llama-stack 0.4.4__py3-none-any.whl → 0.5.0rc1__py3-none-any.whl

llama_stack/core/store/registry.py

@@ -190,7 +190,7 @@ class CachedDiskDistributionRegistry(DiskDistributionRegistry):
 
 
 async def create_dist_registry(
-    metadata_store: KVStoreReference, image_name: str
+    metadata_store: KVStoreReference, distro_name: str
 ) -> tuple[CachedDiskDistributionRegistry, KVStore]:
     # instantiate kvstore for storing and retrieving distribution metadata
     dist_kvstore = await kvstore_impl(metadata_store)

llama_stack/core/utils/exec.py

@@ -17,10 +17,10 @@ from llama_stack.log import get_logger
 log = get_logger(name=__name__, category="core")
 
 
-def formulate_run_args(image_type: str, image_name: str) -> list:
+def formulate_run_args(image_type: str, distro_name: str) -> list:
     # Only venv is supported now
     current_venv = os.environ.get("VIRTUAL_ENV")
-    env_name = image_name or current_venv
+    env_name = distro_name or current_venv
     if not env_name:
         cprint(
             "No current virtual environment detected, please specify a virtual environment name with --image-name",

llama_stack/core/utils/type_inspection.py

@@ -36,10 +36,24 @@ def is_unwrapped_body_param(param_type: Any) -> bool:
         base_type = args[0]
         metadata = args[1:]
 
-        # Look for Body annotation with embed=False
+        # Look for Body annotation; treat embed=None (default) as unwrapped
         # Body() returns a FieldInfo object, so we check for that type and the embed attribute
         for item in metadata:
-            if isinstance(item, FieldInfo) and hasattr(item, "embed") and not item.embed:
+            if isinstance(item, FieldInfo) and hasattr(item, "embed") and item.embed is not True:
                 return inspect.isclass(base_type) and issubclass(base_type, BaseModel)
 
     return False
+
+
+def is_body_param(param_type: Any) -> bool:
+    """
+    Check if a parameter type represents a body parameter (Annotated with Body()).
+    """
+    if get_origin(param_type) is typing.Annotated:
+        args = get_args(param_type)
+        base_type = args[0]
+        metadata = args[1:]
+        for item in metadata:
+            if isinstance(item, FieldInfo):
+                return inspect.isclass(base_type) and issubclass(base_type, BaseModel)
+    return False

llama_stack/distributions/dell/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: dell
+distro_name: dell
 apis:
 - agents
 - datasetio
@@ -108,6 +108,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/dell/run-with-safety.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: dell
+distro_name: dell
 apis:
 - agents
 - datasetio
@@ -112,6 +112,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/nvidia/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: nvidia
+distro_name: nvidia
 apis:
 - agents
 - datasetio
@@ -102,6 +102,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models: []
   shields: []

llama_stack/distributions/nvidia/run-with-safety.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: nvidia
+distro_name: nvidia
 apis:
 - agents
 - datasetio
@@ -113,6 +113,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/oci/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: oci
+distro_name: oci
 apis:
 - agents
 - datasetio
@@ -120,6 +120,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models: []
   shields: []

llama_stack/distributions/open-benchmark/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: open-benchmark
+distro_name: open-benchmark
 apis:
 - agents
 - datasetio
@@ -57,6 +57,11 @@ providers:
       db: ${env.PGVECTOR_DB:=}
       user: ${env.PGVECTOR_USER:=}
       password: ${env.PGVECTOR_PASSWORD:=}
+      distance_metric: COSINE
+      vector_index:
+        type: HNSW
+        m: 16
+        ef_construction: 64
       persistence:
         namespace: vector_io::pgvector
         backend: kv_default
@@ -145,6 +150,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/postgres-demo/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: postgres-demo
+distro_name: postgres-demo
 apis:
 - agents
 - inference

