Merge branch 'main' into feat/add-dana-agent-provider-stub

tests/unit/conversations/test_api_models.py

@@ -5,11 +5,7 @@
 # the root directory of this source tree.
 
 
-from llama_stack.apis.conversations.conversations import (
-    Conversation,
-    ConversationItem,
-    ConversationItemList,
-)
+from llama_stack_api import Conversation, ConversationItem, ConversationItemList
 
 
 def test_conversation_model_defaults():

tests/unit/conversations/test_conversations.py

@@ -12,10 +12,6 @@ from openai.types.conversations.conversation import Conversation as OpenAIConver
 from openai.types.conversations.conversation_item import ConversationItem as OpenAIConversationItem
 from pydantic import TypeAdapter
 
-from llama_stack.apis.agents.openai_responses import (
-    OpenAIResponseInputMessageContentText,
-    OpenAIResponseMessage,
-)
 from llama_stack.core.conversations.conversations import (
     ConversationServiceConfig,
     ConversationServiceImpl,
@@ -28,6 +24,7 @@ from llama_stack.core.storage.datatypes import (
     StorageConfig,
 )
 from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
+from llama_stack_api import OpenAIResponseInputMessageContentText, OpenAIResponseMessage
 
 
 @pytest.fixture

tests/unit/core/routers/test_safety_router.py

@@ -6,10 +6,9 @@
 
 from unittest.mock import AsyncMock
 
-from llama_stack.apis.safety.safety import ModerationObject, ModerationObjectResults
-from llama_stack.apis.shields import ListShieldsResponse, Shield
 from llama_stack.core.datatypes import SafetyConfig
 from llama_stack.core.routers.safety import SafetyRouter
+from llama_stack_api import ListShieldsResponse, ModerationObject, ModerationObjectResults, Shield
 
 
 async def test_run_moderation_uses_default_shield_when_model_missing():

tests/unit/core/routers/test_vector_io.py

@@ -8,8 +8,13 @@ from unittest.mock import AsyncMock, Mock
 
 import pytest
 
-from llama_stack.apis.vector_io import OpenAICreateVectorStoreRequestWithExtraBody
 from llama_stack.core.routers.vector_io import VectorIORouter
+from llama_stack_api import (
+    ModelNotFoundError,
+    ModelType,
+    ModelTypeError,
+    OpenAICreateVectorStoreRequestWithExtraBody,
+)
 
 
 async def test_single_provider_auto_selection():
@@ -21,6 +26,7 @@ async def test_single_provider_auto_selection():
             Mock(identifier="all-MiniLM-L6-v2", model_type="embedding", metadata={"embedding_dimension": 384})
         ]
     )
+    mock_routing_table.get_object_by_identifier = AsyncMock(return_value=Mock(model_type=ModelType.embedding))
     mock_routing_table.register_vector_store = AsyncMock(
         return_value=Mock(identifier="vs_123", provider_id="inline::faiss", provider_resource_id="vs_123")
     )
@@ -48,6 +54,7 @@ async def test_create_vector_stores_multiple_providers_missing_provider_id_error
             Mock(identifier="all-MiniLM-L6-v2", model_type="embedding", metadata={"embedding_dimension": 384})
         ]
     )
+    mock_routing_table.get_object_by_identifier = AsyncMock(return_value=Mock(model_type=ModelType.embedding))
     router = VectorIORouter(mock_routing_table)
     request = OpenAICreateVectorStoreRequestWithExtraBody.model_validate(
         {"name": "test_store", "embedding_model": "all-MiniLM-L6-v2"}
@@ -117,3 +124,32 @@ async def test_update_vector_store_same_provider_id_succeeds():
     provider.openai_update_vector_store.assert_called_once_with(
         vector_store_id="vs_123", name="updated_name", expires_after=None, metadata={"provider_id": "inline::faiss"}
     )
+
+
+async def test_create_vector_store_with_unknown_embedding_model_raises_error():
+    """Test that creating a vector store with an unknown embedding model raises
+    FoundError."""
+    mock_routing_table = Mock(impls_by_provider_id={"provider": "mock"})
+    mock_routing_table.get_object_by_identifier = AsyncMock(return_value=None)
+
+    router = VectorIORouter(mock_routing_table)
+    request = OpenAICreateVectorStoreRequestWithExtraBody.model_validate(
+        {"embedding_model": "unknown-model", "embedding_dimension": 384}
+    )
+
+    with pytest.raises(ModelNotFoundError, match="Model 'unknown-model' not found"):
+        await router.openai_create_vector_store(request)
+
+
+async def test_create_vector_store_with_wrong_model_type_raises_error():
+    """Test that creating a vector store with a non-embedding model raises ModelTypeError."""
+    mock_routing_table = Mock(impls_by_provider_id={"provider": "mock"})
+    mock_routing_table.get_object_by_identifier = AsyncMock(return_value=Mock(model_type=ModelType.llm))
+
+    router = VectorIORouter(mock_routing_table)
+    request = OpenAICreateVectorStoreRequestWithExtraBody.model_validate(
+        {"embedding_model": "text-model", "embedding_dimension": 384}
+    )
+
+    with pytest.raises(ModelTypeError, match="Model 'text-model' is of type"):
+        await router.openai_create_vector_store(request)

