Merge branch 'main' into delete_unused_imports

.github/workflows/unit-tests.yml

@@ -33,7 +33,7 @@ jobs:
 
       - name: Run unit tests
         run: |
-          uv run --python ${{ matrix.python }} --with-editable . --with-editable ".[dev]" --with-editable ".[unit]" pytest --cov=llama_stack -s -v tests/unit/ --junitxml=pytest-report-${{ matrix.python }}.xml
+          uv run --python ${{ matrix.python }} --with-editable . --with-editable ".[dev]" --with-editable ".[unit]" pytest --cov=llama_stack -s -v tests/unit/ --junitxml=pytest-report-${{ matrix.python }}.xml --cov-report=html:htmlcov-${{ matrix.python }}
 
       - name: Upload test results
         if: always()
@@ -43,4 +43,5 @@ jobs:
           path: |
             .pytest_cache/
             pytest-report-${{ matrix.python }}.xml
+            htmlcov-${{ matrix.python }}/
           retention-days: 7

.gitignore

@@ -22,3 +22,4 @@ pyrightconfig.json
 venv/
 pytest-report.xml
 .coverage
+.python-version

.python-version

@@ -1 +0,0 @@
-3.10

CONTRIBUTING.md

@@ -61,6 +61,7 @@ outlined on that page and do not file a public issue.
 
 We use [uv](https://github.com/astral-sh/uv) to manage python dependencies and virtual environments.
 You can install `uv` by following this [guide](https://docs.astral.sh/uv/getting-started/installation/).
+
 You can install the dependencies by running:
 
 ```bash
@@ -70,6 +71,11 @@ uv pip install -e .
 source .venv/bin/activate
 ```
 
+> [!NOTE]
+> You can pin a specific version of Python to use for `uv` by adding a `.python-version` file in the root project directory.
+> Otherwise, `uv` will automatically select a Python version according to the `requires-python` section of the `pyproject.toml`.
+> For more info, see the [uv docs around Python versions](https://docs.astral.sh/uv/concepts/python-versions/).
+
 Note that you can create a dotenv file `.env` that includes necessary environment variables:
 ```
 LLAMA_STACK_BASE_URL=http://localhost:8321

docs/_static/llama-stack-spec.html

@@ -4347,24 +4347,6 @@
                         "type": "string",
                         "description": "Unique identifier for the tool call this response is for"
                     },
-                    "tool_name": {
-                        "oneOf": [
-                            {
-                                "type": "string",
-                                "enum": [
-                                    "brave_search",
-                                    "wolfram_alpha",
-                                    "photogen",
-                                    "code_interpreter"
-                                ],
-                                "title": "BuiltinTool"
-                            },
-                            {
-                                "type": "string"
-                            }
-                        ],
-                        "description": "Name of the tool that was called"
-                    },
                     "content": {
                         "$ref": "#/components/schemas/InterleavedContent",
                         "description": "The response content from the tool"
@@ -4374,7 +4356,6 @@
                 "required": [
                     "role",
                     "call_id",
-                    "tool_name",
                     "content"
                 ],
                 "title": "ToolResponseMessage",

docs/_static/llama-stack-spec.yaml

@@ -2943,17 +2943,6 @@ components:
           type: string
           description: >-
             Unique identifier for the tool call this response is for
-        tool_name:
-          oneOf:
-            - type: string
-              enum:
-                - brave_search
-                - wolfram_alpha
-                - photogen
-                - code_interpreter
-              title: BuiltinTool
-            - type: string
-          description: Name of the tool that was called
         content:
           $ref: '#/components/schemas/InterleavedContent'
           description: The response content from the tool
@@ -2961,7 +2950,6 @@ components:
       required:
         - role
         - call_id
-        - tool_name
         - content
       title: ToolResponseMessage
       description: >-

llama_stack/apis/inference/inference.py

@@ -117,13 +117,11 @@ class ToolResponseMessage(BaseModel):
 
     :param role: Must be "tool" to identify this as a tool response
     :param call_id: Unique identifier for the tool call this response is for
-    :param tool_name: Name of the tool that was called
     :param content: The response content from the tool
     """
 
     role: Literal["tool"] = "tool"
     call_id: str
-    tool_name: Union[BuiltinTool, str]
     content: InterleavedContent
 
 

llama_stack/providers/inline/agents/meta_reference/agent_instance.py

@@ -12,7 +12,7 @@ import secrets
 import string
 import uuid
 from datetime import datetime
-from typing import Any, AsyncGenerator, Dict, List, Optional, Tuple, Union
+from typing import AsyncGenerator, List, Optional, Union
 from urllib.parse import urlparse
 
 import httpx
@@ -153,7 +153,6 @@ class ChatAgent(ShieldRunnerMixin):
                     messages.append(
                         ToolResponseMessage(
                             call_id=response.call_id,
-                            tool_name=response.tool_name,
                             content=response.content,
                         )
                     )
@@ -181,6 +180,7 @@ class ChatAgent(ShieldRunnerMixin):
         return messages
 
     async def create_and_execute_turn(self, request: AgentTurnCreateRequest) -> AsyncGenerator:
+        await self._initialize_tools(request.toolgroups)
         async with tracing.span("create_and_execute_turn") as span:
             span.set_attribute("session_id", request.session_id)
             span.set_attribute("agent_id", self.agent_id)
@@ -191,6 +191,7 @@ class ChatAgent(ShieldRunnerMixin):
                 yield chunk
 
     async def resume_turn(self, request: AgentTurnResumeRequest) -> AsyncGenerator:
+        await self._initialize_tools()
         async with tracing.span("resume_turn") as span:
             span.set_attribute("agent_id", self.agent_id)
             span.set_attribute("session_id", request.session_id)
@@ -219,8 +220,7 @@ class ChatAgent(ShieldRunnerMixin):
         messages = await self.get_messages_from_turns(turns)
         if is_resume:
             tool_response_messages = [
-                ToolResponseMessage(call_id=x.call_id, tool_name=x.tool_name, content=x.content)
+                ToolResponseMessage(call_id=x.call_id, content=x.content) for x in request.tool_responses
-                for x in request.tool_responses
             ]
             messages.extend(tool_response_messages)
             last_turn = turns[-1]
@@ -275,7 +275,6 @@ class ChatAgent(ShieldRunnerMixin):
             sampling_params=self.agent_config.sampling_params,
             stream=request.stream,
             documents=request.documents if not is_resume else None,
-            toolgroups_for_turn=request.toolgroups if not is_resume else None,
         ):
             if isinstance(chunk, CompletionMessage):
                 output_message = chunk
@@ -327,7 +326,6 @@ class ChatAgent(ShieldRunnerMixin):
         sampling_params: SamplingParams,
         stream: bool = False,
         documents: Optional[List[Document]] = None,
-        toolgroups_for_turn: Optional[List[AgentToolGroup]] = None,
     ) -> AsyncGenerator:
         # Doing async generators makes downstream code much simpler and everything amenable to
         # streaming. However, it also makes things complicated here because AsyncGenerators cannot
@@ -350,7 +348,6 @@ class ChatAgent(ShieldRunnerMixin):
             sampling_params,
             stream,
             documents,
-            toolgroups_for_turn,
         ):
             if isinstance(res, bool):
                 return
@@ -451,30 +448,19 @@ class ChatAgent(ShieldRunnerMixin):
         sampling_params: SamplingParams,
         stream: bool = False,
         documents: Optional[List[Document]] = None,
-        toolgroups_for_turn: Optional[List[AgentToolGroup]] = None,
     ) -> AsyncGenerator:
-        # TODO: simplify all of this code, it can be simpler
-        toolgroup_args = {}
-        toolgroups = set()
-        for toolgroup in self.agent_config.toolgroups + (toolgroups_for_turn or []):
-            if isinstance(toolgroup, AgentToolGroupWithArgs):
-                tool_group_name, tool_name = self._parse_toolgroup_name(toolgroup.name)
-                toolgroups.add(tool_group_name)
-                toolgroup_args[tool_group_name] = toolgroup.args
-            else:
-                toolgroups.add(toolgroup)
-
-        tool_defs, tool_to_group = await self._get_tool_defs(toolgroups_for_turn)
         if documents:
-            await self.handle_documents(session_id, documents, input_messages, tool_defs)
+            await self.handle_documents(session_id, documents, input_messages)
 
         session_info = await self.storage.get_session_info(session_id)
         # if the session has a memory bank id, let the memory tool use it