llama_stack/distributions/starter/build.yaml

@@ -0,0 +1,62 @@
+version: 2
+distribution_spec:
+  description: Quick start template for running Llama Stack with several popular providers.
+    This distribution is intended for CPU-only environments.
+  providers:
+    inference:
+    - provider_type: remote::cerebras
+    - provider_type: remote::ollama
+    - provider_type: remote::vllm
+    - provider_type: remote::tgi
+    - provider_type: remote::fireworks
+    - provider_type: remote::together
+    - provider_type: remote::bedrock
+    - provider_type: remote::nvidia
+    - provider_type: remote::openai
+    - provider_type: remote::anthropic
+    - provider_type: remote::gemini
+    - provider_type: remote::vertexai
+    - provider_type: remote::groq
+    - provider_type: remote::sambanova
+    - provider_type: remote::azure
+    - provider_type: inline::sentence-transformers
+    vector_io:
+    - provider_type: inline::faiss
+    - provider_type: inline::sqlite-vec
+    - provider_type: inline::milvus
+    - provider_type: remote::chromadb
+    - provider_type: remote::pgvector
+    - provider_type: remote::qdrant
+    - provider_type: remote::weaviate
+    - provider_type: remote::elasticsearch
+    files:
+    - provider_type: inline::localfs
+    safety:
+    - provider_type: inline::llama-guard
+    - provider_type: inline::code-scanner
+    agents:
+    - provider_type: inline::meta-reference
+    post_training:
+    - provider_type: inline::torchtune-cpu
+    eval:
+    - provider_type: inline::meta-reference
+    datasetio:
+    - provider_type: remote::huggingface
+    - provider_type: inline::localfs
+    scoring:
+    - provider_type: inline::basic
+    - provider_type: inline::llm-as-judge
+    - provider_type: inline::braintrust
+    tool_runtime:
+    - provider_type: remote::brave-search
+    - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
+    - provider_type: remote::model-context-protocol
+    batches:
+    - provider_type: inline::reference
+image_type: venv
+additional_pip_packages:
+- aiosqlite
+- asyncpg
+- psycopg2-binary
+- sqlalchemy[asyncio]

llama_stack/distributions/starter/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: starter
+distro_name: starter
 apis:
 - agents
 - batches
@@ -29,7 +29,9 @@ providers:
       base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
-      tls_verify: ${env.VLLM_TLS_VERIFY:=true}
+      network:
+        tls:
+          verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
@@ -71,7 +73,7 @@ providers:
     provider_type: remote::vertexai
     config:
       project: ${env.VERTEX_AI_PROJECT:=}
-      location: ${env.VERTEX_AI_LOCATION:=us-central1}
+      location: ${env.VERTEX_AI_LOCATION:=global}
   - provider_id: groq
     provider_type: remote::groq
     config:
@@ -127,6 +129,11 @@ providers:
       db: ${env.PGVECTOR_DB:=}
       user: ${env.PGVECTOR_USER:=}
       password: ${env.PGVECTOR_PASSWORD:=}
+      distance_metric: COSINE
+      vector_index:
+        type: HNSW
+        m: 16
+        ef_construction: 64
       persistence:
         namespace: vector_io::pgvector
         backend: kv_default
@@ -145,6 +152,14 @@ providers:
       persistence:
         namespace: vector_io::weaviate
         backend: kv_default
+  - provider_id: ${env.ELASTICSEARCH_URL:+elasticsearch}
+    provider_type: remote::elasticsearch
+    config:
+      elasticsearch_url: ${env.ELASTICSEARCH_URL:=localhost:9200}
+      elasticsearch_api_key: ${env.ELASTICSEARCH_API_KEY:=}
+      persistence:
+        namespace: vector_io::elasticsearch
+        backend: kv_default
   files:
   - provider_id: meta-reference-files
     provider_type: inline::localfs
@@ -252,6 +267,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models: []
   shields:
@@ -325,3 +343,4 @@ vector_stores:
     cleanup_interval_seconds: 86400
 safety:
   default_shield_id: llama-guard
+connectors: []