tests/unit/core/test_stack_validation.py

@@ -10,11 +10,9 @@ from unittest.mock import AsyncMock
 
 import pytest
 
-from llama_stack.apis.models import ListModelsResponse, Model, ModelType
-from llama_stack.apis.shields import ListShieldsResponse, Shield
 from llama_stack.core.datatypes import QualifiedModel, SafetyConfig, StackRunConfig, StorageConfig, VectorStoresConfig
 from llama_stack.core.stack import validate_safety_config, validate_vector_stores_config
-from llama_stack.providers.datatypes import Api
+from llama_stack_api import Api, ListModelsResponse, ListShieldsResponse, Model, ModelType, Shield
 
 
 class TestVectorStoresValidation:

tests/unit/distribution/routers/test_routing_tables.py

@@ -10,14 +10,6 @@ from unittest.mock import AsyncMock
 
 import pytest
 
-from llama_stack.apis.common.content_types import URL
-from llama_stack.apis.common.errors import ModelNotFoundError
-from llama_stack.apis.common.type_system import NumberType
-from llama_stack.apis.datasets.datasets import Dataset, DatasetPurpose, URIDataSource
-from llama_stack.apis.datatypes import Api
-from llama_stack.apis.models import Model, ModelType
-from llama_stack.apis.shields.shields import Shield
-from llama_stack.apis.tools import ListToolDefsResponse, ToolDef, ToolGroup
 from llama_stack.core.datatypes import RegistryEntrySource
 from llama_stack.core.routing_tables.benchmarks import BenchmarksRoutingTable
 from llama_stack.core.routing_tables.datasets import DatasetsRoutingTable
@@ -25,6 +17,21 @@ from llama_stack.core.routing_tables.models import ModelsRoutingTable
 from llama_stack.core.routing_tables.scoring_functions import ScoringFunctionsRoutingTable
 from llama_stack.core.routing_tables.shields import ShieldsRoutingTable
 from llama_stack.core.routing_tables.toolgroups import ToolGroupsRoutingTable
+from llama_stack_api import (
+    URL,
+    Api,
+    Dataset,
+    DatasetPurpose,
+    ListToolDefsResponse,
+    Model,
+    ModelNotFoundError,
+    ModelType,
+    NumberType,
+    Shield,
+    ToolDef,
+    ToolGroup,
+    URIDataSource,
+)
 
 
 class Impl:
@@ -130,7 +137,7 @@ class ToolGroupsImpl(Impl):
     async def unregister_toolgroup(self, toolgroup_id: str):
         return toolgroup_id
 
-    async def list_runtime_tools(self, toolgroup_id, mcp_endpoint):
+    async def list_runtime_tools(self, toolgroup_id, mcp_endpoint, authorization=None):
         return ListToolDefsResponse(
             data=[
                 ToolDef(

tests/unit/distribution/test_api_recordings.py

@@ -11,8 +11,15 @@ from unittest.mock import patch
 import pytest
 from openai import AsyncOpenAI
 
+from llama_stack.testing.api_recorder import (
+    APIRecordingMode,
+    ResponseStorage,
+    api_recording,
+    normalize_inference_request,
+)
+
 # Import the real Pydantic response types instead of using Mocks
-from llama_stack.apis.inference import (
+from llama_stack_api import (
     OpenAIAssistantMessageParam,
     OpenAIChatCompletion,
     OpenAIChoice,
@@ -20,12 +27,6 @@ from llama_stack.apis.inference import (
     OpenAIEmbeddingsResponse,
     OpenAIEmbeddingUsage,
 )
-from llama_stack.testing.api_recorder import (
-    APIRecordingMode,
-    ResponseStorage,
-    api_recording,
-    normalize_inference_request,
-)
 
 
 @pytest.fixture

tests/unit/distribution/test_distribution.py

@@ -22,7 +22,7 @@ from llama_stack.core.storage.datatypes import (
     SqlStoreReference,
     StorageConfig,
 )
-from llama_stack.providers.datatypes import ProviderSpec
+from llama_stack_api import ProviderSpec
 
 
 class SampleConfig(BaseModel):
@@ -312,7 +312,7 @@ pip_packages:
         """Test loading an external provider from a module (success path)."""
         from types import SimpleNamespace
 
-        from llama_stack.providers.datatypes import Api, ProviderSpec
+        from llama_stack_api import Api, ProviderSpec
 
         # Simulate a provider module with get_provider_spec
         fake_spec = ProviderSpec(
@@ -396,7 +396,7 @@ pip_packages:
     def test_external_provider_from_module_building(self, mock_providers):
         """Test loading an external provider from a module during build (building=True, partial spec)."""
         from llama_stack.core.datatypes import BuildConfig, BuildProvider, DistributionSpec
-        from llama_stack.providers.datatypes import Api
+        from llama_stack_api import Api
 
         # No importlib patch needed, should not import module when type of `config` is BuildConfig or DistributionSpec
         build_config = BuildConfig(
@@ -457,7 +457,7 @@ class TestGetExternalProvidersFromModule:
         from types import SimpleNamespace
 
         from llama_stack.core.distribution import get_external_providers_from_module
-        from llama_stack.providers.datatypes import ProviderSpec
+        from llama_stack_api import ProviderSpec
 
         fake_spec = ProviderSpec(
             api=Api.inference,
@@ -594,7 +594,7 @@ class TestGetExternalProvidersFromModule:
         from types import SimpleNamespace
 
         from llama_stack.core.distribution import get_external_providers_from_module
-        from llama_stack.providers.datatypes import ProviderSpec
+        from llama_stack_api import ProviderSpec
 
         spec1 = ProviderSpec(
             api=Api.inference,
@@ -642,7 +642,7 @@ class TestGetExternalProvidersFromModule:
         from types import SimpleNamespace
 
         from llama_stack.core.distribution import get_external_providers_from_module
-        from llama_stack.providers.datatypes import ProviderSpec
+        from llama_stack_api import ProviderSpec
 
         spec1 = ProviderSpec(
             api=Api.inference,
@@ -690,7 +690,7 @@ class TestGetExternalProvidersFromModule:
         from types import SimpleNamespace
 
         from llama_stack.core.distribution import get_external_providers_from_module
-        from llama_stack.providers.datatypes import ProviderSpec
+        from llama_stack_api import ProviderSpec
 