         if session_info and session_info.vector_db_id:
-            if RAG_TOOL_GROUP not in toolgroup_args:
+            for tool_name in self.tool_name_to_args.keys():
-                toolgroup_args[RAG_TOOL_GROUP] = {"vector_db_ids": [session_info.vector_db_id]}
+                if tool_name == MEMORY_QUERY_TOOL:
-            else:
+                    if "vector_db_ids" not in self.tool_name_to_args[tool_name]:
-                toolgroup_args[RAG_TOOL_GROUP]["vector_db_ids"].append(session_info.vector_db_id)
+                        self.tool_name_to_args[tool_name]["vector_db_ids"] = [session_info.vector_db_id]
+                    else:
+                        self.tool_name_to_args[tool_name]["vector_db_ids"].append(session_info.vector_db_id)
 
         output_attachments = []
 
@@ -504,7 +490,7 @@ class ChatAgent(ShieldRunnerMixin):
                 async for chunk in await self.inference_api.chat_completion(
                     self.agent_config.model,
                     input_messages,
-                    tools=tool_defs,
+                    tools=self.tool_defs,
                     tool_prompt_format=self.agent_config.tool_config.tool_prompt_format,
                     response_format=self.agent_config.response_format,
                     stream=True,
@@ -686,12 +672,9 @@ class ChatAgent(ShieldRunnerMixin):
                 ) as span:
                     tool_execution_start_time = datetime.now().astimezone().isoformat()
                     tool_call = message.tool_calls[0]
-                    tool_result = await execute_tool_call_maybe(
+                    tool_result = await self.execute_tool_call_maybe(
-                        self.tool_runtime_api,
                         session_id,
                         tool_call,
-                        toolgroup_args,
-                        tool_to_group,
                     )
                     if tool_result.content is None:
                         raise ValueError(
@@ -700,7 +683,6 @@ class ChatAgent(ShieldRunnerMixin):
                     result_messages = [
                         ToolResponseMessage(
                             call_id=tool_call.call_id,
-                            tool_name=tool_call.tool_name,
                             content=tool_result.content,
                         )
                     ]
@@ -720,7 +702,7 @@ class ChatAgent(ShieldRunnerMixin):
                                 tool_responses=[
                                     ToolResponse(
                                         call_id=result_message.call_id,
-                                        tool_name=result_message.tool_name,
+                                        tool_name=tool_call.tool_name,
                                         content=result_message.content,
                                         metadata=tool_result.metadata,
                                     )
@@ -744,9 +726,16 @@ class ChatAgent(ShieldRunnerMixin):
 
                 input_messages = input_messages + [message, result_message]
 
-    async def _get_tool_defs(
+    async def _initialize_tools(
-        self, toolgroups_for_turn: Optional[List[AgentToolGroup]] = None
+        self,
-    ) -> Tuple[List[ToolDefinition], Dict[str, str]]:
+        toolgroups_for_turn: Optional[List[AgentToolGroup]] = None,
+    ) -> None:
+        toolgroup_to_args = {}
+        for toolgroup in (self.agent_config.toolgroups or []) + (toolgroups_for_turn or []):
+            if isinstance(toolgroup, AgentToolGroupWithArgs):
+                tool_group_name, _ = self._parse_toolgroup_name(toolgroup.name)
+                toolgroup_to_args[tool_group_name] = toolgroup.args
+
         # Determine which tools to include
         tool_groups_to_include = toolgroups_for_turn or self.agent_config.toolgroups or []
         agent_config_toolgroups = []
@@ -755,8 +744,10 @@ class ChatAgent(ShieldRunnerMixin):
             if name not in agent_config_toolgroups:
                 agent_config_toolgroups.append(name)
 
+        toolgroup_to_args = toolgroup_to_args or {}
+
         tool_name_to_def = {}
-        tool_to_group = {}
+        tool_name_to_args = {}
 
         for tool_def in self.agent_config.client_tools:
             if tool_name_to_def.get(tool_def.name, None):
@@ -774,53 +765,38 @@ class ChatAgent(ShieldRunnerMixin):
                     for param in tool_def.parameters
                 },
             )
-            tool_to_group[tool_def.name] = "__client_tools__"
         for toolgroup_name_with_maybe_tool_name in agent_config_toolgroups:
-            toolgroup_name, tool_name = self._parse_toolgroup_name(toolgroup_name_with_maybe_tool_name)
+            toolgroup_name, input_tool_name = self._parse_toolgroup_name(toolgroup_name_with_maybe_tool_name)
             tools = await self.tool_groups_api.list_tools(toolgroup_id=toolgroup_name)
             if not tools.data:
                 available_tool_groups = ", ".join(
                     [t.identifier for t in (await self.tool_groups_api.list_tool_groups()).data]
                 )
                 raise ValueError(f"Toolgroup {toolgroup_name} not found, available toolgroups: {available_tool_groups}")
-            if tool_name is not None and not any(tool.identifier == tool_name for tool in tools.data):
+            if input_tool_name is not None and not any(tool.identifier == input_tool_name for tool in tools.data):
                 raise ValueError(
-                    f"Tool {tool_name} not found in toolgroup {toolgroup_name}. Available tools: {', '.join([tool.identifier for tool in tools.data])}"
+                    f"Tool {input_tool_name} not found in toolgroup {toolgroup_name}. Available tools: {', '.join([tool.identifier for tool in tools.data])}"
                 )
 
             for tool_def in tools.data:
                 if toolgroup_name.startswith("builtin") and toolgroup_name != RAG_TOOL_GROUP:
-                    tool_name = tool_def.identifier
+                    identifier: str | BuiltinTool | None = tool_def.identifier
-                    built_in_type = BuiltinTool.brave_search
+                    if identifier == "web_search":
-                    if tool_name == "web_search":
+                        identifier = BuiltinTool.brave_search
-                        built_in_type = BuiltinTool.brave_search
                     else:
-                        built_in_type = BuiltinTool(tool_name)
+                        identifier = BuiltinTool(identifier)
+                else:
+                    # add if tool_name is unspecified or the tool_def identifier is the same as the tool_name
+                    if input_tool_name in (None, tool_def.identifier):
+                        identifier = tool_def.identifier
+                    else:
+                        identifier = None
 
-                    if tool_name_to_def.get(built_in_type, None):
+                if tool_name_to_def.get(identifier, None):
-                        raise ValueError(f"Tool {built_in_type} already exists")
+                    raise ValueError(f"Tool {identifier} already exists")
-
+                if identifier:
-                    tool_name_to_def[built_in_type] = ToolDefinition(
-                        tool_name=built_in_type,
-                        description=tool_def.description,
-                        parameters={
-                            param.name: ToolParamDefinition(
-                                param_type=param.parameter_type,
-                                description=param.description,
-                                required=param.required,
-                                default=param.default,
-                            )
-                            for param in tool_def.parameters
-                        },
-                    )
-                    tool_to_group[built_in_type] = tool_def.toolgroup_id
-                    continue
-
-                if tool_name_to_def.get(tool_def.identifier, None):
-                    raise ValueError(f"Tool {tool_def.identifier} already exists")
-                if tool_name in (None, tool_def.identifier):
                     tool_name_to_def[tool_def.identifier] = ToolDefinition(
-                        tool_name=tool_def.identifier,
+                        tool_name=identifier,
                         description=tool_def.description,
                         parameters={
                             param.name: ToolParamDefinition(
@@ -832,9 +808,9 @@ class ChatAgent(ShieldRunnerMixin):
                             for param in tool_def.parameters
                         },
                     )
-                    tool_to_group[tool_def.identifier] = tool_def.toolgroup_id
+                    tool_name_to_args[tool_def.identifier] = toolgroup_to_args.get(toolgroup_name, {})
 
-        return list(tool_name_to_def.values()), tool_to_group
+        self.tool_defs, self.tool_name_to_args = list(tool_name_to_def.values()), tool_name_to_args
 
     def _parse_toolgroup_name(self, toolgroup_name_with_maybe_tool_name: str) -> tuple[str, Optional[str]]:
         """Parse a toolgroup name into its components.