llama_stack/distributions/starter/run-with-postgres-store.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: starter
+distro_name: starter
 apis:
 - agents
 - batches
@@ -29,7 +29,9 @@ providers:
       base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
-      tls_verify: ${env.VLLM_TLS_VERIFY:=true}
+      network:
+        tls:
+          verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
@@ -71,7 +73,7 @@ providers:
     provider_type: remote::vertexai
     config:
       project: ${env.VERTEX_AI_PROJECT:=}
-      location: ${env.VERTEX_AI_LOCATION:=us-central1}
+      location: ${env.VERTEX_AI_LOCATION:=global}
   - provider_id: groq
     provider_type: remote::groq
     config:
@@ -127,6 +129,11 @@ providers:
       db: ${env.PGVECTOR_DB:=}
       user: ${env.PGVECTOR_USER:=}
       password: ${env.PGVECTOR_PASSWORD:=}
+      distance_metric: COSINE
+      vector_index:
+        type: HNSW
+        m: 16
+        ef_construction: 64
       persistence:
         namespace: vector_io::pgvector
         backend: kv_default
@@ -145,6 +152,14 @@ providers:
       persistence:
         namespace: vector_io::weaviate
         backend: kv_default
+  - provider_id: ${env.ELASTICSEARCH_URL:+elasticsearch}
+    provider_type: remote::elasticsearch
+    config:
+      elasticsearch_url: ${env.ELASTICSEARCH_URL:=localhost:9200}
+      elasticsearch_api_key: ${env.ELASTICSEARCH_API_KEY:=}
+      persistence:
+        namespace: vector_io::elasticsearch
+        backend: kv_default
   files:
   - provider_id: meta-reference-files
     provider_type: inline::localfs
@@ -261,6 +276,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models: []
   shields:
@@ -334,3 +352,4 @@ vector_stores:
     cleanup_interval_seconds: 86400
 safety:
   default_shield_id: llama-guard
+connectors: []

llama_stack/distributions/starter/starter.py

@@ -32,6 +32,7 @@ from llama_stack.providers.inline.vector_io.sqlite_vec.config import (
 )
 from llama_stack.providers.registry.inference import available_providers
 from llama_stack.providers.remote.vector_io.chroma.config import ChromaVectorIOConfig
+from llama_stack.providers.remote.vector_io.elasticsearch.config import ElasticsearchVectorIOConfig
 from llama_stack.providers.remote.vector_io.pgvector.config import (
     PGVectorVectorIOConfig,
 )
@@ -121,6 +122,7 @@ def get_distribution_template(name: str = "starter") -> DistributionTemplate:
             BuildProvider(provider_type="remote::pgvector"),
             BuildProvider(provider_type="remote::qdrant"),
             BuildProvider(provider_type="remote::weaviate"),
+            BuildProvider(provider_type="remote::elasticsearch"),
         ],
         "files": [BuildProvider(provider_type="inline::localfs")],
         "safety": [
@@ -237,6 +239,15 @@ def get_distribution_template(name: str = "starter") -> DistributionTemplate:
                     cluster_url="${env.WEAVIATE_CLUSTER_URL:=}",
                 ),
             ),
+            Provider(
+                provider_id="${env.ELASTICSEARCH_URL:+elasticsearch}",
+                provider_type="remote::elasticsearch",
+                config=ElasticsearchVectorIOConfig.sample_run_config(
+                    f"~/.llama/distributions/{name}",
+                    elasticsearch_url="${env.ELASTICSEARCH_URL:=localhost:9200}",
+                    elasticsearch_api_key="${env.ELASTICSEARCH_API_KEY:=}",
+                ),
+            ),
         ],
         "files": [files_provider],
     }
@@ -246,6 +257,7 @@ def get_distribution_template(name: str = "starter") -> DistributionTemplate:
         default_models=[],
         default_tool_groups=default_tool_groups,
         default_shields=default_shields,
+        default_connectors=[],
         vector_stores_config=VectorStoresConfig(
             default_provider_id="faiss",
             default_embedding_model=QualifiedModel(
@@ -309,7 +321,7 @@ def get_distribution_template(name: str = "starter") -> DistributionTemplate:
                 "Google Cloud Project ID for Vertex AI",
             ),
             "VERTEX_AI_LOCATION": (
-                "us-central1",
+                "global",
                 "Google Cloud Location for Vertex AI",
             ),
             "SAMBANOVA_API_KEY": (

