Move embedding deps to RAG tool where they are needed

distributions/dependencies.json

@@ -30,9 +30,7 @@
     "sentencepiece",
     "tqdm",
     "transformers",
-    "uvicorn",
+    "uvicorn"
-    "sentence-transformers --no-deps",
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu"
   ],
   "cerebras": [
     "aiosqlite",
@@ -170,9 +168,7 @@
     "sentencepiece",
     "tqdm",
     "transformers",
-    "uvicorn",
+    "uvicorn"
-    "sentence-transformers --no-deps",
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu"
   ],
   "hf-serverless": [
     "aiohttp",
@@ -247,9 +243,7 @@
     "tqdm",
     "transformers",
     "uvicorn",
-    "zmq",
+    "zmq"
-    "sentence-transformers --no-deps",
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu"
   ],
   "meta-reference-quantized-gpu": [
     "accelerate",
@@ -290,9 +284,7 @@
     "tqdm",
     "transformers",
     "uvicorn",
-    "zmq",
+    "zmq"
-    "sentence-transformers --no-deps",
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu"
   ],
   "nvidia": [
     "aiosqlite",
@@ -323,9 +315,7 @@
     "sentencepiece",
     "tqdm",
     "transformers",
-    "uvicorn",
+    "uvicorn"
-    "sentence-transformers --no-deps",
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu"
   ],
   "ollama": [
     "aiohttp",
@@ -335,7 +325,6 @@
     "chardet",
     "chromadb-client",
     "datasets",
-    "faiss-cpu",
     "fastapi",
     "fire",
     "httpx",
@@ -359,9 +348,7 @@
     "sqlite-vec",
     "tqdm",
     "transformers",
-    "uvicorn",
+    "uvicorn"
-    "sentence-transformers --no-deps",
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu"
   ],
   "remote-vllm": [
     "aiosqlite",
@@ -424,9 +411,7 @@
     "sentencepiece",
     "tqdm",
     "transformers",
-    "uvicorn",
+    "uvicorn"
-    "sentence-transformers --no-deps",
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu"
   ],
   "tgi": [
     "aiohttp",

docs/source/distributions/self_hosted_distro/cerebras.md

@@ -8,7 +8,7 @@ The `llamastack/distribution-cerebras` distribution consists of the following pr
 | agents | `inline::meta-reference` |
 | datasetio | `remote::huggingface`, `inline::localfs` |
 | eval | `inline::meta-reference` |
-| inference | `remote::cerebras` |
+| inference | `remote::cerebras`, `inline::sentence-transformers` |
 | safety | `inline::llama-guard` |
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
 | telemetry | `inline::meta-reference` |

docs/source/distributions/self_hosted_distro/dell.md

@@ -19,7 +19,7 @@ The `llamastack/distribution-dell` distribution consists of the following provid
 | agents | `inline::meta-reference` |
 | datasetio | `remote::huggingface`, `inline::localfs` |
 | eval | `inline::meta-reference` |
-| inference | `remote::tgi` |
+| inference | `remote::tgi`, `inline::sentence-transformers` |
 | safety | `inline::llama-guard` |
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
 | telemetry | `inline::meta-reference` |

docs/source/distributions/self_hosted_distro/fireworks.md

@@ -18,7 +18,7 @@ The `llamastack/distribution-fireworks` distribution consists of the following p
 | agents | `inline::meta-reference` |
 | datasetio | `remote::huggingface`, `inline::localfs` |
 | eval | `inline::meta-reference` |
-| inference | `remote::fireworks` |
+| inference | `remote::fireworks`, `inline::sentence-transformers` |
 | safety | `inline::llama-guard` |
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
 | telemetry | `inline::meta-reference` |

docs/source/distributions/self_hosted_distro/ollama.md

@@ -23,7 +23,7 @@ The `llamastack/distribution-ollama` distribution consists of the following prov
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
 | telemetry | `inline::meta-reference` |
 | tool_runtime | `remote::brave-search`, `remote::tavily-search`, `inline::code-interpreter`, `inline::rag-runtime` |
-| vector_io | `inline::faiss`, `inline::sqlite_vec`, `remote::chromadb`, `remote::pgvector` |
+| vector_io | `inline::sqlite-vec`, `remote::chromadb`, `remote::pgvector` |
 
