llama-stack 0.3.3__tar.gz → 0.3.5__tar.gz

{llama_stack-0.3.3/llama_stack.egg-info → llama_stack-0.3.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llama_stack
-Version: 0.3.3
+Version: 0.3.5
 Summary: Llama Stack
 Author-email: Meta Llama <llama-oss@meta.com>
 License: MIT
@@ -22,7 +22,7 @@ Requires-Dist: fire
 Requires-Dist: httpx
 Requires-Dist: jinja2>=3.1.6
 Requires-Dist: jsonschema
-Requires-Dist: llama-stack-client>=0.3.3
+Requires-Dist: llama-stack-client==0.3.5
 Requires-Dist: openai>=1.107
 Requires-Dist: prompt-toolkit
 Requires-Dist: python-dotenv
@@ -41,10 +41,11 @@ Requires-Dist: opentelemetry-exporter-otlp-proto-http>=1.30.0
 Requires-Dist: aiosqlite>=0.21.0
 Requires-Dist: asyncpg
 Requires-Dist: sqlalchemy[asyncio]>=2.0.41
+Requires-Dist: starlette>=0.49.1
 Provides-Extra: ui
 Requires-Dist: streamlit; extra == "ui"
 Requires-Dist: pandas; extra == "ui"
-Requires-Dist: llama-stack-client>=0.3.3; extra == "ui"
+Requires-Dist: llama-stack-client==0.3.5; extra == "ui"
 Requires-Dist: streamlit-option-menu; extra == "ui"
 Dynamic: license-file
 
@@ -60,83 +61,6 @@ Dynamic: license-file
 [**Quick Start**](https://llamastack.github.io/docs/getting_started/quickstart) | [**Documentation**](https://llamastack.github.io/docs) | [**Colab Notebook**](./docs/getting_started.ipynb) | [**Discord**](https://discord.gg/llama-stack)
 
 
-### ✨🎉 Llama 4 Support  🎉✨
-We released [Version 0.2.0](https://github.com/meta-llama/llama-stack/releases/tag/v0.2.0) with support for the Llama 4 herd of models released by Meta.
-
-<details>
-
-<summary>👋 Click here to see how to run Llama 4 models on Llama Stack </summary>
-
-\
-*Note you need 8xH100 GPU-host to run these models*
-
-```bash
-pip install -U llama_stack
-
-MODEL="Llama-4-Scout-17B-16E-Instruct"
-# get meta url from llama.com
-huggingface-cli download meta-llama/$MODEL --local-dir ~/.llama/$MODEL
-
-# install dependencies for the distribution
-llama stack list-deps meta-reference-gpu | xargs -L1 uv pip install
-
-# start a llama stack server
-INFERENCE_MODEL=meta-llama/$MODEL llama stack run meta-reference-gpu
-
-# install client to interact with the server
-pip install llama-stack-client
-```
-### CLI
-```bash
-# Run a chat completion
-MODEL="Llama-4-Scout-17B-16E-Instruct"
-
-llama-stack-client --endpoint http://localhost:8321 \
-inference chat-completion \
---model-id meta-llama/$MODEL \
---message "write a haiku for meta's llama 4 models"
-
-OpenAIChatCompletion(
-    ...
-    choices=[
-        OpenAIChatCompletionChoice(
-            finish_reason='stop',
-            index=0,
-            message=OpenAIChatCompletionChoiceMessageOpenAIAssistantMessageParam(
-                role='assistant',
-                content='...**Silent minds awaken,**  \n**Whispers of billions of words,**  \n**Reasoning breaks the night.**  \n\n—  \n*This haiku blends the essence of LLaMA 4\'s capabilities with nature-inspired metaphor, evoking its vast training data and transformative potential.*',
-                ...
-            ),
-            ...
-        )
-    ],
-    ...
-)
-```
-### Python SDK
-```python
-from llama_stack_client import LlamaStackClient
-
-client = LlamaStackClient(base_url=f"http://localhost:8321")
-
-model_id = "meta-llama/Llama-4-Scout-17B-16E-Instruct"
-prompt = "Write a haiku about coding"
-
-print(f"User> {prompt}")
-response = client.chat.completions.create(
-    model=model_id,
-    messages=[
-        {"role": "system", "content": "You are a helpful assistant."},
-        {"role": "user", "content": prompt},
-    ],
-)
-print(f"Assistant> {response.choices[0].message.content}")
-```
-As more providers start supporting Llama 4, you can use them in Llama Stack as well. We are adding to the list. Stay tuned!
-
-
-</details>
-
 ### 🚀 One-Line Installer 🚀
 