llama_stack/distributions/starter-gpu/build.yaml

@@ -0,0 +1,62 @@
+version: 2
+distribution_spec:
+  description: Quick start template for running Llama Stack with several popular providers.
+    This distribution is intended for GPU-enabled environments.
+  providers:
+    inference:
+    - provider_type: remote::cerebras
+    - provider_type: remote::ollama
+    - provider_type: remote::vllm
+    - provider_type: remote::tgi
+    - provider_type: remote::fireworks
+    - provider_type: remote::together
+    - provider_type: remote::bedrock
+    - provider_type: remote::nvidia
+    - provider_type: remote::openai
+    - provider_type: remote::anthropic
+    - provider_type: remote::gemini
+    - provider_type: remote::vertexai
+    - provider_type: remote::groq
+    - provider_type: remote::sambanova
+    - provider_type: remote::azure
+    - provider_type: inline::sentence-transformers
+    vector_io:
+    - provider_type: inline::faiss
+    - provider_type: inline::sqlite-vec
+    - provider_type: inline::milvus
+    - provider_type: remote::chromadb
+    - provider_type: remote::pgvector
+    - provider_type: remote::qdrant
+    - provider_type: remote::weaviate
+    - provider_type: remote::elasticsearch
+    files:
+    - provider_type: inline::localfs
+    safety:
+    - provider_type: inline::llama-guard
+    - provider_type: inline::code-scanner
+    agents:
+    - provider_type: inline::meta-reference
+    post_training:
+    - provider_type: inline::huggingface-gpu
+    eval:
+    - provider_type: inline::meta-reference
+    datasetio:
+    - provider_type: remote::huggingface
+    - provider_type: inline::localfs
+    scoring:
+    - provider_type: inline::basic
+    - provider_type: inline::llm-as-judge
+    - provider_type: inline::braintrust
+    tool_runtime:
+    - provider_type: remote::brave-search
+    - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
+    - provider_type: remote::model-context-protocol
+    batches:
+    - provider_type: inline::reference
+image_type: venv
+additional_pip_packages:
+- aiosqlite
+- asyncpg
+- psycopg2-binary
+- sqlalchemy[asyncio]

llama_stack/distributions/starter-gpu/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: starter-gpu
+distro_name: starter-gpu
 apis:
 - agents
 - batches
@@ -29,7 +29,9 @@ providers:
       base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
-      tls_verify: ${env.VLLM_TLS_VERIFY:=true}
+      network:
+        tls:
+          verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
@@ -71,7 +73,7 @@ providers:
     provider_type: remote::vertexai
     config:
       project: ${env.VERTEX_AI_PROJECT:=}
-      location: ${env.VERTEX_AI_LOCATION:=us-central1}
+      location: ${env.VERTEX_AI_LOCATION:=global}
   - provider_id: groq
     provider_type: remote::groq
     config:
@@ -127,6 +129,11 @@ providers:
       db: ${env.PGVECTOR_DB:=}
       user: ${env.PGVECTOR_USER:=}
       password: ${env.PGVECTOR_PASSWORD:=}
+      distance_metric: COSINE
+      vector_index:
+        type: HNSW
+        m: 16
+        ef_construction: 64
       persistence:
         namespace: vector_io::pgvector
         backend: kv_default
@@ -145,6 +152,14 @@ providers:
       persistence:
         namespace: vector_io::weaviate
         backend: kv_default
+  - provider_id: ${env.ELASTICSEARCH_URL:+elasticsearch}
+    provider_type: remote::elasticsearch
+    config:
+      elasticsearch_url: ${env.ELASTICSEARCH_URL:=localhost:9200}
+      elasticsearch_api_key: ${env.ELASTICSEARCH_API_KEY:=}
+      persistence:
+        namespace: vector_io::elasticsearch
+        backend: kv_default
   files:
   - provider_id: meta-reference-files
     provider_type: inline::localfs
@@ -255,6 +270,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models: []
   shields:
@@ -328,3 +346,4 @@ vector_stores:
     cleanup_interval_seconds: 86400
 safety:
   default_shield_id: llama-guard
+connectors: []