 
 You should use this distribution if you have a regular desktop machine without very powerful GPUs. Of course, if you have powerful GPUs, you can still continue using this distribution since Ollama supports GPU acceleration.

docs/source/distributions/self_hosted_distro/remote-vllm.md

@@ -17,7 +17,7 @@ The `llamastack/distribution-remote-vllm` distribution consists of the following
 | agents | `inline::meta-reference` |
 | datasetio | `remote::huggingface`, `inline::localfs` |
 | eval | `inline::meta-reference` |
-| inference | `remote::vllm` |
+| inference | `remote::vllm`, `inline::sentence-transformers` |
 | safety | `inline::llama-guard` |
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
 | telemetry | `inline::meta-reference` |

docs/source/distributions/self_hosted_distro/tgi.md

@@ -19,7 +19,7 @@ The `llamastack/distribution-tgi` distribution consists of the following provide
 | agents | `inline::meta-reference` |
 | datasetio | `remote::huggingface`, `inline::localfs` |
 | eval | `inline::meta-reference` |
-| inference | `remote::tgi` |
+| inference | `remote::tgi`, `inline::sentence-transformers` |
 | safety | `inline::llama-guard` |
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
 | telemetry | `inline::meta-reference` |

docs/source/distributions/self_hosted_distro/together.md

@@ -18,7 +18,7 @@ The `llamastack/distribution-together` distribution consists of the following pr
 | agents | `inline::meta-reference` |
 | datasetio | `remote::huggingface`, `inline::localfs` |
 | eval | `inline::meta-reference` |
-| inference | `remote::together` |
+| inference | `remote::together`, `inline::sentence-transformers` |
 | safety | `inline::llama-guard` |
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
 | telemetry | `inline::meta-reference` |

llama_stack/providers/registry/inference.py

@@ -61,7 +61,10 @@ def available_providers() -> List[ProviderSpec]:
         InlineProviderSpec(
             api=Api.inference,
             provider_type="inline::sentence-transformers",
-            pip_packages=["sentence-transformers"],
+            pip_packages=[
+                "torch torchvision --index-url https://download.pytorch.org/whl/cpu",
+                "sentence-transformers --no-deps",
+            ],
             module="llama_stack.providers.inline.inference.sentence_transformers",
             config_class="llama_stack.providers.inline.inference.sentence_transformers.config.SentenceTransformersInferenceConfig",
         ),

llama_stack/providers/registry/tool_runtime.py

@@ -20,7 +20,18 @@ def available_providers() -> List[ProviderSpec]:
         InlineProviderSpec(
             api=Api.tool_runtime,
             provider_type="inline::rag-runtime",
-            pip_packages=[],
+            pip_packages=[
+                "blobfile",
+                "chardet",
+                "pypdf",
+                "tqdm",
+                "numpy",
+                "scikit-learn",
+                "scipy",
+                "nltk",
+                "sentencepiece",
+                "transformers",
+            ],
             module="llama_stack.providers.inline.tool_runtime.rag",
             config_class="llama_stack.providers.inline.tool_runtime.rag.config.RagToolRuntimeConfig",
             api_dependencies=[Api.vector_io, Api.inference],

llama_stack/providers/registry/vector_io.py

@@ -14,33 +14,13 @@ from llama_stack.providers.datatypes import (
     remote_provider_spec,
 )
 
