feat: create dynamic model registration for Anthropic remote inference provider

llama_stack/providers/remote/inference/anthropic/anthropic.py

@@ -4,11 +4,17 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+import logging
+
+from anthropic import AsyncAnthropic, NotFoundError
+
 from llama_stack.providers.utils.inference.litellm_openai_mixin import LiteLLMOpenAIMixin
 
 from .config import AnthropicConfig
 from .models import MODEL_ENTRIES
 
+logger = logging.getLogger(__name__)
+
 
 class AnthropicInferenceAdapter(LiteLLMOpenAIMixin):
     def __init__(self, config: AnthropicConfig) -> None:
@@ -19,9 +25,35 @@ class AnthropicInferenceAdapter(LiteLLMOpenAIMixin):
             provider_data_api_key_field="anthropic_api_key",
         )
         self.config = config
+        self._client: AsyncAnthropic | None = None
 
     async def initialize(self) -> None:
         await super().initialize()
 
     async def shutdown(self) -> None:
+        # Clean up the client connection pool
+        if self._client:
+            await self._client.aclose()
+            self._client = None
         await super().shutdown()
+
+    @property
+    def client(self) -> AsyncAnthropic:
+        if self._client is None:
+            api_key = self.config.api_key if self.config.api_key else "no-key"
+            self._client = AsyncAnthropic(api_key=api_key)
+        return self._client
+
+    async def check_model_availability(self, model: str) -> bool:
+        try:
+            retrieved_model = await self.client.models.retrieve(model)
+            logger.info(f"Model {retrieved_model.id} is available on Anthropic")
+            return True
+
+        except NotFoundError:
+            logger.info(f"Model {model} was not found on Anthropic")
+
+        except Exception as e:
+            logger.error(f"Failed to check model availability for {model} on Anthropic: {e}")
+
+        return False

pyproject.toml

@@ -22,6 +22,7 @@ classifiers = [
 ]
 dependencies = [
     "aiohttp",
+    "anthropic>=0.58.2",
     "fastapi>=0.115.0,<1.0",                  # server
     "fire",                                   # for MCP in LLS client
     "httpx",

requirements.txt

@@ -10,8 +10,11 @@ aiosqlite==0.21.0
     # via llama-stack
 annotated-types==0.7.0
     # via pydantic
+anthropic==0.58.2
+    # via llama-stack
 anyio==4.8.0
     # via
+    #   anthropic
     #   httpx
     #   llama-api-client
     #   llama-stack-client
@@ -50,6 +53,7 @@ deprecated==1.2.18
     #   opentelemetry-semantic-conventions
 distro==1.9.0
     # via
+    #   anthropic
     #   llama-api-client
     #   llama-stack-client
     #   openai
@@ -82,6 +86,7 @@ httpcore==1.0.9
     # via httpx
 httpx==0.28.1
     # via
+    #   anthropic
     #   llama-api-client
     #   llama-stack
     #   llama-stack-client
@@ -99,7 +104,9 @@ importlib-metadata==8.5.0
 jinja2==3.1.6
     # via llama-stack
 jiter==0.8.2
-    # via openai
+    # via
+    #   anthropic
+    #   openai
 jsonschema==4.23.0
     # via llama-stack
 jsonschema-specifications==2024.10.1
@@ -169,6 +176,7 @@ pycparser==2.22 ; platform_python_implementation != 'PyPy'
     # via cffi
 pydantic==2.10.6
     # via
+    #   anthropic
     #   fastapi
     #   llama-api-client
     #   llama-stack
@@ -220,6 +228,7 @@ six==1.17.0
     #   python-dateutil
 sniffio==1.3.1
     # via
+    #   anthropic
     #   anyio
     #   llama-api-client
     #   llama-stack-client
@@ -243,6 +252,7 @@ tqdm==4.67.1
 typing-extensions==4.12.2
     # via
     #   aiosqlite
+    #   anthropic
     #   anyio
     #   fastapi
     #   huggingface-hub