llama_stack/distributions/starter-gpu/run-with-postgres-store.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: starter-gpu
+distro_name: starter-gpu
 apis:
 - agents
 - batches
@@ -29,7 +29,9 @@ providers:
       base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
-      tls_verify: ${env.VLLM_TLS_VERIFY:=true}
+      network:
+        tls:
+          verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
@@ -71,7 +73,7 @@ providers:
     provider_type: remote::vertexai
     config:
       project: ${env.VERTEX_AI_PROJECT:=}
-      location: ${env.VERTEX_AI_LOCATION:=us-central1}
+      location: ${env.VERTEX_AI_LOCATION:=global}
   - provider_id: groq
     provider_type: remote::groq
     config:
@@ -127,6 +129,11 @@ providers:
       db: ${env.PGVECTOR_DB:=}
       user: ${env.PGVECTOR_USER:=}
       password: ${env.PGVECTOR_PASSWORD:=}
+      distance_metric: COSINE
+      vector_index:
+        type: HNSW
+        m: 16
+        ef_construction: 64
       persistence:
         namespace: vector_io::pgvector
         backend: kv_default
@@ -145,6 +152,14 @@ providers:
       persistence:
         namespace: vector_io::weaviate
         backend: kv_default
+  - provider_id: ${env.ELASTICSEARCH_URL:+elasticsearch}
+    provider_type: remote::elasticsearch
+    config:
+      elasticsearch_url: ${env.ELASTICSEARCH_URL:=localhost:9200}
+      elasticsearch_api_key: ${env.ELASTICSEARCH_API_KEY:=}
+      persistence:
+        namespace: vector_io::elasticsearch
+        backend: kv_default
   files:
   - provider_id: meta-reference-files
     provider_type: inline::localfs
@@ -264,6 +279,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models: []
   shields:
@@ -337,3 +355,4 @@ vector_stores:
     cleanup_interval_seconds: 86400
 safety:
   default_shield_id: llama-guard
+connectors: []

llama_stack/distributions/template.py

@@ -36,7 +36,7 @@ from llama_stack.core.storage.kvstore.config import SqliteKVStoreConfig
 from llama_stack.core.storage.sqlstore.sqlstore import SqliteSqlStoreConfig
 from llama_stack.core.utils.dynamic import instantiate_class_type
 from llama_stack.providers.utils.inference.model_registry import ProviderModelEntry
-from llama_stack_api import DatasetPurpose, ModelType
+from llama_stack_api import ConnectorInput, DatasetPurpose, ModelType
 
 
 def filter_empty_values(obj: Any) -> Any:
@@ -181,6 +181,7 @@ class RunConfigSettings(BaseModel):
     default_tool_groups: list[ToolGroupInput] | None = None
     default_datasets: list[DatasetInput] | None = None
     default_benchmarks: list[BenchmarkInput] | None = None
+    default_connectors: list[ConnectorInput] | None = None
     vector_stores_config: VectorStoresConfig | None = None
     safety_config: SafetyConfig | None = None
     storage_backends: dict[str, Any] | None = None
@@ -255,6 +256,10 @@ class RunConfigSettings(BaseModel):
                 backend="kv_default",
                 namespace="prompts",
             ).model_dump(exclude_none=True),
+            "connectors": KVStoreReference(
+                backend="kv_default",
+                namespace="connectors",
+            ).model_dump(exclude_none=True),
         }
 
         storage_config = dict(
@@ -265,7 +270,7 @@ class RunConfigSettings(BaseModel):
         # Return a dict that matches StackRunConfig structure
         config = {
             "version": LLAMA_STACK_RUN_CONFIG_VERSION,
-            "image_name": name,
+            "distro_name": name,
             "container_image": container_image,
             "apis": apis,
             "providers": provider_configs,
@@ -290,6 +295,9 @@ class RunConfigSettings(BaseModel):
         if self.safety_config:
             config["safety"] = self.safety_config.model_dump(exclude_none=True)
 
+        if self.default_connectors is not None:
+            config["connectors"] = [c.model_dump(exclude_none=True) for c in self.default_connectors]
+
         return config
 
 

llama_stack/distributions/watsonx/config.yaml

@@ -1,5 +1,5 @@
 version: 2
-image_name: watsonx
+distro_name: watsonx
 apis:
 - agents
 - datasetio
@@ -118,6 +118,9 @@ storage:
     prompts:
       namespace: prompts
       backend: kv_default
+    connectors:
+      namespace: connectors
+      backend: kv_default
 registered_resources:
   models: []
   shields: []

llama_stack/log.py

@@ -51,6 +51,7 @@ CATEGORIES = [
     "post_training",
     "scoring",
     "tests",
+    "connectors",
 ]
 UNCATEGORIZED = "uncategorized"