Merge branch 'main' into vectordb_name

llama_stack/providers/remote/inference/groq/groq.py

@@ -38,24 +38,18 @@ class GroqInferenceAdapter(LiteLLMOpenAIMixin):
             provider_data_api_key_field="groq_api_key",
         )
         self.config = config
-        self._openai_client = None
 
     async def initialize(self):
         await super().initialize()
 
     async def shutdown(self):
         await super().shutdown()
-        if self._openai_client:
-            await self._openai_client.close()
-            self._openai_client = None
 
     def _get_openai_client(self) -> AsyncOpenAI:
-        if not self._openai_client:
-            self._openai_client = AsyncOpenAI(
-                base_url=f"{self.config.url}/openai/v1",
-                api_key=self.config.api_key,
-            )
-        return self._openai_client
+        return AsyncOpenAI(
+            base_url=f"{self.config.url}/openai/v1",
+            api_key=self.get_api_key(),
+        )
 
     async def openai_chat_completion(
         self,

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

@@ -59,9 +59,6 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
         # if we do not set this, users will be exposed to the
         # litellm specific model names, an abstraction leak.
         self.is_openai_compat = True
-        self._openai_client = AsyncOpenAI(
-            api_key=self.config.api_key,
-        )
 
     async def initialize(self) -> None:
         await super().initialize()
@@ -69,6 +66,11 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
     async def shutdown(self) -> None:
         await super().shutdown()
 
+    def _get_openai_client(self) -> AsyncOpenAI:
+        return AsyncOpenAI(
+            api_key=self.get_api_key(),
+        )
+
     async def openai_completion(
         self,
         model: str,
@@ -120,7 +122,7 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
             user=user,
             suffix=suffix,
         )
-        return await self._openai_client.completions.create(**params)
+        return await self._get_openai_client().completions.create(**params)
 
     async def openai_chat_completion(
         self,
@@ -176,7 +178,7 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
             top_p=top_p,
             user=user,
         )
-        return await self._openai_client.chat.completions.create(**params)
+        return await self._get_openai_client().chat.completions.create(**params)
 
     async def openai_embeddings(
         self,
@@ -204,7 +206,7 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
             params["user"] = user
 
         # Call OpenAI embeddings API
-        response = await self._openai_client.embeddings.create(**params)
+        response = await self._get_openai_client().embeddings.create(**params)
 
         data = []
         for i, embedding_data in enumerate(response.data):

llama_stack/providers/remote/inference/sambanova/sambanova.py

@@ -7,6 +7,7 @@
 import json
 from collections.abc import Iterable
 
+import requests
 from openai.types.chat import (
     ChatCompletionAssistantMessageParam as OpenAIChatCompletionAssistantMessage,
 )
@@ -56,6 +57,7 @@ from llama_stack.apis.inference import (
     ToolResponseMessage,
     UserMessage,
 )
+from llama_stack.apis.models import Model
 from llama_stack.log import get_logger
 from llama_stack.models.llama.datatypes import BuiltinTool
 from llama_stack.providers.utils.inference.litellm_openai_mixin import LiteLLMOpenAIMixin
@@ -176,10 +178,11 @@ class SambaNovaInferenceAdapter(LiteLLMOpenAIMixin):
 
     def __init__(self, config: SambaNovaImplConfig):
         self.config = config
+        self.environment_available_models = []
         LiteLLMOpenAIMixin.__init__(
             self,
             model_entries=MODEL_ENTRIES,
-            api_key_from_config=self.config.api_key,
+            api_key_from_config=self.config.api_key.get_secret_value() if self.config.api_key else None,
             provider_data_api_key_field="sambanova_api_key",
         )
 
@@ -246,6 +249,22 @@ class SambaNovaInferenceAdapter(LiteLLMOpenAIMixin):
             **get_sampling_options(request.sampling_params),
         }
 
+    async def register_model(self, model: Model) -> Model:
+        model_id = self.get_provider_model_id(model.provider_resource_id)
+
+        list_models_url = self.config.url + "/models"
+        if len(self.environment_available_models) == 0:
+            try:
+                response = requests.get(list_models_url)
+                response.raise_for_status()
+            except requests.exceptions.RequestException as e:
+                raise RuntimeError(f"Request to {list_models_url} failed") from e
+            self.environment_available_models = [model.get("id") for model in response.json().get("data", {})]
+
+        if model_id.split("sambanova/")[-1] not in self.environment_available_models:
+            logger.warning(f"Model {model_id} not available in {list_models_url}")
+        return model
+
     async def initialize(self):
         await super().initialize()
 