         # Module returns both inline and remote variants
         spec1 = ProviderSpec(
@@ -829,7 +829,7 @@ class TestGetExternalProvidersFromModule:
         from types import SimpleNamespace
 
         from llama_stack.core.distribution import get_external_providers_from_module
-        from llama_stack.providers.datatypes import ProviderSpec
+        from llama_stack_api import ProviderSpec
 
         inference_spec = ProviderSpec(
             api=Api.inference,

tests/unit/files/test_files.py

@@ -7,9 +7,6 @@
 
 import pytest
 
-from llama_stack.apis.common.errors import ResourceNotFoundError
-from llama_stack.apis.common.responses import Order
-from llama_stack.apis.files import OpenAIFilePurpose
 from llama_stack.core.access_control.access_control import default_policy
 from llama_stack.core.storage.datatypes import SqliteSqlStoreConfig, SqlStoreReference
 from llama_stack.providers.inline.files.localfs import (
@@ -17,6 +14,7 @@ from llama_stack.providers.inline.files.localfs import (
     LocalfsFilesImplConfig,
 )
 from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
+from llama_stack_api import OpenAIFilePurpose, Order, ResourceNotFoundError
 
 
 class MockUploadFile:

tests/unit/providers/batches/test_reference.py

@@ -59,8 +59,7 @@ from unittest.mock import AsyncMock, MagicMock
 
 import pytest
 
-from llama_stack.apis.batches import BatchObject
-from llama_stack.apis.common.errors import ConflictError, ResourceNotFoundError
+from llama_stack_api import BatchObject, ConflictError, ResourceNotFoundError
 
 
 class TestReferenceBatchesImpl:

tests/unit/providers/batches/test_reference_idempotency.py

@@ -44,7 +44,7 @@ import asyncio
 
 import pytest
 
-from llama_stack.apis.common.errors import ConflictError
+from llama_stack_api import ConflictError
 
 
 class TestReferenceBatchesIdempotency:

tests/unit/providers/files/test_s3_files.py

@@ -9,8 +9,7 @@ from unittest.mock import patch
 import pytest
 from botocore.exceptions import ClientError
 
-from llama_stack.apis.common.errors import ResourceNotFoundError
-from llama_stack.apis.files import OpenAIFilePurpose
+from llama_stack_api import OpenAIFilePurpose, ResourceNotFoundError
 
 
 class TestS3FilesImpl:
@@ -228,7 +227,7 @@ class TestS3FilesImpl:
 
             mock_now.return_value = 0
 
-            from llama_stack.apis.files import ExpiresAfter
+            from llama_stack_api import ExpiresAfter
 
             sample_text_file.filename = "test_expired_file"
             uploaded = await s3_provider.openai_upload_file(
@@ -260,7 +259,7 @@ class TestS3FilesImpl:
 
     async def test_unsupported_expires_after_anchor(self, s3_provider, sample_text_file):
         """Unsupported anchor value should raise ValueError."""
-        from llama_stack.apis.files import ExpiresAfter
+        from llama_stack_api import ExpiresAfter
 
         sample_text_file.filename = "test_unsupported_expires_after_anchor"
 
@@ -273,7 +272,7 @@ class TestS3FilesImpl:
 
     async def test_nonint_expires_after_seconds(self, s3_provider, sample_text_file):
         """Non-integer seconds in expires_after should raise ValueError."""
-        from llama_stack.apis.files import ExpiresAfter
+        from llama_stack_api import ExpiresAfter
 
         sample_text_file.filename = "test_nonint_expires_after_seconds"
 
@@ -286,7 +285,7 @@ class TestS3FilesImpl:
 
     async def test_expires_after_seconds_out_of_bounds(self, s3_provider, sample_text_file):
         """Seconds outside allowed range should raise ValueError."""
-        from llama_stack.apis.files import ExpiresAfter
+        from llama_stack_api import ExpiresAfter
 
         with pytest.raises(ValueError, match="greater than or equal to 3600"):
             await s3_provider.openai_upload_file(

tests/unit/providers/files/test_s3_files_auth.py

@@ -8,10 +8,9 @@ from unittest.mock import patch
 
 import pytest
 
-from llama_stack.apis.common.errors import ResourceNotFoundError
-from llama_stack.apis.files import OpenAIFilePurpose
 from llama_stack.core.datatypes import User
 from llama_stack.providers.remote.files.s3.files import S3FilesImpl
+from llama_stack_api import OpenAIFilePurpose, ResourceNotFoundError
 
 
 async def test_listing_hides_other_users_file(s3_provider, sample_text_file):

tests/unit/providers/inference/test_bedrock_adapter.py

@@ -10,9 +10,9 @@ from unittest.mock import AsyncMock, MagicMock
 import pytest
 from openai import AuthenticationError
 