@@ -853,15 +829,46 @@ class ChatAgent(ShieldRunnerMixin):
             tool_group, tool_name = split_names[0], None
         return tool_group, tool_name
 
+    async def execute_tool_call_maybe(
+        self,
+        session_id: str,
+        tool_call: ToolCall,
+    ) -> ToolInvocationResult:
+        tool_name = tool_call.tool_name
+        registered_tool_names = [tool_def.tool_name for tool_def in self.tool_defs]
+        if tool_name not in registered_tool_names:
+            raise ValueError(
+                f"Tool {tool_name} not found in provided tools, registered tools: {', '.join([str(x) for x in registered_tool_names])}"
+            )
+        if isinstance(tool_name, BuiltinTool):
+            if tool_name == BuiltinTool.brave_search:
+                tool_name_str = WEB_SEARCH_TOOL
+            else:
+                tool_name_str = tool_name.value
+        else:
+            tool_name_str = tool_name
+
+        logger.info(f"executing tool call: {tool_name_str} with args: {tool_call.arguments}")
+        result = await self.tool_runtime_api.invoke_tool(
+            tool_name=tool_name_str,
+            kwargs={
+                "session_id": session_id,
+                # get the arguments generated by the model and augment with toolgroup arg overrides for the agent
+                **tool_call.arguments,
+                **self.tool_name_to_args.get(tool_name_str, {}),
+            },
+        )
+        logger.debug(f"tool call {tool_name_str} completed with result: {result}")
+        return result
+
     async def handle_documents(
         self,
         session_id: str,
         documents: List[Document],
         input_messages: List[Message],
-        tool_defs: Dict[str, ToolDefinition],
     ) -> None:
-        memory_tool = any(tool_def.tool_name == MEMORY_QUERY_TOOL for tool_def in tool_defs)
+        memory_tool = any(tool_def.tool_name == MEMORY_QUERY_TOOL for tool_def in self.tool_defs)
-        code_interpreter_tool = any(tool_def.tool_name == BuiltinTool.code_interpreter for tool_def in tool_defs)
+        code_interpreter_tool = any(tool_def.tool_name == BuiltinTool.code_interpreter for tool_def in self.tool_defs)
         content_items = []
         url_items = []
         pattern = re.compile("^(https?://|file://|data:)")
@@ -989,42 +996,10 @@ async def attachment_message(tempdir: str, urls: List[URL]) -> ToolResponseMessa
 
     return ToolResponseMessage(
         call_id="",
-        tool_name=BuiltinTool.code_interpreter,
         content=content,
     )
 
 
-async def execute_tool_call_maybe(
-    tool_runtime_api: ToolRuntime,
-    session_id: str,
-    tool_call: ToolCall,
-    toolgroup_args: Dict[str, Dict[str, Any]],
-    tool_to_group: Dict[str, str],
-) -> ToolInvocationResult:
-    name = tool_call.tool_name
-    group_name = tool_to_group.get(name, None)
-    if group_name is None:
-        raise ValueError(f"Tool {name} not found in any tool group")
-    if isinstance(name, BuiltinTool):
-        if name == BuiltinTool.brave_search:
-            name = WEB_SEARCH_TOOL
-        else:
-            name = name.value
-
-    logger.info(f"executing tool call: {name} with args: {tool_call.arguments}")
-    result = await tool_runtime_api.invoke_tool(
-        tool_name=name,
-        kwargs={
-            "session_id": session_id,
-            # get the arguments generated by the model and augment with toolgroup arg overrides for the agent
-            **tool_call.arguments,
-            **toolgroup_args.get(group_name, {}),
-        },
-    )
-    logger.info(f"tool call {name} completed with result: {result}")
-    return result
-
-
 def _interpret_content_as_attachment(
     content: str,
 ) -> Optional[Attachment]:

llama_stack/providers/inline/datasetio/localfs/config.py

@@ -3,9 +3,10 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
-from llama_stack.distribution.utils.config_dirs import RUNTIME_BASE_DIR
 from llama_stack.providers.utils.kvstore.config import (
     KVStoreConfig,
     SqliteKVStoreConfig,
@@ -13,6 +14,13 @@ from llama_stack.providers.utils.kvstore.config import (
 
 
 class LocalFSDatasetIOConfig(BaseModel):
-    kvstore: KVStoreConfig = SqliteKVStoreConfig(
+    kvstore: KVStoreConfig
-        db_path=(RUNTIME_BASE_DIR / "localfs_datasetio.db").as_posix()
+
-    )  # Uses SQLite config specific to localfs storage
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "kvstore": SqliteKVStoreConfig.sample_run_config(
+                __distro_dir__=__distro_dir__,
+                db_name="localfs_datasetio.db",
+            )
+        }

llama_stack/providers/inline/eval/meta_reference/config.py

@@ -3,9 +3,10 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
-from llama_stack.distribution.utils.config_dirs import RUNTIME_BASE_DIR
 from llama_stack.providers.utils.kvstore.config import (
     KVStoreConfig,
     SqliteKVStoreConfig,
@@ -13,6 +14,13 @@ from llama_stack.providers.utils.kvstore.config import (
 
 
 class MetaReferenceEvalConfig(BaseModel):
-    kvstore: KVStoreConfig = SqliteKVStoreConfig(
+    kvstore: KVStoreConfig
-        db_path=(RUNTIME_BASE_DIR / "meta_reference_eval.db").as_posix()
+
-    )  # Uses SQLite config specific to Meta Reference Eval storage
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "kvstore": SqliteKVStoreConfig.sample_run_config(
+                __distro_dir__=__distro_dir__,
+                db_name="meta_reference_eval.db",
+            )
+        }

llama_stack/providers/inline/inference/vllm/config.py

@@ -4,6 +4,8 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+from typing import Any, Dict
+
 from pydantic import BaseModel, Field
 
 from llama_stack.schema_utils import json_schema_type
@@ -40,7 +42,7 @@ class VLLMConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls):
+    def sample_run_config(cls, **kwargs: Any) -> Dict[str, Any]:
         return {
             "tensor_parallel_size": "${env.TENSOR_PARALLEL_SIZE:1}",
             "max_tokens": "${env.MAX_TOKENS:4096}",

llama_stack/providers/inline/post_training/torchtune/config.py

@@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from typing import Literal, Optional
+from typing import Any, Dict, Literal, Optional
 
 from pydantic import BaseModel
 
@@ -12,3 +12,9 @@ from pydantic import BaseModel
 class TorchtunePostTrainingConfig(BaseModel):
     torch_seed: Optional[int] = None
     checkpoint_format: Optional[Literal["meta", "huggingface"]] = "meta"
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "checkpoint_format": "meta",
+        }

llama_stack/providers/inline/safety/code_scanner/config.py

@@ -4,8 +4,12 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
 
 class CodeScannerConfig(BaseModel):
-    pass
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {}

llama_stack/providers/inline/safety/llama_guard/config.py

@@ -4,10 +4,16 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from typing import List
+from typing import Any, Dict, List
 
 from pydantic import BaseModel
 
 
 class LlamaGuardConfig(BaseModel):
     excluded_categories: List[str] = []
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "excluded_categories": [],
+        }

llama_stack/providers/inline/safety/prompt_guard/config.py

