chore: turn OpenAIMixin into a pydantic.BaseModel (#3671)

# What does this PR do?

- implement get_api_key instead of relying on
LiteLLMOpenAIMixin.get_api_key
 - remove use of LiteLLMOpenAIMixin
 - add default initialize/shutdown methods to OpenAIMixin
 - remove __init__s to allow proper pydantic construction
- remove dead code from vllm adapter and associated / duplicate unit
tests
 - update vllm adapter to use openaimixin for model registration
 - remove ModelRegistryHelper from fireworks & together adapters
 - remove Inference from nvidia adapter
 - complete type hints on embedding_model_metadata
- allow extra fields on OpenAIMixin, for model_store, __provider_id__,
etc
 - new recordings for ollama
 - enhance the list models error handling
- update cerebras (remove cerebras-cloud-sdk) and anthropic (custom
model listing) inference adapters
 - parametrized test_inference_client_caching
- remove cerebras, databricks, fireworks, together from blanket mypy
exclude
 - removed unnecessary litellm deps

## Test Plan

ci

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

@@ -17,6 +17,6 @@ async def get_adapter_impl(config: FireworksImplConfig, _deps):
     from .fireworks import FireworksInferenceAdapter
 
     assert isinstance(config, FireworksImplConfig), f"Unexpected config type: {type(config)}"
-    impl = FireworksInferenceAdapter(config)
+    impl = FireworksInferenceAdapter(config=config)
     await impl.initialize()
     return impl

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

@@ -5,124 +5,26 @@
 # the root directory of this source tree.
 
 
-from fireworks.client import Fireworks
-
-from llama_stack.apis.inference import (
-    ChatCompletionRequest,
-    Inference,
-    LogProbConfig,
-    ResponseFormat,
-    ResponseFormatType,
-    SamplingParams,
-)
-from llama_stack.core.request_headers import NeedsRequestProviderData
 from llama_stack.log import get_logger
-from llama_stack.providers.utils.inference.model_registry import (
-    ModelRegistryHelper,
-)
-from llama_stack.providers.utils.inference.openai_compat import (
-    convert_message_to_openai_dict,
-    get_sampling_options,
-)
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
-from llama_stack.providers.utils.inference.prompt_adapter import (
-    chat_completion_request_to_prompt,
-    request_has_media,
-)
 
 from .config import FireworksImplConfig
 
 logger = get_logger(name=__name__, category="inference::fireworks")
 
 
-class FireworksInferenceAdapter(OpenAIMixin, Inference, NeedsRequestProviderData):
-    embedding_model_metadata = {
+class FireworksInferenceAdapter(OpenAIMixin):
+    config: FireworksImplConfig
+
+    embedding_model_metadata: dict[str, dict[str, int]] = {
         "nomic-ai/nomic-embed-text-v1.5": {"embedding_dimension": 768, "context_length": 8192},
         "accounts/fireworks/models/qwen3-embedding-8b": {"embedding_dimension": 4096, "context_length": 40960},
     }
 
-    def __init__(self, config: FireworksImplConfig) -> None:
-        ModelRegistryHelper.__init__(self)
-        self.config = config
-        self.allowed_models = config.allowed_models
-
-    async def initialize(self) -> None:
-        pass
-
-    async def shutdown(self) -> None:
-        pass
+    provider_data_api_key_field: str = "fireworks_api_key"
 
     def get_api_key(self) -> str:
-        config_api_key = self.config.api_key.get_secret_value() if self.config.api_key else None
-        if config_api_key:
-            return config_api_key
-        else:
-            provider_data = self.get_request_provider_data()
-            if provider_data is None or not provider_data.fireworks_api_key:
-                raise ValueError(
-                    'Pass Fireworks API Key in the header X-LlamaStack-Provider-Data as { "fireworks_api_key": <your api key>}'
-                )
-            return provider_data.fireworks_api_key
+        return self.config.api_key.get_secret_value() if self.config.api_key else None  # type: ignore[return-value]
 
     def get_base_url(self) -> str:
         return "https://api.fireworks.ai/inference/v1"
-
-    def _get_client(self) -> Fireworks:
-        fireworks_api_key = self.get_api_key()
-        return Fireworks(api_key=fireworks_api_key)
-
-    def _build_options(
-        self,
-        sampling_params: SamplingParams | None,
-        fmt: ResponseFormat | None,
-        logprobs: LogProbConfig | None,
-    ) -> dict:
-        options = get_sampling_options(sampling_params)
-        options.setdefault("max_tokens", 512)
-
-        if fmt:
-            if fmt.type == ResponseFormatType.json_schema.value:
-                options["response_format"] = {
-                    "type": "json_object",
-                    "schema": fmt.json_schema,
-                }
-            elif fmt.type == ResponseFormatType.grammar.value:
-                options["response_format"] = {
-                    "type": "grammar",
-                    "grammar": fmt.bnf,
-                }
-            else:
-                raise ValueError(f"Unknown response format {fmt.type}")
-
-        if logprobs and logprobs.top_k:
-            options["logprobs"] = logprobs.top_k
-            if options["logprobs"] <= 0 or options["logprobs"] >= 5:
-                raise ValueError("Required range: 0 < top_k < 5")
-
-        return options
-
-    async def _get_params(self, request: ChatCompletionRequest) -> dict:
-        input_dict = {}
-        media_present = request_has_media(request)
-
-        llama_model = self.get_llama_model(request.model)
-        # TODO: tools are never added to the request, so we need to add them here
-        if media_present or not llama_model:
-            input_dict["messages"] = [await convert_message_to_openai_dict(m, download=True) for m in request.messages]
-        else:
-            input_dict["prompt"] = await chat_completion_request_to_prompt(request, llama_model)
-
-        # Fireworks always prepends with BOS
-        if "prompt" in input_dict:
-            if input_dict["prompt"].startswith("<|begin_of_text|>"):
-                input_dict["prompt"] = input_dict["prompt"][len("<|begin_of_text|>") :]
-
-        params = {
-            "model": request.model,
-            **input_dict,
-            "stream": bool(request.stream),
-            **self._build_options(request.sampling_params, request.response_format, request.logprobs),
-        }
-        logger.debug(f"params to fireworks: {params}")
-
-        return params