-from llama_stack.apis.inference import OpenAIChatCompletionRequestWithExtraBody
 from llama_stack.providers.remote.inference.bedrock.bedrock import BedrockInferenceAdapter
 from llama_stack.providers.remote.inference.bedrock.config import BedrockConfig
+from llama_stack_api import OpenAIChatCompletionRequestWithExtraBody
 
 
 def test_adapter_initialization():

tests/unit/providers/inference/test_remote_vllm.py

@@ -10,7 +10,13 @@ from unittest.mock import AsyncMock, MagicMock, PropertyMock, patch
 
 import pytest
 
-from llama_stack.apis.inference import (
+from llama_stack.core.routers.inference import InferenceRouter
+from llama_stack.core.routing_tables.models import ModelsRoutingTable
+from llama_stack.providers.remote.inference.vllm.config import VLLMInferenceAdapterConfig
+from llama_stack.providers.remote.inference.vllm.vllm import VLLMInferenceAdapter
+from llama_stack_api import (
+    HealthStatus,
+    Model,
     OpenAIAssistantMessageParam,
     OpenAIChatCompletion,
     OpenAIChatCompletionRequestWithExtraBody,
@@ -20,12 +26,6 @@ from llama_stack.apis.inference import (
     OpenAICompletionRequestWithExtraBody,
     ToolChoice,
 )
-from llama_stack.apis.models import Model
-from llama_stack.core.routers.inference import InferenceRouter
-from llama_stack.core.routing_tables.models import ModelsRoutingTable
-from llama_stack.providers.datatypes import HealthStatus
-from llama_stack.providers.remote.inference.vllm.config import VLLMInferenceAdapterConfig
-from llama_stack.providers.remote.inference.vllm.vllm import VLLMInferenceAdapter
 
 # These are unit test for the remote vllm provider
 # implementation. This should only contain tests which are specific to

tests/unit/providers/inline/agents/meta_reference/responses/test_streaming.py

@@ -8,11 +8,11 @@ from unittest.mock import AsyncMock
 
 import pytest
 
-from llama_stack.apis.tools import ToolDef
 from llama_stack.providers.inline.agents.meta_reference.responses.streaming import (
     convert_tooldef_to_chat_tool,
 )
 from llama_stack.providers.inline.agents.meta_reference.responses.types import ChatCompletionContext
+from llama_stack_api import ToolDef
 
 
 @pytest.fixture

tests/unit/providers/nvidia/test_datastore.py

@@ -9,10 +9,9 @@ from unittest.mock import patch
 
 import pytest
 
-from llama_stack.apis.datasets import Dataset, DatasetPurpose, URIDataSource
-from llama_stack.apis.resource import ResourceType
 from llama_stack.providers.remote.datasetio.nvidia.config import NvidiaDatasetIOConfig
 from llama_stack.providers.remote.datasetio.nvidia.datasetio import NvidiaDatasetIOAdapter
+from llama_stack_api import Dataset, DatasetPurpose, ResourceType, URIDataSource
 
 
 @pytest.fixture

tests/unit/providers/nvidia/test_eval.py

@@ -9,14 +9,20 @@ from unittest.mock import MagicMock, patch
 
 import pytest
 
-from llama_stack.apis.benchmarks import Benchmark
-from llama_stack.apis.common.job_types import Job, JobStatus
-from llama_stack.apis.eval.eval import BenchmarkConfig, EvaluateResponse, ModelCandidate, SamplingParams
-from llama_stack.apis.inference.inference import TopPSamplingStrategy
-from llama_stack.apis.resource import ResourceType
 from llama_stack.models.llama.sku_types import CoreModelId
 from llama_stack.providers.remote.eval.nvidia.config import NVIDIAEvalConfig
 from llama_stack.providers.remote.eval.nvidia.eval import NVIDIAEvalImpl
+from llama_stack_api import (
+    Benchmark,
+    BenchmarkConfig,
+    EvaluateResponse,
+    Job,
+    JobStatus,
+    ModelCandidate,
+    ResourceType,
+    SamplingParams,
+    TopPSamplingStrategy,
+)
 
 MOCK_DATASET_ID = "default/test-dataset"
 MOCK_BENCHMARK_ID = "test-benchmark"

tests/unit/providers/nvidia/test_parameters.py

@@ -10,7 +10,12 @@ from unittest.mock import patch
 
 import pytest
 
-from llama_stack.apis.post_training.post_training import (
+from llama_stack.core.library_client import convert_pydantic_to_json_value
+from llama_stack.providers.remote.post_training.nvidia.post_training import (
+    NvidiaPostTrainingAdapter,
+    NvidiaPostTrainingConfig,
+)
+from llama_stack_api import (
     DataConfig,
     DatasetFormat,
     EfficiencyConfig,
@@ -19,11 +24,6 @@ from llama_stack.apis.post_training.post_training import (
     OptimizerType,
     TrainingConfig,
 )
-from llama_stack.core.library_client import convert_pydantic_to_json_value
-from llama_stack.providers.remote.post_training.nvidia.post_training import (
-    NvidiaPostTrainingAdapter,
-    NvidiaPostTrainingConfig,
-)
 
 
 class TestNvidiaParameters:

tests/unit/providers/nvidia/test_rerank_inference.py

@@ -9,10 +9,10 @@ from unittest.mock import AsyncMock, MagicMock, patch
 import aiohttp
 import pytest
 