@@ -5,6 +5,7 @@
 # the root directory of this source tree.
 
 from enum import Enum
+from typing import Any, Dict
 
 from pydantic import BaseModel, field_validator
 
@@ -23,3 +24,9 @@ class PromptGuardConfig(BaseModel):
         if v not in [t.value for t in PromptGuardType]:
             raise ValueError(f"Unknown prompt guard type: {v}")
         return v
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "guard_type": "injection",
+        }

llama_stack/providers/inline/scoring/basic/config.py

@@ -3,7 +3,12 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
 
-class BasicScoringConfig(BaseModel): ...
+class BasicScoringConfig(BaseModel):
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {}

llama_stack/providers/inline/scoring/llm_as_judge/config.py

@@ -3,7 +3,12 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
 
-class LlmAsJudgeScoringConfig(BaseModel): ...
+class LlmAsJudgeScoringConfig(BaseModel):
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {}

llama_stack/providers/inline/telemetry/sample/__init__.py

@@ -1,17 +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 typing import Any
-
-from .config import SampleConfig
-
-
-async def get_adapter_impl(config: SampleConfig, _deps) -> Any:
-    from .sample import SampleTelemetryImpl
-
-    impl = SampleTelemetryImpl(config)
-    await impl.initialize()
-    return impl

llama_stack/providers/inline/telemetry/sample/config.py

@@ -1,12 +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 pydantic import BaseModel
-
-
-class SampleConfig(BaseModel):
-    host: str = "localhost"
-    port: int = 9999

llama_stack/providers/inline/telemetry/sample/sample.py

@@ -1,17 +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.apis.telemetry import Telemetry
-
-from .config import SampleConfig
-
-
-class SampleTelemetryImpl(Telemetry):
-    def __init__(self, config: SampleConfig):
-        self.config = config
-
-    async def initialize(self):
-        pass

llama_stack/providers/inline/tool_runtime/code_interpreter/code_execution.py

@@ -76,6 +76,7 @@ class CodeExecutionRequest:
     only_last_cell_fail: bool = True
     seed: int = 0
     strip_fpaths_in_stderr: bool = True
+    use_bwrap: bool = True
 
 
 class CodeExecutor:
@@ -103,8 +104,6 @@ _set_seeds()\
 
         script = "\n\n".join([seeds_prefix] + [CODE_ENV_PREFIX] + scripts)
         with tempfile.TemporaryDirectory() as dpath:
-            bwrap_prefix = "bwrap " + generate_bwrap_command(bind_dirs=[dpath])
-            cmd = [*bwrap_prefix.split(), sys.executable, "-c", script]
             code_fpath = os.path.join(dpath, "code.py")
             with open(code_fpath, "w") as f:
                 f.write(script)
@@ -118,6 +117,13 @@ _set_seeds()\
                     MPLBACKEND="module://matplotlib_custom_backend",
                     PYTHONPATH=f"{DIRNAME}:{python_path}",
                 )
+
+                if req.use_bwrap:
+                    bwrap_prefix = "bwrap " + generate_bwrap_command(bind_dirs=[dpath])
+                    cmd = [*bwrap_prefix.split(), sys.executable, "-c", script]
+                else:
+                    cmd = [sys.executable, "-c", script]
+
                 stdout, stderr, returncode = do_subprocess(
                     cmd=cmd,
                     env=env,

llama_stack/providers/inline/tool_runtime/code_interpreter/code_interpreter.py

@@ -6,6 +6,7 @@
 
 
 import logging
+import os
 import tempfile
 from typing import Any, Dict, List, Optional
 
@@ -61,7 +62,9 @@ class CodeInterpreterToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime):
 
     async def invoke_tool(self, tool_name: str, kwargs: Dict[str, Any]) -> ToolInvocationResult:
         script = kwargs["code"]
-        req = CodeExecutionRequest(scripts=[script])
+        # Use environment variable to control bwrap usage
+        force_disable_bwrap = os.environ.get("DISABLE_CODE_SANDBOX", "").lower() in ("1", "true", "yes")
+        req = CodeExecutionRequest(scripts=[script], use_bwrap=not force_disable_bwrap)
         res = self.code_executor.execute(req)
         pieces = [res["process_status"]]
         for out_type in ["stdout", "stderr"]:

llama_stack/providers/inline/tool_runtime/code_interpreter/config.py

@@ -4,8 +4,12 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
 
 class CodeInterpreterToolConfig(BaseModel):
-    pass
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {}

llama_stack/providers/inline/tool_runtime/rag/config.py

@@ -4,8 +4,12 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
 
 class RagToolRuntimeConfig(BaseModel):
-    pass
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {}

llama_stack/providers/inline/vector_io/chroma/config.py

@@ -13,5 +13,5 @@ class ChromaVectorIOConfig(BaseModel):
     db_path: str
 
     @classmethod
-    def sample_config(cls) -> Dict[str, Any]:
+    def sample_run_config(cls, db_path: str = "${env.CHROMADB_PATH}", **kwargs: Any) -> Dict[str, Any]:
-        return {"db_path": "{env.CHROMADB_PATH}"}
+        return {"db_path": db_path}

llama_stack/providers/registry/agents.py

@@ -7,11 +7,9 @@
 from typing import List
 
 from llama_stack.providers.datatypes import (
-    AdapterSpec,
     Api,
     InlineProviderSpec,
     ProviderSpec,
-    remote_provider_spec,
 )
 from llama_stack.providers.utils.kvstore import kvstore_dependencies
 
@@ -39,13 +37,4 @@ def available_providers() -> List[ProviderSpec]:
                 Api.tool_groups,
             ],
         ),
-        remote_provider_spec(
-            api=Api.agents,
-            adapter=AdapterSpec(
-                adapter_type="sample",
-                pip_packages=[],
-                module="llama_stack.providers.remote.agents.sample",
-                config_class="llama_stack.providers.remote.agents.sample.SampleConfig",
-            ),
-        ),
     ]

llama_stack/providers/registry/inference.py

@@ -68,15 +68,6 @@ def available_providers() -> List[ProviderSpec]:
             module="llama_stack.providers.inline.inference.sentence_transformers",
             config_class="llama_stack.providers.inline.inference.sentence_transformers.config.SentenceTransformersInferenceConfig",
         ),
-        remote_provider_spec(
-            api=Api.inference,
-            adapter=AdapterSpec(
-                adapter_type="sample",
-                pip_packages=[],
-                module="llama_stack.providers.remote.inference.sample",
-                config_class="llama_stack.providers.remote.inference.sample.SampleConfig",
-            ),
-        ),
         remote_provider_spec(
             api=Api.inference,
             adapter=AdapterSpec(

llama_stack/providers/registry/safety.py

@@ -27,27 +27,6 @@ def available_providers() -> List[ProviderSpec]:
             module="llama_stack.providers.inline.safety.prompt_guard",
             config_class="llama_stack.providers.inline.safety.prompt_guard.PromptGuardConfig",
         ),
-        InlineProviderSpec(
-            api=Api.safety,
-            provider_type="inline::meta-reference",
-            pip_packages=[
-                "transformers",
-                "torch --index-url https://download.pytorch.org/whl/cpu",
-            ],
-            module="llama_stack.providers.inline.safety.meta_reference",
-            config_class="llama_stack.providers.inline.safety.meta_reference.SafetyConfig",
-            api_dependencies=[
-                Api.inference,
-            ],
-            deprecation_error="""
-Provider `inline::meta-reference` for API `safety` does not work with the latest Llama Stack.