-EMBEDDING_DEPS = [
-    "blobfile",
-    "chardet",
-    "pypdf",
-    "tqdm",
-    "numpy",
-    "scikit-learn",
-    "scipy",
-    "nltk",
-    "sentencepiece",
-    "transformers",
-    # this happens to work because special dependencies are always installed last
-    # so if there was a regular torch installed first, this would be ignored
-    # we need a better way to do this to identify potential conflicts, etc.
-    # for now, this lets us significantly reduce the size of the container which
-    # does not have any "local" inference code (and hence does not need GPU-enabled torch)
-    "torch torchvision --index-url https://download.pytorch.org/whl/cpu",
-    "sentence-transformers --no-deps",
-]
-
 
 def available_providers() -> List[ProviderSpec]:
     return [
         InlineProviderSpec(
             api=Api.vector_io,
             provider_type="inline::meta-reference",
-            pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
+            pip_packages=["faiss-cpu"],
             module="llama_stack.providers.inline.vector_io.faiss",
             config_class="llama_stack.providers.inline.vector_io.faiss.FaissVectorIOConfig",
             deprecation_warning="Please use the `inline::faiss` provider instead.",
@@ -49,15 +29,15 @@ def available_providers() -> List[ProviderSpec]:
         InlineProviderSpec(
             api=Api.vector_io,
             provider_type="inline::faiss",
-            pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
+            pip_packages=["faiss-cpu"],
             module="llama_stack.providers.inline.vector_io.faiss",
             config_class="llama_stack.providers.inline.vector_io.faiss.FaissVectorIOConfig",
             api_dependencies=[Api.inference],
         ),
         InlineProviderSpec(
             api=Api.vector_io,
-            provider_type="inline::sqlite_vec",
+            provider_type="inline::sqlite-vec",
-            pip_packages=EMBEDDING_DEPS + ["sqlite-vec"],
+            pip_packages=["sqlite-vec"],
             module="llama_stack.providers.inline.vector_io.sqlite_vec",
             config_class="llama_stack.providers.inline.vector_io.sqlite_vec.SQLiteVectorIOConfig",
             api_dependencies=[Api.inference],
@@ -66,7 +46,7 @@ def available_providers() -> List[ProviderSpec]:
             Api.vector_io,
             AdapterSpec(
                 adapter_type="chromadb",
-                pip_packages=EMBEDDING_DEPS + ["chromadb-client"],
+                pip_packages=["chromadb-client"],
                 module="llama_stack.providers.remote.vector_io.chroma",
                 config_class="llama_stack.providers.remote.vector_io.chroma.ChromaVectorIOConfig",
             ),
@@ -75,7 +55,7 @@ def available_providers() -> List[ProviderSpec]:
         InlineProviderSpec(
             api=Api.vector_io,
             provider_type="inline::chromadb",
-            pip_packages=EMBEDDING_DEPS + ["chromadb"],
+            pip_packages=["chromadb"],
             module="llama_stack.providers.inline.vector_io.chroma",
             config_class="llama_stack.providers.inline.vector_io.chroma.ChromaVectorIOConfig",
             api_dependencies=[Api.inference],
@@ -84,7 +64,7 @@ def available_providers() -> List[ProviderSpec]:
             Api.vector_io,
             AdapterSpec(
                 adapter_type="pgvector",
-                pip_packages=EMBEDDING_DEPS + ["psycopg2-binary"],
+                pip_packages=["psycopg2-binary"],
                 module="llama_stack.providers.remote.vector_io.pgvector",
                 config_class="llama_stack.providers.remote.vector_io.pgvector.PGVectorVectorIOConfig",
             ),
@@ -94,7 +74,7 @@ def available_providers() -> List[ProviderSpec]:
             Api.vector_io,
             AdapterSpec(
                 adapter_type="weaviate",
-                pip_packages=EMBEDDING_DEPS + ["weaviate-client"],
+                pip_packages=["weaviate-client"],
                 module="llama_stack.providers.remote.vector_io.weaviate",
                 config_class="llama_stack.providers.remote.vector_io.weaviate.WeaviateVectorIOConfig",
                 provider_data_validator="llama_stack.providers.remote.vector_io.weaviate.WeaviateRequestProviderData",
@@ -115,7 +95,7 @@ def available_providers() -> List[ProviderSpec]:
             Api.vector_io,
             AdapterSpec(
                 adapter_type="qdrant",
-                pip_packages=EMBEDDING_DEPS + ["qdrant-client"],
+                pip_packages=["qdrant-client"],
                 module="llama_stack.providers.remote.vector_io.qdrant",
                 config_class="llama_stack.providers.remote.vector_io.qdrant.QdrantVectorIOConfig",
             ),

