feat: add static embedding metadata to dynamic model listings for providers using OpenAIMixin (#3547)

# What does this PR do?

- remove auto-download of ollama embedding models
- add embedding model metadata to dynamic listing w/ unit test
- add support and tests for allowed_models
- removed inference provider models.py files where dynamic listing is
enabled
- store embedding metadata in embedding_model_metadata field on
inference providers
- make model_entries optional on ModelRegistryHelper and
LiteLLMOpenAIMixin
- make OpenAIMixin a ModelRegistryHelper
- skip base64 embedding test for remote::ollama, always returns floats
- only use OpenAI client for ollama model listing
- remove unused build_model_entry function
- remove unused get_huggingface_repo function


## Test Plan

ci w/ new tests

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

@@ -1,103 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-#
-# This source code is licensed under the terms described in the LICENSE file in
-# the root directory of this source tree.
-
-from llama_stack.models.llama.sku_types import CoreModelId
-from llama_stack.providers.utils.inference.model_registry import (
-    ProviderModelEntry,
-    build_hf_repo_model_entry,
-)
-
-SAFETY_MODELS_ENTRIES = [
-    build_hf_repo_model_entry(
-        "meta-llama/Llama-Guard-3-8B",
-        CoreModelId.llama_guard_3_8b.value,
-    ),
-    build_hf_repo_model_entry(
-        "meta-llama/Llama-Guard-3-11B-Vision-Turbo",
-        CoreModelId.llama_guard_3_11b_vision.value,
-    ),
-]
-
-# source: https://docs.together.ai/docs/serverless-models#embedding-models
-EMBEDDING_MODEL_ENTRIES = {
-    "togethercomputer/m2-bert-80M-32k-retrieval": ProviderModelEntry(
-        provider_model_id="togethercomputer/m2-bert-80M-32k-retrieval",
-        metadata={
-            "embedding_dimension": 768,
-            "context_length": 32768,
-        },
-    ),
-    "BAAI/bge-large-en-v1.5": ProviderModelEntry(
-        provider_model_id="BAAI/bge-large-en-v1.5",
-        metadata={
-            "embedding_dimension": 1024,
-            "context_length": 512,
-        },
-    ),
-    "BAAI/bge-base-en-v1.5": ProviderModelEntry(
-        provider_model_id="BAAI/bge-base-en-v1.5",
-        metadata={
-            "embedding_dimension": 768,
-            "context_length": 512,
-        },
-    ),
-    "Alibaba-NLP/gte-modernbert-base": ProviderModelEntry(
-        provider_model_id="Alibaba-NLP/gte-modernbert-base",
-        metadata={
-            "embedding_dimension": 768,
-            "context_length": 8192,
-        },
-    ),
-    "intfloat/multilingual-e5-large-instruct": ProviderModelEntry(
-        provider_model_id="intfloat/multilingual-e5-large-instruct",
-        metadata={
-            "embedding_dimension": 1024,
-            "context_length": 512,
-        },
-    ),
-}
-MODEL_ENTRIES = (
-    [
-        build_hf_repo_model_entry(
-            "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo",
-            CoreModelId.llama3_1_8b_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo",
-            CoreModelId.llama3_1_70b_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo",
-            CoreModelId.llama3_1_405b_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Llama-3.2-3B-Instruct-Turbo",
-            CoreModelId.llama3_2_3b_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Llama-3.2-11B-Vision-Instruct-Turbo",
-            CoreModelId.llama3_2_11b_vision_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Llama-3.2-90B-Vision-Instruct-Turbo",
-            CoreModelId.llama3_2_90b_vision_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Llama-3.3-70B-Instruct-Turbo",
-            CoreModelId.llama3_3_70b_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Llama-4-Scout-17B-16E-Instruct",
-            CoreModelId.llama4_scout_17b_16e_instruct.value,
-        ),
-        build_hf_repo_model_entry(
-            "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8",
-            CoreModelId.llama4_maverick_17b_128e_instruct.value,
-        ),
-    ]
-    + SAFETY_MODELS_ENTRIES
-    + list(EMBEDDING_MODEL_ENTRIES.values())
-)

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

@@ -56,15 +56,22 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 )
 
 from .config import TogetherImplConfig
-from .models import EMBEDDING_MODEL_ENTRIES, MODEL_ENTRIES
 
 logger = get_logger(name=__name__, category="inference::together")
 
 
 class TogetherInferenceAdapter(OpenAIMixin, ModelRegistryHelper, Inference, NeedsRequestProviderData):
+    embedding_model_metadata = {
+        "togethercomputer/m2-bert-80M-32k-retrieval": {"embedding_dimension": 768, "context_length": 32768},
+        "BAAI/bge-large-en-v1.5": {"embedding_dimension": 1024, "context_length": 512},
+        "BAAI/bge-base-en-v1.5": {"embedding_dimension": 768, "context_length": 512},
+        "Alibaba-NLP/gte-modernbert-base": {"embedding_dimension": 768, "context_length": 8192},
+        "intfloat/multilingual-e5-large-instruct": {"embedding_dimension": 1024, "context_length": 512},
+    }
+
     def __init__(self, config: TogetherImplConfig) -> None:
-        ModelRegistryHelper.__init__(self, MODEL_ENTRIES, config.allowed_models)
         self.config = config
+        self.allowed_models = config.allowed_models
         self._model_cache: dict[str, Model] = {}
 
     def get_api_key(self):
@@ -264,15 +271,16 @@ class TogetherInferenceAdapter(OpenAIMixin, ModelRegistryHelper, Inference, Need
         # Together's /v1/models is not compatible with OpenAI's /v1/models. Together support ticket #13355 -> will not fix, use Together's own client
         for m in await self._get_client().models.list():
             if m.type == "embedding":
-                if m.id not in EMBEDDING_MODEL_ENTRIES:
+                if m.id not in self.embedding_model_metadata:
                     logger.warning(f"Unknown embedding dimension for model {m.id}, skipping.")
                     continue
+                metadata = self.embedding_model_metadata[m.id]
                 self._model_cache[m.id] = Model(
                     provider_id=self.__provider_id__,
-                    provider_resource_id=EMBEDDING_MODEL_ENTRIES[m.id].provider_model_id,
+                    provider_resource_id=m.id,
                     identifier=m.id,
                     model_type=ModelType.embedding,
-                    metadata=EMBEDDING_MODEL_ENTRIES[m.id].metadata,
+                    metadata=metadata,
                 )
             else:
                 self._model_cache[m.id] = Model(