-
-- if you are using Llama Guard v3, please use the `inline::llama-guard` provider instead.
-- if you are using Prompt Guard, please use the `inline::prompt-guard` provider instead.
-- if you are using Code Scanner, please use the `inline::code-scanner` provider instead.
-
-            """,
-        ),
         InlineProviderSpec(
             api=Api.safety,
             provider_type="inline::llama-guard",
@@ -67,15 +46,6 @@ Provider `inline::meta-reference` for API `safety` does not work with the latest
             module="llama_stack.providers.inline.safety.code_scanner",
             config_class="llama_stack.providers.inline.safety.code_scanner.CodeScannerConfig",
         ),
-        remote_provider_spec(
-            api=Api.safety,
-            adapter=AdapterSpec(
-                adapter_type="sample",
-                pip_packages=[],
-                module="llama_stack.providers.remote.safety.sample",
-                config_class="llama_stack.providers.remote.safety.sample.SampleConfig",
-            ),
-        ),
         remote_provider_spec(
             api=Api.safety,
             adapter=AdapterSpec(

llama_stack/providers/registry/telemetry.py

@@ -7,11 +7,9 @@
 from typing import List
 
 from llama_stack.providers.datatypes import (
-    AdapterSpec,
     Api,
     InlineProviderSpec,
     ProviderSpec,
-    remote_provider_spec,
 )
 
 
@@ -28,13 +26,4 @@ def available_providers() -> List[ProviderSpec]:
             module="llama_stack.providers.inline.telemetry.meta_reference",
             config_class="llama_stack.providers.inline.telemetry.meta_reference.config.TelemetryConfig",
         ),
-        remote_provider_spec(
-            api=Api.telemetry,
-            adapter=AdapterSpec(
-                adapter_type="sample",
-                pip_packages=[],
-                module="llama_stack.providers.remote.telemetry.sample",
-                config_class="llama_stack.providers.remote.telemetry.sample.SampleConfig",
-            ),
-        ),
     ]

llama_stack/providers/registry/vector_io.py

@@ -92,16 +92,6 @@ def available_providers() -> List[ProviderSpec]:
             ),
             api_dependencies=[Api.inference],
         ),
-        remote_provider_spec(
-            api=Api.vector_io,
-            adapter=AdapterSpec(
-                adapter_type="sample",
-                pip_packages=[],
-                module="llama_stack.providers.remote.vector_io.sample",
-                config_class="llama_stack.providers.remote.vector_io.sample.SampleVectorIOConfig",
-            ),
-            api_dependencies=[],
-        ),
         remote_provider_spec(
             Api.vector_io,
             AdapterSpec(

llama_stack/providers/remote/agents/sample/__init__.py

@@ -1,17 +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 typing import Any
-
-from .config import SampleConfig
-
-
-async def get_adapter_impl(config: SampleConfig, _deps) -> Any:
-    from .sample import SampleAgentsImpl
-
-    impl = SampleAgentsImpl(config)
-    await impl.initialize()
-    return impl

llama_stack/providers/remote/agents/sample/config.py

@@ -1,12 +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 pydantic import BaseModel
-
-
-class SampleConfig(BaseModel):
-    host: str = "localhost"
-    port: int = 9999

llama_stack/providers/remote/agents/sample/sample.py

@@ -1,17 +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.apis.agents import Agents
-
-from .config import SampleConfig
-
-
-class SampleAgentsImpl(Agents):
-    def __init__(self, config: SampleConfig):
-        self.config = config
-
-    async def initialize(self):
-        pass

llama_stack/providers/remote/datasetio/huggingface/config.py

@@ -3,9 +3,10 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
-from llama_stack.distribution.utils.config_dirs import RUNTIME_BASE_DIR
 from llama_stack.providers.utils.kvstore.config import (
     KVStoreConfig,
     SqliteKVStoreConfig,
@@ -13,6 +14,13 @@ from llama_stack.providers.utils.kvstore.config import (
 
 
 class HuggingfaceDatasetIOConfig(BaseModel):
-    kvstore: KVStoreConfig = SqliteKVStoreConfig(
+    kvstore: KVStoreConfig
-        db_path=(RUNTIME_BASE_DIR / "huggingface_datasetio.db").as_posix()
+
-    )  # Uses SQLite config specific to HF storage
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "kvstore": SqliteKVStoreConfig.sample_run_config(
+                __distro_dir__=__distro_dir__,
+                db_name="huggingface_datasetio.db",
+            )
+        }

llama_stack/providers/remote/inference/databricks/config.py

@@ -4,6 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+from typing import Any, Dict
 
 from pydantic import BaseModel, Field
 
@@ -20,3 +21,15 @@ class DatabricksImplConfig(BaseModel):
         default=None,
         description="The Databricks API token",
     )
+
+    @classmethod
+    def sample_run_config(
+        cls,
+        url: str = "${env.DATABRICKS_URL}",
+        api_token: str = "${env.DATABRICKS_API_TOKEN}",
+        **kwargs: Any,
+    ) -> Dict[str, Any]:
+        return {
+            "url": url,
+            "api_token": api_token,
+        }

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