-from llama_stack.apis.models import ModelType
 from llama_stack.providers.remote.inference.nvidia.config import NVIDIAConfig
 from llama_stack.providers.remote.inference.nvidia.nvidia import NVIDIAInferenceAdapter
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
+from llama_stack_api import ModelType
 
 
 class MockResponse:

tests/unit/providers/nvidia/test_safety.py

@@ -10,15 +10,16 @@ from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
 
-from llama_stack.apis.inference import (
-    OpenAIAssistantMessageParam,
-    OpenAIUserMessageParam,
-)
-from llama_stack.apis.resource import ResourceType
-from llama_stack.apis.safety import RunShieldResponse, ViolationLevel
-from llama_stack.apis.shields import Shield
 from llama_stack.providers.remote.safety.nvidia.config import NVIDIASafetyConfig
 from llama_stack.providers.remote.safety.nvidia.nvidia import NVIDIASafetyAdapter
+from llama_stack_api import (
+    OpenAIAssistantMessageParam,
+    OpenAIUserMessageParam,
+    ResourceType,
+    RunShieldResponse,
+    Shield,
+    ViolationLevel,
+)
 
 
 class FakeNVIDIASafetyAdapter(NVIDIASafetyAdapter):

tests/unit/providers/nvidia/test_supervised_fine_tuning.py

@@ -10,15 +10,6 @@ from unittest.mock import patch
 
 import pytest
 
-from llama_stack.apis.post_training.post_training import (
-    DataConfig,
-    DatasetFormat,
-    LoraFinetuningConfig,
-    OptimizerConfig,
-    OptimizerType,
-    QATFinetuningConfig,
-    TrainingConfig,
-)
 from llama_stack.core.library_client import convert_pydantic_to_json_value
 from llama_stack.providers.remote.post_training.nvidia.post_training import (
     ListNvidiaPostTrainingJobs,
@@ -27,6 +18,15 @@ from llama_stack.providers.remote.post_training.nvidia.post_training import (
     NvidiaPostTrainingJob,
     NvidiaPostTrainingJobStatusResponse,
 )
+from llama_stack_api import (
+    DataConfig,
+    DatasetFormat,
+    LoraFinetuningConfig,
+    OptimizerConfig,
+    OptimizerType,
+    QATFinetuningConfig,
+    TrainingConfig,
+)
 
 
 @pytest.fixture

tests/unit/providers/test_bedrock.py

@@ -7,9 +7,9 @@
 from types import SimpleNamespace
 from unittest.mock import AsyncMock, PropertyMock, patch
 
-from llama_stack.apis.inference import OpenAIChatCompletionRequestWithExtraBody
 from llama_stack.providers.remote.inference.bedrock.bedrock import BedrockInferenceAdapter
 from llama_stack.providers.remote.inference.bedrock.config import BedrockConfig
+from llama_stack_api import OpenAIChatCompletionRequestWithExtraBody
 
 
 def test_can_create_adapter():

tests/unit/providers/utils/inference/test_openai_mixin.py

@@ -12,11 +12,10 @@ from unittest.mock import AsyncMock, MagicMock, Mock, PropertyMock, patch
 import pytest
 from pydantic import BaseModel, Field
 
-from llama_stack.apis.inference import Model, OpenAIChatCompletionRequestWithExtraBody, OpenAIUserMessageParam
-from llama_stack.apis.models import ModelType
 from llama_stack.core.request_headers import request_provider_data_context
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
+from llama_stack_api import Model, ModelType, OpenAIChatCompletionRequestWithExtraBody, OpenAIUserMessageParam
 
 
 class OpenAIMixinImpl(OpenAIMixin):

tests/unit/providers/utils/inference/test_prompt_adapter.py

@@ -4,14 +4,11 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from llama_stack.apis.inference import (
-    OpenAIAssistantMessageParam,
-    OpenAIUserMessageParam,
-)
 from llama_stack.models.llama.datatypes import RawTextItem
 from llama_stack.providers.utils.inference.prompt_adapter import (
     convert_openai_message_to_raw_message,
 )
+from llama_stack_api import OpenAIAssistantMessageParam, OpenAIUserMessageParam
 
 
 class TestConvertOpenAIMessageToRawMessage:

tests/unit/providers/utils/memory/test_vector_store.py

@@ -8,9 +8,8 @@ from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
 
-from llama_stack.apis.common.content_types import URL, TextContentItem
-from llama_stack.apis.tools import RAGDocument
 from llama_stack.providers.utils.memory.vector_store import content_from_data_and_mime_type, content_from_doc
+from llama_stack_api import URL, RAGDocument, TextContentItem
 
 
 async def test_content_from_doc_with_url():

tests/unit/providers/utils/test_model_registry.py

@@ -35,8 +35,8 @@
 
 import pytest
 
-from llama_stack.apis.models import Model
 from llama_stack.providers.utils.inference.model_registry import ModelRegistryHelper, ProviderModelEntry
+from llama_stack_api import Model
 
 
 @pytest.fixture

tests/unit/providers/vector_io/conftest.py

@@ -10,8 +10,6 @@ from unittest.mock import AsyncMock, MagicMock, patch
 import numpy as np
 import pytest
 