 To try Llama Stack locally, run:

{llama_stack-0.3.3 → llama_stack-0.3.5}/README.md

@@ -10,83 +10,6 @@
 [**Quick Start**](https://llamastack.github.io/docs/getting_started/quickstart) | [**Documentation**](https://llamastack.github.io/docs) | [**Colab Notebook**](./docs/getting_started.ipynb) | [**Discord**](https://discord.gg/llama-stack)
 
 
-### ✨🎉 Llama 4 Support  🎉✨
-We released [Version 0.2.0](https://github.com/meta-llama/llama-stack/releases/tag/v0.2.0) with support for the Llama 4 herd of models released by Meta.
-
-<details>
-
-<summary>👋 Click here to see how to run Llama 4 models on Llama Stack </summary>
-
-\
-*Note you need 8xH100 GPU-host to run these models*
-
-```bash
-pip install -U llama_stack
-
-MODEL="Llama-4-Scout-17B-16E-Instruct"
-# get meta url from llama.com
-huggingface-cli download meta-llama/$MODEL --local-dir ~/.llama/$MODEL
-
-# install dependencies for the distribution
-llama stack list-deps meta-reference-gpu | xargs -L1 uv pip install
-
-# start a llama stack server
-INFERENCE_MODEL=meta-llama/$MODEL llama stack run meta-reference-gpu
-
-# install client to interact with the server
-pip install llama-stack-client
-```
-### CLI
-```bash
-# Run a chat completion
-MODEL="Llama-4-Scout-17B-16E-Instruct"
-
-llama-stack-client --endpoint http://localhost:8321 \
-inference chat-completion \
---model-id meta-llama/$MODEL \
---message "write a haiku for meta's llama 4 models"
-
-OpenAIChatCompletion(
-    ...
-    choices=[
-        OpenAIChatCompletionChoice(
-            finish_reason='stop',
-            index=0,
-            message=OpenAIChatCompletionChoiceMessageOpenAIAssistantMessageParam(
-                role='assistant',
-                content='...**Silent minds awaken,**  \n**Whispers of billions of words,**  \n**Reasoning breaks the night.**  \n\n—  \n*This haiku blends the essence of LLaMA 4\'s capabilities with nature-inspired metaphor, evoking its vast training data and transformative potential.*',
-                ...
-            ),
-            ...
-        )
-    ],
-    ...
-)
-```
-### Python SDK
-```python
-from llama_stack_client import LlamaStackClient
-
-client = LlamaStackClient(base_url=f"http://localhost:8321")
-
-model_id = "meta-llama/Llama-4-Scout-17B-16E-Instruct"
-prompt = "Write a haiku about coding"
-
-print(f"User> {prompt}")
-response = client.chat.completions.create(
-    model=model_id,
-    messages=[
-        {"role": "system", "content": "You are a helpful assistant."},
-        {"role": "user", "content": prompt},
-    ],
-)
-print(f"Assistant> {response.choices[0].message.content}")
-```
-As more providers start supporting Llama 4, you can use them in Llama Stack as well. We are adding to the list. Stay tuned!
-
-
-</details>
-
 ### 🚀 One-Line Installer 🚀
 
 To try Llama Stack locally, run:

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack/core/resolver.py

@@ -391,6 +391,9 @@ async def instantiate_provider(
         method = "get_adapter_impl"
         args = [config, deps]
 
+        if "policy" in inspect.signature(getattr(module, method)).parameters:
+            args.append(policy)
+
     elif isinstance(provider_spec, AutoRoutedProviderSpec):
         method = "get_auto_router_impl"
 