@@ -5,10 +5,11 @@
 # the root directory of this source tree.
 
 from .config import RunpodImplConfig
-from .runpod import RunpodInferenceAdapter
 
 
 async def get_adapter_impl(config: RunpodImplConfig, _deps):
+    from .runpod import RunpodInferenceAdapter
+
     assert isinstance(config, RunpodImplConfig), f"Unexpected config type: {type(config)}"
     impl = RunpodInferenceAdapter(config)
     await impl.initialize()

llama_stack/providers/remote/inference/runpod/config.py

@@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from typing import Optional
+from typing import Any, Dict, Optional
 
 from pydantic import BaseModel, Field
 
@@ -21,3 +21,10 @@ class RunpodImplConfig(BaseModel):
         default=None,
         description="The API token",
     )
+
+    @classmethod
+    def sample_run_config(cls, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "url": "${env.RUNPOD_URL:}",
+            "api_token": "${env.RUNPOD_API_TOKEN:}",
+        }

llama_stack/providers/remote/inference/runpod/runpod.py

@@ -8,7 +8,6 @@ from typing import AsyncGenerator
 from openai import OpenAI
 
 from llama_stack.apis.inference import *  # noqa: F403
-from llama_stack.models.llama.datatypes import Message
 
 # from llama_stack.providers.datatypes import ModelsProtocolPrivate
 from llama_stack.providers.utils.inference.model_registry import ModelRegistryHelper

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

@@ -1,17 +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 typing import Any
-
-from .config import SampleConfig
-
-
-async def get_adapter_impl(config: SampleConfig, _deps) -> Any:
-    from .sample import SampleInferenceImpl
-
-    impl = SampleInferenceImpl(config)
-    await impl.initialize()
-    return impl

llama_stack/providers/remote/inference/sample/config.py

@@ -1,12 +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 pydantic import BaseModel
-
-
-class SampleConfig(BaseModel):
-    host: str = "localhost"
-    port: int = 9999

llama_stack/providers/remote/inference/sample/sample.py

@@ -1,23 +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.apis.inference import Inference
-from llama_stack.apis.models import Model
-
-from .config import SampleConfig
-
-
-class SampleInferenceImpl(Inference):
-    def __init__(self, config: SampleConfig):
-        self.config = config
-
-    async def register_model(self, model: Model) -> None:
-        # these are the model names the Llama Stack will use to route requests to this provider
-        # perform validation here if necessary
-        pass
-
-    async def initialize(self):
-        pass

llama_stack/providers/remote/safety/sample/__init__.py

@@ -1,17 +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 typing import Any
-
-from .config import SampleConfig
-
-
-async def get_adapter_impl(config: SampleConfig, _deps) -> Any:
-    from .sample import SampleSafetyImpl
-
-    impl = SampleSafetyImpl(config)
-    await impl.initialize()
-    return impl

llama_stack/providers/remote/safety/sample/config.py

@@ -1,12 +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 pydantic import BaseModel
-
-
-class SampleConfig(BaseModel):
-    host: str = "localhost"
-    port: int = 9999

llama_stack/providers/remote/safety/sample/sample.py

@@ -1,23 +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.apis.safety import Safety
-from llama_stack.apis.shields import Shield
-
-from .config import SampleConfig
-
-
-class SampleSafetyImpl(Safety):
-    def __init__(self, config: SampleConfig):
-        self.config = config
-
-    async def register_shield(self, shield: Shield) -> None:
-        # these are the safety shields the Llama Stack will use to route requests to this provider
-        # perform validation here if necessary
-        pass
-
-    async def initialize(self):
-        pass

llama_stack/providers/remote/tool_runtime/bing_search/config.py

@@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from typing import Optional
+from typing import Any, Dict, Optional
 
 from pydantic import BaseModel
 
@@ -14,3 +14,9 @@ class BingSearchToolConfig(BaseModel):
 
     api_key: Optional[str] = None
     top_k: int = 3
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "api_key": "${env.BING_API_KEY:}",
+        }

llama_stack/providers/remote/tool_runtime/model_context_protocol/config.py

@@ -4,8 +4,12 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
 
 class ModelContextProtocolConfig(BaseModel):
-    pass
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {}

llama_stack/providers/remote/tool_runtime/wolfram_alpha/config.py

@@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from typing import Optional
+from typing import Any, Dict, Optional
 
 from pydantic import BaseModel
 
@@ -13,3 +13,9 @@ class WolframAlphaToolConfig(BaseModel):
     """Configuration for WolframAlpha Tool Runtime"""
 
     api_key: Optional[str] = None
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "api_key": "${env.WOLFRAM_ALPHA_API_KEY:}",
+        }

llama_stack/providers/remote/vector_io/qdrant/config.py

@@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from typing import Optional
+from typing import Any, Dict, Optional
 
 from pydantic import BaseModel
 
@@ -24,3 +24,9 @@ class QdrantVectorIOConfig(BaseModel):
     timeout: Optional[int] = None
     host: Optional[str] = None
     path: Optional[str] = None
+
+    @classmethod
+    def sample_run_config(cls, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "api_key": "${env.QDRANT_API_KEY}",
+        }

llama_stack/providers/remote/vector_io/sample/__init__.py

@@ -1,17 +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 typing import Any
-
-from .config import SampleVectorIOConfig
-
-
-async def get_adapter_impl(config: SampleVectorIOConfig, _deps) -> Any:
-    from .sample import SampleVectorIOImpl
-
-    impl = SampleVectorIOImpl(config)
-    await impl.initialize()
-    return impl

llama_stack/providers/remote/vector_io/sample/config.py

@@ -1,12 +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 pydantic import BaseModel
-
-
-class SampleVectorIOConfig(BaseModel):
-    host: str = "localhost"
-    port: int = 9999

llama_stack/providers/remote/vector_io/sample/sample.py

@@ -1,26 +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.apis.vector_dbs import VectorDB
-from llama_stack.apis.vector_io import VectorIO
-
-from .config import SampleVectorIOConfig
-
-
-class SampleVectorIOImpl(VectorIO):
-    def __init__(self, config: SampleVectorIOConfig):
-        self.config = config
-
-    async def register_vector_db(self, vector_db: VectorDB) -> None:
-        # these are the vector dbs the Llama Stack will use to route requests to this provider
-        # perform validation here if necessary
-        pass
-
-    async def initialize(self):
-        pass
-
-    async def shutdown(self):
-        pass

llama_stack/providers/remote/vector_io/weaviate/config.py

@@ -4,6 +4,8 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+from typing import Any, Dict
+
 from pydantic import BaseModel
 
 
@@ -13,4 +15,6 @@ class WeaviateRequestProviderData(BaseModel):
 
 
 class WeaviateVectorIOConfig(BaseModel):
-    pass
+    @classmethod
+    def sample_run_config(cls, **kwargs: Any) -> Dict[str, Any]:
+        return {}

llama_stack/templates/bedrock/run.yaml

@@ -45,14 +45,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/bedrock}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/bedrock}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/bedrock}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/cerebras/run.yaml

@@ -23,7 +23,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   vector_io:
   - provider_id: faiss
     provider_type: inline::faiss
@@ -43,14 +44,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/cerebras}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/cerebras}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/cerebras}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/ci-tests/run.yaml

@@ -28,7 +28,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -47,14 +48,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ci-tests}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ci-tests}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ci-tests}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

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

@@ -31,7 +31,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dell}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dell}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dell}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/dell/run.yaml

@@ -27,7 +27,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -46,14 +47,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dell}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dell}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dell}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/dev/run.yaml