-from llama_stack.apis.vector_io import Chunk, ChunkMetadata, QueryChunksResponse
-from llama_stack.apis.vector_stores import VectorStore
 from llama_stack.core.storage.datatypes import KVStoreReference, SqliteKVStoreConfig
 from llama_stack.providers.inline.vector_io.faiss.config import FaissVectorIOConfig
 from llama_stack.providers.inline.vector_io.faiss.faiss import FaissIndex, FaissVectorIOAdapter
@@ -20,6 +18,7 @@ from llama_stack.providers.inline.vector_io.sqlite_vec.sqlite_vec import SQLiteV
 from llama_stack.providers.remote.vector_io.pgvector.config import PGVectorVectorIOConfig
 from llama_stack.providers.remote.vector_io.pgvector.pgvector import PGVectorIndex, PGVectorVectorIOAdapter
 from llama_stack.providers.utils.kvstore import register_kvstore_backends
+from llama_stack_api import Chunk, ChunkMetadata, QueryChunksResponse, VectorStore
 
 EMBEDDING_DIMENSION = 768
 COLLECTION_PREFIX = "test_collection"

tests/unit/providers/vector_io/test_faiss.py

@@ -10,15 +10,12 @@ from unittest.mock import MagicMock, patch
 import numpy as np
 import pytest
 
-from llama_stack.apis.files import Files
-from llama_stack.apis.vector_io import Chunk, QueryChunksResponse
-from llama_stack.apis.vector_stores import VectorStore
-from llama_stack.providers.datatypes import HealthStatus
 from llama_stack.providers.inline.vector_io.faiss.config import FaissVectorIOConfig
 from llama_stack.providers.inline.vector_io.faiss.faiss import (
     FaissIndex,
     FaissVectorIOAdapter,
 )
+from llama_stack_api import Chunk, Files, HealthStatus, QueryChunksResponse, VectorStore
 
 # This test is a unit test for the FaissVectorIOAdapter class. This should only contain
 # tests which are specific to this class. More general (API-level) tests should be placed in

tests/unit/providers/vector_io/test_sqlite_vec.py

@@ -9,12 +9,12 @@ import asyncio
 import numpy as np
 import pytest
 
-from llama_stack.apis.vector_io import Chunk, QueryChunksResponse
 from llama_stack.providers.inline.vector_io.sqlite_vec.sqlite_vec import (
     SQLiteVecIndex,
     SQLiteVecVectorIOAdapter,
     _create_sqlite_connection,
 )
+from llama_stack_api import Chunk, QueryChunksResponse
 
 # This test is a unit test for the SQLiteVecVectorIOAdapter class. This should only contain
 # tests which are specific to this class. More general (API-level) tests should be placed in

tests/unit/providers/vector_io/test_vector_io_openai_vector_stores.py

@@ -11,17 +11,17 @@ from unittest.mock import AsyncMock, patch
 import numpy as np
 import pytest
 
-from llama_stack.apis.common.errors import VectorStoreNotFoundError
-from llama_stack.apis.vector_io import (
+from llama_stack.providers.inline.vector_io.sqlite_vec.sqlite_vec import VECTOR_DBS_PREFIX
+from llama_stack_api import (
     Chunk,
     OpenAICreateVectorStoreFileBatchRequestWithExtraBody,
     OpenAICreateVectorStoreRequestWithExtraBody,
     QueryChunksResponse,
+    VectorStore,
     VectorStoreChunkingStrategyAuto,
     VectorStoreFileObject,
+    VectorStoreNotFoundError,
 )
-from llama_stack.apis.vector_stores import VectorStore
-from llama_stack.providers.inline.vector_io.sqlite_vec.sqlite_vec import VECTOR_DBS_PREFIX
 
 # This test is a unit test for the inline VectorIO providers. This should only contain
 # tests which are specific to this class. More general (API-level) tests should be placed in
@@ -222,7 +222,7 @@ async def test_insert_chunks_missing_db_raises(vector_io_adapter):
 
 async def test_insert_chunks_with_missing_document_id(vector_io_adapter):
     """Ensure no KeyError when document_id is missing or in different places."""
-    from llama_stack.apis.vector_io import Chunk, ChunkMetadata
+    from llama_stack_api import Chunk, ChunkMetadata
 
     fake_index = AsyncMock()
     vector_io_adapter.cache["db1"] = fake_index
@@ -255,10 +255,9 @@ async def test_insert_chunks_with_missing_document_id(vector_io_adapter):
 
 async def test_document_id_with_invalid_type_raises_error():
     """Ensure TypeError is raised when document_id is not a string."""
-    from llama_stack.apis.vector_io import Chunk
-
     # Integer document_id should raise TypeError
     from llama_stack.providers.utils.vector_io.vector_utils import generate_chunk_id
+    from llama_stack_api import Chunk
 
     chunk = Chunk(content="test", chunk_id=generate_chunk_id("test", "test"), metadata={"document_id": 12345})
     with pytest.raises(TypeError) as exc_info:

tests/unit/providers/vector_io/test_vector_utils.py

@@ -4,8 +4,8 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from llama_stack.apis.vector_io import Chunk, ChunkMetadata
 from llama_stack.providers.utils.vector_io.vector_utils import generate_chunk_id
+from llama_stack_api import Chunk, ChunkMetadata
 
 # This test is a unit test for the chunk_utils.py helpers. This should only contain
 # tests which are specific to this file. More general (API-level) tests should be placed in

tests/unit/rag/test_rag_query.py

@@ -8,13 +8,8 @@ from unittest.mock import AsyncMock, MagicMock
 
 import pytest
 
-from llama_stack.apis.tools.rag_tool import RAGQueryConfig
-from llama_stack.apis.vector_io import (
-    Chunk,
-    ChunkMetadata,
-    QueryChunksResponse,
-)
 from llama_stack.providers.inline.tool_runtime.rag.memory import MemoryToolRuntimeImpl
+from llama_stack_api import Chunk, ChunkMetadata, QueryChunksResponse, RAGQueryConfig
 
 
 class TestRagQuery:

tests/unit/rag/test_vector_store.py

@@ -13,12 +13,6 @@ from unittest.mock import AsyncMock, MagicMock
 import numpy as np
 import pytest
 