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack/core/routers/inference.py

@@ -49,10 +49,17 @@ from llama_stack.apis.inference import (
 )
 from llama_stack.apis.models import Model, ModelType
 from llama_stack.apis.telemetry import MetricEvent, MetricInResponse, Telemetry
+from llama_stack.core.access_control.access_control import is_action_allowed
+from llama_stack.core.datatypes import ModelWithOwner
+from llama_stack.core.request_headers import get_authenticated_user
 from llama_stack.log import get_logger
 from llama_stack.models.llama.llama3.chat_format import ChatFormat
 from llama_stack.models.llama.llama3.tokenizer import Tokenizer
-from llama_stack.providers.datatypes import HealthResponse, HealthStatus, RoutingTable
+from llama_stack.providers.datatypes import (
+    HealthResponse,
+    HealthStatus,
+    RoutingTable,
+)
 from llama_stack.providers.utils.inference.inference_store import InferenceStore
 from llama_stack.providers.utils.telemetry.tracing import enqueue_event, get_current_span
 
@@ -186,15 +193,41 @@ class InferenceRouter(Inference):
             provider = await self.routing_table.get_provider_impl(model.identifier)
             return provider, model.provider_resource_id
 
+        # Handles cases where clients use the provider format directly
+        return await self._get_provider_by_fallback(model_id, expected_model_type)
+
+    async def _get_provider_by_fallback(self, model_id: str, expected_model_type: str) -> tuple[Inference, str]:
+        """
+        Handle fallback case where model_id is in provider_id/provider_resource_id format.
+        """
         splits = model_id.split("/", maxsplit=1)
         if len(splits) != 2:
             raise ModelNotFoundError(model_id)
 
         provider_id, provider_resource_id = splits
+
+        # Check if provider exists
         if provider_id not in self.routing_table.impls_by_provider_id:
             logger.warning(f"Provider {provider_id} not found for model {model_id}")
             raise ModelNotFoundError(model_id)
 
+        # Create a temporary model object for RBAC check
+        temp_model = ModelWithOwner(
+            identifier=model_id,
+            provider_id=provider_id,
+            provider_resource_id=provider_resource_id,
+            model_type=expected_model_type,
+            metadata={},  # Empty metadata for temporary object
+        )
+
+        # Perform RBAC check
+        user = get_authenticated_user()
+        if not is_action_allowed(self.routing_table.policy, "read", temp_model, user):
+            logger.debug(
+                f"Access denied to model '{model_id}' via fallback path for user {user.principal if user else 'anonymous'}"
+            )
+            raise ModelNotFoundError(model_id)
+
         return self.routing_table.impls_by_provider_id[provider_id], provider_resource_id
 
     async def openai_completion(

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack/providers/registry/agents.py

@@ -23,7 +23,7 @@ def available_providers() -> list[ProviderSpec]:
                 "pillow",
                 "pandas",
                 "scikit-learn",
-                "mcp>=1.8.1",
+                "mcp>=1.23.0",
             ]
             + kvstore_dependencies(),  # TODO make this dynamic based on the kvstore config
             module="llama_stack.providers.inline.agents.meta_reference",

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack/providers/registry/tool_runtime.py

@@ -80,7 +80,7 @@ def available_providers() -> list[ProviderSpec]:
             provider_type="remote::model-context-protocol",
             module="llama_stack.providers.remote.tool_runtime.model_context_protocol",
             config_class="llama_stack.providers.remote.tool_runtime.model_context_protocol.config.MCPProviderConfig",
-            pip_packages=["mcp>=1.8.1"],
+            pip_packages=["mcp>=1.23.0"],
             provider_data_validator="llama_stack.providers.remote.tool_runtime.model_context_protocol.config.MCPProviderDataValidator",
             description="Model Context Protocol (MCP) tool for standardized tool calling and context management.",
         ),

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack/providers/remote/inference/watsonx/watsonx.py