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

@@ -68,19 +68,12 @@ class TogetherInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProvi
     def __init__(self, config: TogetherImplConfig) -> None:
         ModelRegistryHelper.__init__(self, MODEL_ENTRIES)
         self.config = config
-        self._client = None
-        self._openai_client = None
 
     async def initialize(self) -> None:
         pass
 
     async def shutdown(self) -> None:
-        if self._client:
-            # Together client has no close method, so just set to None
-            self._client = None
-        if self._openai_client:
-            await self._openai_client.close()
-            self._openai_client = None
+        pass
 
     async def completion(
         self,
@@ -108,29 +101,25 @@ class TogetherInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProvi
             return await self._nonstream_completion(request)
 
     def _get_client(self) -> AsyncTogether:
-        if not self._client:
-            together_api_key = None
-            config_api_key = self.config.api_key.get_secret_value() if self.config.api_key else None
-            if config_api_key:
-                together_api_key = config_api_key
-            else:
-                provider_data = self.get_request_provider_data()
-                if provider_data is None or not provider_data.together_api_key:
-                    raise ValueError(
-                        'Pass Together API Key in the header X-LlamaStack-Provider-Data as { "together_api_key": <your api key>}'
-                    )
-                together_api_key = provider_data.together_api_key
-            self._client = AsyncTogether(api_key=together_api_key)
-        return self._client
+        together_api_key = None
+        config_api_key = self.config.api_key.get_secret_value() if self.config.api_key else None
+        if config_api_key:
+            together_api_key = config_api_key
+        else:
+            provider_data = self.get_request_provider_data()
+            if provider_data is None or not provider_data.together_api_key:
+                raise ValueError(
+                    'Pass Together API Key in the header X-LlamaStack-Provider-Data as { "together_api_key": <your api key>}'
+                )
+            together_api_key = provider_data.together_api_key
+        return AsyncTogether(api_key=together_api_key)
 
     def _get_openai_client(self) -> AsyncOpenAI:
-        if not self._openai_client:
-            together_client = self._get_client().client
-            self._openai_client = AsyncOpenAI(
-                base_url=together_client.base_url,
-                api_key=together_client.api_key,
-            )
-        return self._openai_client
+        together_client = self._get_client().client
+        return AsyncOpenAI(
+            base_url=together_client.base_url,
+            api_key=together_client.api_key,
+        )
 
     async def _nonstream_completion(self, request: CompletionRequest) -> ChatCompletionResponse:
         params = await self._get_params(request)

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

@@ -33,6 +33,7 @@ CANNED_RESPONSE_TEXT = "I can't answer that. Can I help with something else?"
 class SambaNovaSafetyAdapter(Safety, ShieldsProtocolPrivate, NeedsRequestProviderData):
     def __init__(self, config: SambaNovaSafetyConfig) -> None:
         self.config = config
+        self.environment_available_models = []
 
     async def initialize(self) -> None:
         pass
@@ -54,18 +55,18 @@ class SambaNovaSafetyAdapter(Safety, ShieldsProtocolPrivate, NeedsRequestProvide
 
     async def register_shield(self, shield: Shield) -> None:
         list_models_url = self.config.url + "/models"
-        try:
-            response = requests.get(list_models_url)
-            response.raise_for_status()
-        except requests.exceptions.RequestException as e:
-            raise RuntimeError(f"Request to {list_models_url} failed") from e
-        available_models = [model.get("id") for model in response.json().get("data", {})]
+        if len(self.environment_available_models) == 0:
+            try:
+                response = requests.get(list_models_url)
+                response.raise_for_status()
+            except requests.exceptions.RequestException as e:
+                raise RuntimeError(f"Request to {list_models_url} failed") from e
+            self.environment_available_models = [model.get("id") for model in response.json().get("data", {})]
         if (
-            len(available_models) == 0
-            or "guard" not in shield.provider_resource_id.lower()
-            or shield.provider_resource_id.split("sambanova/")[-1] not in available_models
+            "guard" not in shield.provider_resource_id.lower()
+            or shield.provider_resource_id.split("sambanova/")[-1] not in self.environment_available_models
         ):
-            raise ValueError(f"Shield {shield.provider_resource_id} not found in SambaNova")
+            logger.warning(f"Shield {shield.provider_resource_id} not available in {list_models_url}")
 
     async def run_shield(
         self, shield_id: str, messages: list[Message], params: dict[str, Any] | None = None