llama_stack/providers/tests/vector_io/fixtures.py

@@ -61,7 +61,7 @@ def vector_io_sqlite_vec() -> ProviderFixture:
         providers=[
             Provider(
                 provider_id="sqlite_vec",
-                provider_type="inline::sqlite_vec",
+                provider_type="inline::sqlite-vec",
                 config=SQLiteVectorIOConfig(
                     kvstore=SqliteKVStoreConfig(db_path=temp_file.name).model_dump(),
                 ).model_dump(),

llama_stack/templates/cerebras/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - remote::cerebras
+    - inline::sentence-transformers
     safety:
     - inline::llama-guard
     vector_io:

llama_stack/templates/cerebras/cerebras.py

@@ -20,7 +20,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["remote::cerebras"],
+        "inference": ["remote::cerebras", "inline::sentence-transformers"],
         "safety": ["inline::llama-guard"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "agents": ["inline::meta-reference"],

llama_stack/templates/dell/build.yaml

@@ -5,6 +5,7 @@ distribution_spec:
   providers:
     inference:
     - remote::tgi
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/dell/dell.py

@@ -20,7 +20,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["remote::tgi"],
+        "inference": ["remote::tgi", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],

llama_stack/templates/fireworks/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - remote::fireworks
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/fireworks/fireworks.py

@@ -25,7 +25,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["remote::fireworks"],
+        "inference": ["remote::fireworks", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],

llama_stack/templates/hf-serverless/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - remote::hf::serverless
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/hf-serverless/hf_serverless.py

@@ -21,7 +21,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["remote::hf::serverless"],
+        "inference": ["remote::hf::serverless", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],

llama_stack/templates/ollama/build.yaml

@@ -5,8 +5,7 @@ distribution_spec:
     inference:
     - remote::ollama
     vector_io:
-    - inline::faiss
+    - inline::sqlite-vec
-    - inline::sqlite_vec
     - remote::chromadb
     - remote::pgvector
     safety:

llama_stack/templates/ollama/ollama.py

@@ -13,10 +13,6 @@ from llama_stack.distribution.datatypes import (
     ShieldInput,
     ToolGroupInput,
 )
-from llama_stack.providers.inline.inference.sentence_transformers import (
-    SentenceTransformersInferenceConfig,
-)
-from llama_stack.providers.inline.vector_io.faiss.config import FaissVectorIOConfig
 from llama_stack.providers.inline.vector_io.sqlite_vec.config import SQLiteVectorIOConfig
 from llama_stack.providers.remote.inference.ollama import OllamaImplConfig
 from llama_stack.templates.template import DistributionTemplate, RunConfigSettings
@@ -25,7 +21,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 def get_distribution_template() -> DistributionTemplate:
     providers = {
         "inference": ["remote::ollama"],
-        "vector_io": ["inline::faiss", "inline::sqlite_vec", "remote::chromadb", "remote::pgvector"],
+        "vector_io": ["inline::sqlite-vec", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],
         "telemetry": ["inline::meta-reference"],
@@ -45,19 +41,9 @@ def get_distribution_template() -> DistributionTemplate:
         provider_type="remote::ollama",
         config=OllamaImplConfig.sample_run_config(),
     )
-    embedding_provider = Provider(
-        provider_id="sentence-transformers",
-        provider_type="inline::sentence-transformers",
-        config=SentenceTransformersInferenceConfig.sample_run_config(),
-    )
-    vector_io_provider_faiss = Provider(
-        provider_id="faiss",
-        provider_type="inline::faiss",
-        config=FaissVectorIOConfig.sample_run_config(f"distributions/{name}"),
-    )
     vector_io_provider_sqlite = Provider(
-        provider_id="sqlite_vec",
+        provider_id="sqlite-vec",
-        provider_type="inline::sqlite_vec",
+        provider_type="inline::sqlite-vec",
         config=SQLiteVectorIOConfig.sample_run_config(f"distributions/{name}"),
     )
 