@@ -283,8 +283,8 @@ class WatsonXInferenceAdapter(LiteLLMOpenAIMixin):
             # ...
             provider_resource_id = f"{self.__provider_id__}/{model_spec['model_id']}"
             if "embedding" in functions:
-                embedding_dimension = model_spec["model_limits"]["embedding_dimension"]
-                context_length = model_spec["model_limits"]["max_sequence_length"]
+                embedding_dimension = model_spec.get("model_limits", {}).get("embedding_dimension", 0)
+                context_length = model_spec.get("model_limits", {}).get("max_sequence_length", 0)
                 embedding_metadata = {
                     "embedding_dimension": embedding_dimension,
                     "context_length": context_length,
@@ -306,10 +306,6 @@ class WatsonXInferenceAdapter(LiteLLMOpenAIMixin):
                     metadata={},
                     model_type=ModelType.llm,
                 )
-                # In theory, I guess it is possible that a model could be both an embedding model and a text chat model.
-                # In that case, the cache will record the generator Model object, and the list which we return will have
-                # both the generator Model object and the text chat Model object.  That's fine because the cache is
-                # only used for check_model_availability() anyway.
                 self._model_cache[provider_resource_id] = model
                 models.append(model)
         return models

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack/providers/utils/inference/inference_store.py

@@ -56,7 +56,7 @@ class InferenceStore:
             logger.debug("Write queue disabled for SQLite (WAL mode handles concurrency)")
 
         await self.sql_store.create_table(
-            "chat_completions",
+            self.reference.table_name,
             {
                 "id": ColumnDefinition(type=ColumnType.STRING, primary_key=True),
                 "created": ColumnType.INTEGER,
@@ -66,14 +66,6 @@ class InferenceStore:
             },
         )
 
-        if self.enable_write_queue:
-            self._queue = asyncio.Queue(maxsize=self._max_write_queue_size)
-            for _ in range(self._num_writers):
-                self._worker_tasks.append(asyncio.create_task(self._worker_loop()))
-            logger.debug(
-                f"Inference store write queue enabled with {self._num_writers} writers, max queue size {self._max_write_queue_size}"
-            )
-
     async def shutdown(self) -> None:
         if not self._worker_tasks:
             return
@@ -161,7 +153,7 @@ class InferenceStore:
 
         try:
             await self.sql_store.insert(
-                table="chat_completions",
+                table=self.reference.table_name,
                 data=record_data,
             )
         except IntegrityError as e:
@@ -173,7 +165,7 @@ class InferenceStore:
             error_message = str(e.orig) if e.orig else str(e)
             if self._is_unique_constraint_error(error_message):
                 # Update the existing record instead
-                await self.sql_store.update(table="chat_completions", data=record_data, where={"id": data["id"]})
+                await self.sql_store.update(table=self.reference.table_name, data=record_data, where={"id": data["id"]})
             else:
                 # Re-raise if it's not a unique constraint error
                 raise
@@ -217,7 +209,7 @@ class InferenceStore:
             where_conditions["model"] = model
 
         paginated_result = await self.sql_store.fetch_all(
-            table="chat_completions",
+            table=self.reference.table_name,
             where=where_conditions if where_conditions else None,
             order_by=[("created", order.value)],
             cursor=("id", after) if after else None,
@@ -246,7 +238,7 @@ class InferenceStore:
             raise ValueError("Inference store is not initialized")
 
         row = await self.sql_store.fetch_one(
-            table="chat_completions",
+            table=self.reference.table_name,
             where={"id": completion_id},
         )
 

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack/providers/utils/responses/responses_store.py

@@ -3,7 +3,6 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
-import asyncio
 
 from llama_stack.apis.agents import (
     Order,
@@ -18,12 +17,12 @@ from llama_stack.apis.agents.openai_responses import (
 )
 from llama_stack.apis.inference import OpenAIMessageParam
 from llama_stack.core.datatypes import AccessRule
-from llama_stack.core.storage.datatypes import ResponsesStoreReference, SqlStoreReference, StorageBackendType
+from llama_stack.core.storage.datatypes import ResponsesStoreReference, SqlStoreReference
 from llama_stack.log import get_logger
 
 from ..sqlstore.api import ColumnDefinition, ColumnType
 from ..sqlstore.authorized_sqlstore import AuthorizedSqlStore
-from ..sqlstore.sqlstore import _SQLSTORE_BACKENDS, sqlstore_impl
+from ..sqlstore.sqlstore import sqlstore_impl
 
 logger = get_logger(name=__name__, category="openai_responses")
 
