 
-log = get_logger(name=__name__, category="kvstore")
+log = get_logger(name=__name__, category="providers::utils")
 
 
 class MongoDBKVStoreImpl(KVStore):

llama_stack/providers/utils/kvstore/postgres/postgres.py

@@ -14,7 +14,7 @@ from llama_stack.log import get_logger
 from ..api import KVStore
 from ..config import PostgresKVStoreConfig
 
-log = get_logger(name=__name__, category="kvstore")
+log = get_logger(name=__name__, category="providers::utils")
 
 
 class PostgresKVStoreImpl(KVStore):

llama_stack/providers/utils/memory/openai_vector_store_mixin.py

@@ -44,7 +44,7 @@ from llama_stack.providers.utils.memory.vector_store import (
     make_overlapped_chunks,
 )
 
-logger = get_logger(name=__name__, category="memory")
+logger = get_logger(name=__name__, category="providers::utils")
 
 # Constants for OpenAI vector stores
 CHUNK_MULTIPLIER = 5

llama_stack/providers/utils/memory/vector_store.py

@@ -33,7 +33,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 )
 from llama_stack.providers.utils.vector_io.vector_utils import generate_chunk_id
 
-log = get_logger(name=__name__, category="memory")
+log = get_logger(name=__name__, category="providers::utils")
 
 
 class ChunkForDeletion(BaseModel):

llama_stack/providers/utils/scheduler.py

@@ -17,7 +17,7 @@ from pydantic import BaseModel
 
 from llama_stack.log import get_logger
 
-logger = get_logger(name=__name__, category="scheduler")
+logger = get_logger(name=__name__, category="providers::utils")
 
 
 # TODO: revisit the list of possible statuses when defining a more coherent

llama_stack/providers/utils/sqlstore/authorized_sqlstore.py

@@ -17,7 +17,7 @@ from llama_stack.log import get_logger
 from .api import ColumnDefinition, ColumnType, PaginatedResponse, SqlStore
 from .sqlstore import SqlStoreType
 
-logger = get_logger(name=__name__, category="authorized_sqlstore")
+logger = get_logger(name=__name__, category="providers::utils")
 
 # Hardcoded copy of the default policy that our SQL filtering implements
 # WARNING: If default_policy() changes, this constant must be updated accordingly

llama_stack/providers/utils/sqlstore/sqlalchemy_sqlstore.py

@@ -30,7 +30,7 @@ from llama_stack.log import get_logger
 from .api import ColumnDefinition, ColumnType, SqlStore
 from .sqlstore import SqlAlchemySqlStoreConfig
 
-logger = get_logger(name=__name__, category="sqlstore")
+logger = get_logger(name=__name__, category="providers::utils")
 
 TYPE_MAPPING: dict[ColumnType, Any] = {
     ColumnType.INTEGER: Integer,