@@ -104,19 +90,16 @@ def get_distribution_template() -> DistributionTemplate:
         run_configs={
             "run.yaml": RunConfigSettings(
                 provider_overrides={
-                    "inference": [inference_provider, embedding_provider],
+                    "inference": [inference_provider],
-                    "vector_io": [vector_io_provider_faiss, vector_io_provider_sqlite],
+                    "vector_io": [vector_io_provider_sqlite],
                 },
-                default_models=[inference_model, embedding_model],
+                default_models=[inference_model],
                 default_tool_groups=default_tool_groups,
             ),
             "run-with-safety.yaml": RunConfigSettings(
                 provider_overrides={
-                    "inference": [
+                    "inference": [inference_provider],
-                        inference_provider,
+                    "vector_io": [vector_io_provider_sqlite],
-                        embedding_provider,
-                    ],
-                    "vector_io": [vector_io_provider_faiss, vector_io_provider_faiss],
                     "safety": [
                         Provider(
                             provider_id="llama-guard",

llama_stack/templates/ollama/run-with-safety.yaml

@@ -16,24 +16,11 @@ providers:
     provider_type: remote::ollama
     config:
       url: ${env.OLLAMA_URL:http://localhost:11434}
-  - provider_id: sentence-transformers
-    provider_type: inline::sentence-transformers
-    config: {}
   vector_io:
-  - provider_id: faiss
+  - provider_id: sqlite-vec
-    provider_type: inline::faiss
+    provider_type: inline::sqlite-vec
     config:
-      kvstore:
+      db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/sqlite_vec.db
-        type: sqlite
-        namespace: null
-        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/faiss_store.db
-  - provider_id: faiss
-    provider_type: inline::faiss
-    config:
-      kvstore:
-        type: sqlite
-        namespace: null
-        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/faiss_store.db
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard

llama_stack/templates/ollama/run.yaml

@@ -16,19 +16,9 @@ providers:
     provider_type: remote::ollama
     config:
       url: ${env.OLLAMA_URL:http://localhost:11434}
-  - provider_id: sentence-transformers
-    provider_type: inline::sentence-transformers
-    config: {}
   vector_io:
-  - provider_id: faiss
+  - provider_id: sqlite-vec
-    provider_type: inline::faiss
+    provider_type: inline::sqlite-vec
-    config:
-      kvstore:
-        type: sqlite
-        namespace: null
-        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/faiss_store.db
-  - provider_id: sqlite_vec
-    provider_type: inline::sqlite_vec
     config:
       db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/sqlite_vec.db
   safety:
@@ -97,12 +87,6 @@ models:
   model_id: ${env.INFERENCE_MODEL}
   provider_id: ollama
   model_type: llm
-- metadata:
-    embedding_dimension: 384
-  model_id: all-MiniLM-L6-v2
-  provider_id: ollama
-  provider_model_id: all-minilm:latest
-  model_type: embedding
 shields: []
 vector_dbs: []
 datasets: []

llama_stack/templates/remote-vllm/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - remote::vllm
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/remote-vllm/vllm.py

@@ -23,7 +23,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["remote::vllm"],
+        "inference": ["remote::vllm", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],

llama_stack/templates/tgi/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - remote::tgi
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/tgi/tgi.py

@@ -23,7 +23,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["remote::tgi"],
+        "inference": ["remote::tgi", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],

llama_stack/templates/together/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - remote::together
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/together/together.py

@@ -25,7 +25,7 @@ from llama_stack.templates.template import DistributionTemplate, RunConfigSettin
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["remote::together"],
+        "inference": ["remote::together", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],

llama_stack/templates/vllm-gpu/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - inline::vllm
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/vllm-gpu/vllm.py

@@ -20,7 +20,7 @@ from llama_stack.templates.template import (
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["inline::vllm"],
+        "inference": ["inline::vllm", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],