@@ -60,13 +59,6 @@ class ResponsesStore:
         base_store = sqlstore_impl(self.reference)
         self.sql_store = AuthorizedSqlStore(base_store, self.policy)
 
-        # Disable write queue for SQLite since WAL mode handles concurrency
-        # Keep it enabled for other backends (like Postgres) for performance
-        backend_config = _SQLSTORE_BACKENDS.get(self.reference.backend)
-        if backend_config and backend_config.type == StorageBackendType.SQL_SQLITE:
-            self.enable_write_queue = False
-            logger.debug("Write queue disabled for SQLite (WAL mode handles concurrency)")
-
         await self.sql_store.create_table(
             "openai_responses",
             {
@@ -85,14 +77,6 @@ class ResponsesStore:
             },
         )
 
-        if self.enable_write_queue:
-            self._queue = asyncio.Queue(maxsize=self._max_write_queue_size)
-            for _ in range(self._num_writers):
-                self._worker_tasks.append(asyncio.create_task(self._worker_loop()))
-            logger.debug(
-                f"Responses store write queue enabled with {self._num_writers} writers, max queue size {self._max_write_queue_size}"
-            )
-
     async def shutdown(self) -> None:
         return
 

{llama_stack-0.3.3 → llama_stack-0.3.5/llama_stack.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llama_stack
-Version: 0.3.3
+Version: 0.3.5
 Summary: Llama Stack
 Author-email: Meta Llama <llama-oss@meta.com>
 License: MIT
@@ -22,7 +22,7 @@ Requires-Dist: fire
 Requires-Dist: httpx
 Requires-Dist: jinja2>=3.1.6
 Requires-Dist: jsonschema
-Requires-Dist: llama-stack-client>=0.3.3
+Requires-Dist: llama-stack-client==0.3.5
 Requires-Dist: openai>=1.107
 Requires-Dist: prompt-toolkit
 Requires-Dist: python-dotenv
@@ -41,10 +41,11 @@ Requires-Dist: opentelemetry-exporter-otlp-proto-http>=1.30.0
 Requires-Dist: aiosqlite>=0.21.0
 Requires-Dist: asyncpg
 Requires-Dist: sqlalchemy[asyncio]>=2.0.41
+Requires-Dist: starlette>=0.49.1
 Provides-Extra: ui
 Requires-Dist: streamlit; extra == "ui"
 Requires-Dist: pandas; extra == "ui"
-Requires-Dist: llama-stack-client>=0.3.3; extra == "ui"
+Requires-Dist: llama-stack-client==0.3.5; extra == "ui"
 Requires-Dist: streamlit-option-menu; extra == "ui"
 Dynamic: license-file
 
@@ -60,83 +61,6 @@ Dynamic: license-file
 [**Quick Start**](https://llamastack.github.io/docs/getting_started/quickstart) | [**Documentation**](https://llamastack.github.io/docs) | [**Colab Notebook**](./docs/getting_started.ipynb) | [**Discord**](https://discord.gg/llama-stack)
 