-from llama_stack.apis.inference.inference import (
-    OpenAIEmbeddingData,
-    OpenAIEmbeddingsRequestWithExtraBody,
-)
-from llama_stack.apis.tools import RAGDocument
-from llama_stack.apis.vector_io import Chunk
 from llama_stack.providers.utils.memory.vector_store import (
     URL,
     VectorStoreWithIndex,
@@ -27,6 +21,7 @@ from llama_stack.providers.utils.memory.vector_store import (
     make_overlapped_chunks,
 )
 from llama_stack.providers.utils.vector_io.vector_utils import generate_chunk_id
+from llama_stack_api import Chunk, OpenAIEmbeddingData, OpenAIEmbeddingsRequestWithExtraBody, RAGDocument
 
 DUMMY_PDF_PATH = Path(os.path.abspath(__file__)).parent / "fixtures" / "dummy.pdf"
 # Depending on the machine, this can get parsed a couple of ways

tests/unit/registry/test_registry.py

@@ -7,8 +7,6 @@
 
 import pytest
 
-from llama_stack.apis.inference import Model
-from llama_stack.apis.vector_stores import VectorStore
 from llama_stack.core.datatypes import VectorStoreWithOwner
 from llama_stack.core.storage.datatypes import KVStoreReference, SqliteKVStoreConfig
 from llama_stack.core.store.registry import (
@@ -17,6 +15,7 @@ from llama_stack.core.store.registry import (
     DiskDistributionRegistry,
 )
 from llama_stack.providers.utils.kvstore import kvstore_impl, register_kvstore_backends
+from llama_stack_api import Model, VectorStore
 
 
 @pytest.fixture
@@ -304,8 +303,8 @@ async def test_double_registration_different_objects(disk_dist_registry):
 
 async def test_double_registration_with_cache(cached_disk_dist_registry):
     """Test double registration behavior with caching enabled."""
-    from llama_stack.apis.models import ModelType
     from llama_stack.core.datatypes import ModelWithOwner
+    from llama_stack_api import ModelType
 
     model1 = ModelWithOwner(
         identifier="test_model",

tests/unit/registry/test_registry_acl.py

@@ -5,9 +5,9 @@
 # the root directory of this source tree.
 
 
-from llama_stack.apis.models import ModelType
 from llama_stack.core.datatypes import ModelWithOwner, User
 from llama_stack.core.store.registry import CachedDiskDistributionRegistry
+from llama_stack_api import ModelType
 
 
 async def test_registry_cache_with_acl(cached_disk_dist_registry):

tests/unit/server/test_access_control.py

@@ -10,11 +10,10 @@ import pytest
 import yaml
 from pydantic import TypeAdapter, ValidationError
 
-from llama_stack.apis.datatypes import Api
-from llama_stack.apis.models import ModelType
 from llama_stack.core.access_control.access_control import AccessDeniedError, is_action_allowed
 from llama_stack.core.datatypes import AccessRule, ModelWithOwner, User
 from llama_stack.core.routing_tables.models import ModelsRoutingTable
+from llama_stack_api import Api, ModelType
 
 
 class AsyncMock(MagicMock):

tests/unit/server/test_auth.py

@@ -144,7 +144,7 @@ def middleware_with_mocks(mock_auth_endpoint):
     middleware = AuthenticationMiddleware(mock_app, auth_config, {})
 
     # Mock the route_impls to simulate finding routes with required scopes
-    from llama_stack.schema_utils import WebMethod
+    from llama_stack_api import WebMethod
 
     routes = {
         ("POST", "/test/scoped"): WebMethod(route="/test/scoped", method="POST", required_scope="test.read"),

tests/unit/server/test_resolver.py

@@ -11,7 +11,6 @@ from unittest.mock import AsyncMock, MagicMock
 
 from pydantic import BaseModel, Field
 
-from llama_stack.apis.inference import Inference
 from llama_stack.core.datatypes import Api, Provider, StackRunConfig
 from llama_stack.core.resolver import resolve_impls
 from llama_stack.core.routers.inference import InferenceRouter
@@ -25,9 +24,9 @@ from llama_stack.core.storage.datatypes import (
     SqlStoreReference,
     StorageConfig,
 )
-from llama_stack.providers.datatypes import InlineProviderSpec, ProviderSpec
 from llama_stack.providers.utils.kvstore import register_kvstore_backends
 from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
+from llama_stack_api import Inference, InlineProviderSpec, ProviderSpec
 
 
 def add_protocol_methods(cls: type, protocol: type[Protocol]) -> None:

tests/unit/server/test_server.py

@@ -12,7 +12,7 @@ from pydantic import ValidationError
 
 from llama_stack.core.access_control.access_control import AccessDeniedError
 from llama_stack.core.datatypes import AuthenticationRequiredError
-from llama_stack.core.server.server import translate_exception
+from llama_stack.core.server.server import remove_disabled_providers, translate_exception
 