@@ -57,7 +57,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -76,14 +77,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dev}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dev}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dev}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

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

@@ -56,14 +56,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/fireworks}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/fireworks}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/fireworks}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -88,7 +100,8 @@ providers:
       max_results: 3
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
   - provider_id: code-interpreter
     provider_type: inline::code-interpreter
     config: {}

llama_stack/templates/fireworks/run.yaml

@@ -31,7 +31,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/fireworks}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/fireworks}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/fireworks}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -82,7 +95,8 @@ providers:
       max_results: 3
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
   - provider_id: code-interpreter
     provider_type: inline::code-interpreter
     config: {}

llama_stack/templates/groq/run.yaml

@@ -31,7 +31,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/groq}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/groq}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/groq}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/hf-endpoint/run-with-safety.yaml

@@ -36,7 +36,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -55,14 +56,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-endpoint}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-endpoint}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-endpoint}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/hf-endpoint/run.yaml

@@ -31,7 +31,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-endpoint}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-endpoint}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-endpoint}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/hf-serverless/run-with-safety.yaml

@@ -36,7 +36,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -55,14 +56,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-serverless}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-serverless}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-serverless}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/hf-serverless/run.yaml

@@ -31,7 +31,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-serverless}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-serverless}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/hf-serverless}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/meta-reference-gpu/run-with-safety.yaml

@@ -38,7 +38,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -57,14 +58,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-gpu}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-gpu}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-gpu}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/meta-reference-gpu/run.yaml

@@ -32,7 +32,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -51,14 +52,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-gpu}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-gpu}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-gpu}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/meta-reference-quantized-gpu/run.yaml

@@ -34,7 +34,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -53,14 +54,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-quantized-gpu}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-quantized-gpu}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/meta-reference-quantized-gpu}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/nvidia/run.yaml

@@ -28,7 +28,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -47,14 +48,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/nvidia}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/nvidia}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/nvidia}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

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

@@ -49,14 +49,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -90,7 +102,8 @@ providers:
     config: {}
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
 metadata_store:
   type: sqlite
   db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/registry.db

llama_stack/templates/ollama/run.yaml

@@ -27,7 +27,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -46,14 +47,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -87,7 +100,8 @@ providers:
     config: {}
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
 metadata_store:
   type: sqlite
   db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/ollama}/registry.db

llama_stack/templates/open-benchmark/run.yaml

@@ -54,7 +54,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -73,14 +74,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/open-benchmark}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/open-benchmark}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/open-benchmark}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/remote-vllm/run-with-safety.yaml

@@ -38,7 +38,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -98,7 +111,8 @@ providers:
     config: {}
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
 metadata_store:
   type: sqlite
   db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/registry.db

llama_stack/templates/remote-vllm/run.yaml

@@ -32,7 +32,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -44,14 +45,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -92,7 +105,8 @@ providers:
     config: {}
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
 metadata_store:
   type: sqlite
   db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/registry.db

llama_stack/templates/sambanova/run.yaml

@@ -37,7 +37,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference

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

@@ -31,7 +31,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/tgi}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/tgi}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/tgi}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

llama_stack/templates/tgi/run.yaml

@@ -30,7 +30,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -49,14 +50,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/tgi}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/tgi}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/tgi}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

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

@@ -56,14 +56,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -97,7 +109,8 @@ providers:
     config: {}
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
 metadata_store:
   type: sqlite
   db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/registry.db

llama_stack/templates/together/run.yaml

@@ -31,7 +31,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -50,14 +51,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -91,7 +104,8 @@ providers:
     config: {}
   - provider_id: wolfram-alpha
     provider_type: remote::wolfram-alpha
-    config: {}
+    config:
+      api_key: ${env.WOLFRAM_ALPHA_API_KEY:}
 metadata_store:
   type: sqlite
   db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/together}/registry.db

llama_stack/templates/vllm-gpu/run.yaml

@@ -35,7 +35,8 @@ providers:
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
-    config: {}
+    config:
+      excluded_categories: []
   agents:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
@@ -54,14 +55,26 @@ providers:
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/vllm-gpu}/meta_reference_eval.db
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/vllm-gpu}/huggingface_datasetio.db
   - provider_id: localfs
     provider_type: inline::localfs
-    config: {}
+    config:
+      kvstore:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/vllm-gpu}/localfs_datasetio.db
   scoring:
   - provider_id: basic
     provider_type: inline::basic

tests/integration/agents/test_agents.py

@@ -186,7 +186,7 @@ def test_builtin_tool_web_search(llama_stack_client_with_mocked_inference, agent
         messages=[
             {
                 "role": "user",
-                "content": "Search the web and tell me who the current CEO of Meta is.",
+                "content": "Search the web and tell me who the founder of Meta is.",
             }
         ],
         session_id=session_id,

tests/integration/conftest.py

@@ -6,12 +6,17 @@
 import inspect
 import itertools
 import os
+import platform
 import textwrap
 
 from dotenv import load_dotenv
 
+from llama_stack.log import get_logger
+
 from .report import Report
 
+logger = get_logger(__name__, category="tests")
+
 
 def pytest_configure(config):
     config.option.tbstyle = "short"
@@ -24,6 +29,10 @@ def pytest_configure(config):
         key, value = env_var.split("=", 1)