 
-### ✨🎉 Llama 4 Support  🎉✨
-We released [Version 0.2.0](https://github.com/meta-llama/llama-stack/releases/tag/v0.2.0) with support for the Llama 4 herd of models released by Meta.
-
-<details>
-
-<summary>👋 Click here to see how to run Llama 4 models on Llama Stack </summary>
-
-\
-*Note you need 8xH100 GPU-host to run these models*
-
-```bash
-pip install -U llama_stack
-
-MODEL="Llama-4-Scout-17B-16E-Instruct"
-# get meta url from llama.com
-huggingface-cli download meta-llama/$MODEL --local-dir ~/.llama/$MODEL
-
-# install dependencies for the distribution
-llama stack list-deps meta-reference-gpu | xargs -L1 uv pip install
-
-# start a llama stack server
-INFERENCE_MODEL=meta-llama/$MODEL llama stack run meta-reference-gpu
-
-# install client to interact with the server
-pip install llama-stack-client
-```
-### CLI
-```bash
-# Run a chat completion
-MODEL="Llama-4-Scout-17B-16E-Instruct"
-
-llama-stack-client --endpoint http://localhost:8321 \
-inference chat-completion \
---model-id meta-llama/$MODEL \
---message "write a haiku for meta's llama 4 models"
-
-OpenAIChatCompletion(
-    ...
-    choices=[
-        OpenAIChatCompletionChoice(
-            finish_reason='stop',
-            index=0,
-            message=OpenAIChatCompletionChoiceMessageOpenAIAssistantMessageParam(
-                role='assistant',
-                content='...**Silent minds awaken,**  \n**Whispers of billions of words,**  \n**Reasoning breaks the night.**  \n\n—  \n*This haiku blends the essence of LLaMA 4\'s capabilities with nature-inspired metaphor, evoking its vast training data and transformative potential.*',
-                ...
-            ),
-            ...
-        )
-    ],
-    ...
-)
-```
-### Python SDK
-```python
-from llama_stack_client import LlamaStackClient
-
-client = LlamaStackClient(base_url=f"http://localhost:8321")
-
-model_id = "meta-llama/Llama-4-Scout-17B-16E-Instruct"
-prompt = "Write a haiku about coding"
-
-print(f"User> {prompt}")
-response = client.chat.completions.create(
-    model=model_id,
-    messages=[
-        {"role": "system", "content": "You are a helpful assistant."},
-        {"role": "user", "content": prompt},
-    ],
-)
-print(f"Assistant> {response.choices[0].message.content}")
-```
-As more providers start supporting Llama 4, you can use them in Llama Stack as well. We are adding to the list. Stay tuned!
-
-
-</details>
-
 ### 🚀 One-Line Installer 🚀
 
 To try Llama Stack locally, run:

{llama_stack-0.3.3 → llama_stack-0.3.5}/llama_stack.egg-info/requires.txt

@@ -4,7 +4,7 @@ fire
 httpx
 jinja2>=3.1.6
 jsonschema
-llama-stack-client>=0.3.3
+llama-stack-client==0.3.5
 openai>=1.107
 prompt-toolkit
 python-dotenv
@@ -23,9 +23,10 @@ opentelemetry-exporter-otlp-proto-http>=1.30.0
 aiosqlite>=0.21.0
 asyncpg
 sqlalchemy[asyncio]>=2.0.41
+starlette>=0.49.1
 
 [ui]
 streamlit
 pandas
-llama-stack-client>=0.3.3
+llama-stack-client==0.3.5
 streamlit-option-menu

{llama_stack-0.3.3 → llama_stack-0.3.5}/pyproject.toml

@@ -7,7 +7,7 @@ required-version = ">=0.7.0"
 
 [project]
 name = "llama_stack"
-version = "0.3.3"
+version = "0.3.5"
 authors = [{ name = "Meta Llama", email = "llama-oss@meta.com" }]
 description = "Llama Stack"
 readme = "README.md"
@@ -30,7 +30,7 @@ dependencies = [
     "httpx",
     "jinja2>=3.1.6",
     "jsonschema",
-    "llama-stack-client>=0.3.3",
+    "llama-stack-client==0.3.5",
     "openai>=1.107",                                  # for expires_after support
     "prompt-toolkit",
     "python-dotenv",
@@ -49,13 +49,14 @@ dependencies = [
     "aiosqlite>=0.21.0",                              # server - for metadata store
     "asyncpg",                                        # for metadata store
     "sqlalchemy[asyncio]>=2.0.41",                    # server - for conversations
+    "starlette>=0.49.1",
 ]
 
 [project.optional-dependencies]
 ui = [
     "streamlit",
     "pandas",
-    "llama-stack-client>=0.3.3",
+    "llama-stack-client==0.3.5",
     "streamlit-option-menu",
 ]