 
 class TestTranslateException:
@@ -194,3 +194,70 @@ class TestTranslateException:
         assert isinstance(result3, HTTPException)
         assert result3.status_code == 403
         assert result3.detail == "Permission denied: Access denied"
+
+
+class TestRemoveDisabledProviders:
+    """Test cases for the remove_disabled_providers function."""
+
+    def test_remove_explicitly_disabled_provider(self):
+        """Test that providers with provider_id='__disabled__' are removed."""
+        config = {
+            "providers": {
+                "inference": [
+                    {"provider_id": "openai", "provider_type": "remote::openai", "config": {}},
+                    {"provider_id": "__disabled__", "provider_type": "remote::vllm", "config": {}},
+                ]
+            }
+        }
+        result = remove_disabled_providers(config)
+        assert len(result["providers"]["inference"]) == 1
+        assert result["providers"]["inference"][0]["provider_id"] == "openai"
+
+    def test_remove_empty_provider_id(self):
+        """Test that providers with empty provider_id are removed."""
+        config = {
+            "providers": {
+                "inference": [
+                    {"provider_id": "openai", "provider_type": "remote::openai", "config": {}},
+                    {"provider_id": "", "provider_type": "remote::vllm", "config": {}},
+                ]
+            }
+        }
+        result = remove_disabled_providers(config)
+        assert len(result["providers"]["inference"]) == 1
+        assert result["providers"]["inference"][0]["provider_id"] == "openai"
+
+    def test_keep_models_with_none_provider_model_id(self):
+        """Test that models with None provider_model_id are NOT removed."""
+        config = {
+            "registered_resources": {
+                "models": [
+                    {
+                        "model_id": "llama-3-2-3b",
+                        "provider_id": "vllm-inference",
+                        "model_type": "llm",
+                        "provider_model_id": None,
+                        "metadata": {},
+                    },
+                    {
+                        "model_id": "gpt-4o-mini",
+                        "provider_id": "openai",
+                        "model_type": "llm",
+                        "provider_model_id": None,
+                        "metadata": {},
+                    },
+                    {
+                        "model_id": "granite-embedding-125m",
+                        "provider_id": "sentence-transformers",
+                        "model_type": "embedding",
+                        "provider_model_id": "ibm-granite/granite-embedding-125m-english",
+                        "metadata": {"embedding_dimension": 768},
+                    },
+                ]
+            }
+        }
+        result = remove_disabled_providers(config)
+        assert len(result["registered_resources"]["models"]) == 3
+        assert result["registered_resources"]["models"][0]["model_id"] == "llama-3-2-3b"
+        assert result["registered_resources"]["models"][1]["model_id"] == "gpt-4o-mini"
+        assert result["registered_resources"]["models"][2]["model_id"] == "granite-embedding-125m"

tests/unit/server/test_sse.py

@@ -10,8 +10,8 @@ from unittest.mock import AsyncMock, MagicMock
 
 import pytest
 
-from llama_stack.apis.common.responses import PaginatedResponse
 from llama_stack.core.server.server import create_dynamic_typed_route, create_sse_event, sse_generator
+from llama_stack_api import PaginatedResponse
 
 
 @pytest.fixture

tests/unit/tools/test_tools_json_schema.py

@@ -11,8 +11,8 @@ Tests the new input_schema and output_schema fields.
 
 from pydantic import ValidationError
 
-from llama_stack.apis.tools import ToolDef
 from llama_stack.models.llama.datatypes import BuiltinTool, ToolDefinition
+from llama_stack_api import ToolDef
 
 
 class TestToolDefValidation:

tests/unit/utils/inference/test_inference_store.py

@@ -8,16 +8,16 @@ import time
 
 import pytest
 
-from llama_stack.apis.inference import (
+from llama_stack.core.storage.datatypes import InferenceStoreReference, SqliteSqlStoreConfig
+from llama_stack.providers.utils.inference.inference_store import InferenceStore
+from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
+from llama_stack_api import (
     OpenAIAssistantMessageParam,
     OpenAIChatCompletion,
     OpenAIChoice,
     OpenAIUserMessageParam,
     Order,
 )
-from llama_stack.core.storage.datatypes import InferenceStoreReference, SqliteSqlStoreConfig
-from llama_stack.providers.utils.inference.inference_store import InferenceStore
-from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
 
 
 @pytest.fixture(autouse=True)

tests/unit/utils/responses/test_responses_store.py

@@ -10,15 +10,10 @@ from uuid import uuid4
 
 import pytest
 
-from llama_stack.apis.agents import Order
-from llama_stack.apis.agents.openai_responses import (
-    OpenAIResponseInput,
-    OpenAIResponseObject,
-)
-from llama_stack.apis.inference import OpenAIMessageParam, OpenAIUserMessageParam
 from llama_stack.core.storage.datatypes import ResponsesStoreReference, SqliteSqlStoreConfig
 from llama_stack.providers.utils.responses.responses_store import ResponsesStore
 from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
+from llama_stack_api import OpenAIMessageParam, OpenAIResponseInput, OpenAIResponseObject, OpenAIUserMessageParam, Order
 
 
 def build_store(db_path: str, policy: list | None = None) -> ResponsesStore:
@@ -46,7 +41,7 @@ def create_test_response_object(
 
 def create_test_response_input(content: str, input_id: str) -> OpenAIResponseInput:
     """Helper to create a test response input."""
-    from llama_stack.apis.agents.openai_responses import OpenAIResponseMessage
+    from llama_stack_api import OpenAIResponseMessage
 
     return OpenAIResponseMessage(
         id=input_id,