         os.environ[key] = value
 
+    if platform.system() == "Darwin":  # Darwin is the system name for macOS
+        os.environ["DISABLE_CODE_SANDBOX"] = "1"
+        logger.info("Setting DISABLE_CODE_SANDBOX=1 for macOS")
+
     if config.getoption("--report"):
         config.pluginmanager.register(Report(config))
 

tests/integration/fixtures/recorded_responses/invoke_tool.json

@@ -413,23 +413,23 @@
             "type": "text"
           },
           {
-            "text": "Result 1:\nDocument_id:3e31d\nContent:  conversational data, :func:`~torchtune.datasets.chat_dataset` seems to be a good fit. For any\ncustom local dataset we always need to specify ``source``, ``data_files``, and ``split`` for any dataset\nbuilder in torchtune. For :func:`~torchtune.datasets.chat_dataset`, we additionally need to specify\n``conversation_column`` and ``conversation_style``. Our data follows the ``\"sharegpt\"`` format, so\nwe can specify that here. Altogether, our :func:`~torchtune.datasets.chat_dataset` call should\nlook like so:\n\n.. code-block:: python\n\n    from torchtune.datasets import chat_dataset\n    from torchtune.models.llama3 import llama3_tokenizer\n\n    tokenizer = llama3_tokenizer(\"/tmp/Meta-Llama-3-8B-Instruct/original/tokenizer.model\")\n    ds = chat_dataset(\n        tokenizer=tokenizer,\n        source=\"json\",\n        data_files=\"data/my_data.json\",\n        split=\"train\",\n        conversation_column=\"dialogue\",\n        conversation_style=\"sharegpt\",\n    )\n\n.. code-block:: yaml\n\n    # In config\n    tokenizer:\n      _component_: torchtune.models.llama3.llama3_tokenizer\n      path: /tmp/Meta-Llama-3-8B-Instruct/original/tokenizer.model\n\n    dataset:\n      _component_: torchtune.datasets.chat_dataset\n      source: json\n      data_files: data/my_data.json\n      split: train\n      conversation_column: dialogue\n      conversation_style: sharegpt\n\n.. note::\n    You can pass in any keyword argument for `load_dataset <https://huggingface.co/docs/datasets/v2.20.0/en/package_reference/loading_methods#datasets.load_dataset>`_ into all our\n    Dataset classes and they will honor them. This is useful for common parameters\n    such as specifying the data split with :code:`split` or configuration with\n    :code:`name`\n\nIf you needed to add a prompt template, you would simply pass it into the tokenizer.\nSince we're fine-tuning Llama3, the tokenizer will handle all formatting for\nus and prompt templates are optional. Other models such as Mistral's :class:`~torchtune.models.mistral._tokenizer.MistralTokenizer`,\nuse a chat template by default (:class:`~torchtune.models.mistral.MistralChatTemplate`) to format\nall messages according to their `recommendations <https://\n",
+            "text": "Result 1:\nDocument_id:bbddb\nContent:  conversational data, :func:`~torchtune.datasets.chat_dataset` seems to be a good fit. For any\ncustom local dataset we always need to specify ``source``, ``data_files``, and ``split`` for any dataset\nbuilder in torchtune. For :func:`~torchtune.datasets.chat_dataset`, we additionally need to specify\n``conversation_column`` and ``conversation_style``. Our data follows the ``\"sharegpt\"`` format, so\nwe can specify that here. Altogether, our :func:`~torchtune.datasets.chat_dataset` call should\nlook like so:\n\n.. code-block:: python\n\n    from torchtune.datasets import chat_dataset\n    from torchtune.models.llama3 import llama3_tokenizer\n\n    tokenizer = llama3_tokenizer(\"/tmp/Meta-Llama-3-8B-Instruct/original/tokenizer.model\")\n    ds = chat_dataset(\n        tokenizer=tokenizer,\n        source=\"json\",\n        data_files=\"data/my_data.json\",\n        split=\"train\",\n        conversation_column=\"dialogue\",\n        conversation_style=\"sharegpt\",\n    )\n\n.. code-block:: yaml\n\n    # In config\n    tokenizer:\n      _component_: torchtune.models.llama3.llama3_tokenizer\n      path: /tmp/Meta-Llama-3-8B-Instruct/original/tokenizer.model\n\n    dataset:\n      _component_: torchtune.datasets.chat_dataset\n      source: json\n      data_files: data/my_data.json\n      split: train\n      conversation_column: dialogue\n      conversation_style: sharegpt\n\n.. note::\n    You can pass in any keyword argument for `load_dataset <https://huggingface.co/docs/datasets/v2.20.0/en/package_reference/loading_methods#datasets.load_dataset>`_ into all our\n    Dataset classes and they will honor them. This is useful for common parameters\n    such as specifying the data split with :code:`split` or configuration with\n    :code:`name`\n\nIf you needed to add a prompt template, you would simply pass it into the tokenizer.\nSince we're fine-tuning Llama3, the tokenizer will handle all formatting for\nus and prompt templates are optional. Other models such as Mistral's :class:`~torchtune.models.mistral._tokenizer.MistralTokenizer`,\nuse a chat template by default (:class:`~torchtune.models.mistral.MistralChatTemplate`) to format\nall messages according to their `recommendations <https://\n",
             "type": "text"
           },
           {
-            "text": "Result 2:\nDocument_id:14c27\nContent: .. _lora_finetune_label:\n\n============================\nFine-Tuning Llama2 with LoRA\n============================\n\nThis guide will teach you about `LoRA <https://arxiv.org/abs/2106.09685>`_, a parameter-efficient finetuning technique,\nand show you how you can use torchtune to finetune a Llama2 model with LoRA.\nIf you already know what LoRA is and want to get straight to running\nyour own LoRA finetune in torchtune, you can jump to :ref:`LoRA finetuning recipe in torchtune<lora_recipe_label>`.\n\n.. grid:: 2\n\n    .. grid-item-card:: :octicon:`mortar-board;1em;` What you will learn\n\n      * What LoRA is and how it saves memory during finetuning\n      * An overview of LoRA components in torchtune\n      * How to run a LoRA finetune using torchtune\n      * How to experiment with different LoRA configurations\n\n    .. grid-item-card:: :octicon:`list-unordered;1em;` Prerequisites\n\n      * Be familiar with :ref:`torchtune<overview_label>`\n      * Make sure to :ref:`install torchtune<install_label>`\n      * Make sure you have downloaded the :ref:`Llama2-7B model weights<download_llama_label>`\n\nWhat is LoRA?\n-------------\n\n`LoRA <https://arxiv.org/abs/2106.09685>`_ is an adapter-based method for\nparameter-efficient finetuning that adds trainable low-rank decomposition matrices to different layers of a neural network,\nthen freezes the network's remaining parameters. LoRA is most commonly applied to\ntransformer models, in which case it is common to add the low-rank matrices\nto some of the linear projections in each transformer layer's self-attention.\n\n.. note::\n\n    If you're unfamiliar, check out these references for the `definition of rank <https://en.wikipedia.org/wiki/Rank_(linear_algebra)>`_\n    and discussion of `low-rank approximations <https://en.wikipedia.org/wiki/Low-rank_approximation>`_.\n\nBy finetuning with LoRA (as opposed to finetuning all model parameters),\nyou can expect to see memory savings due to a substantial reduction in the\nnumber of parameters with gradients. When using an optimizer with momentum,\nlike `AdamW <https://py\n",
+            "text": "Result 2:\nDocument_id:15b86\nContent: .. _lora_finetune_label:\n\n============================\nFine-Tuning Llama2 with LoRA\n============================\n\nThis guide will teach you about `LoRA <https://arxiv.org/abs/2106.09685>`_, a parameter-efficient finetuning technique,\nand show you how you can use torchtune to finetune a Llama2 model with LoRA.\nIf you already know what LoRA is and want to get straight to running\nyour own LoRA finetune in torchtune, you can jump to :ref:`LoRA finetuning recipe in torchtune<lora_recipe_label>`.\n\n.. grid:: 2\n\n    .. grid-item-card:: :octicon:`mortar-board;1em;` What you will learn\n\n      * What LoRA is and how it saves memory during finetuning\n      * An overview of LoRA components in torchtune\n      * How to run a LoRA finetune using torchtune\n      * How to experiment with different LoRA configurations\n\n    .. grid-item-card:: :octicon:`list-unordered;1em;` Prerequisites\n\n      * Be familiar with :ref:`torchtune<overview_label>`\n      * Make sure to :ref:`install torchtune<install_label>`\n      * Make sure you have downloaded the :ref:`Llama2-7B model weights<download_llama_label>`\n\nWhat is LoRA?\n-------------\n\n`LoRA <https://arxiv.org/abs/2106.09685>`_ is an adapter-based method for\nparameter-efficient finetuning that adds trainable low-rank decomposition matrices to different layers of a neural network,\nthen freezes the network's remaining parameters. LoRA is most commonly applied to\ntransformer models, in which case it is common to add the low-rank matrices\nto some of the linear projections in each transformer layer's self-attention.\n\n.. note::\n\n    If you're unfamiliar, check out these references for the `definition of rank <https://en.wikipedia.org/wiki/Rank_(linear_algebra)>`_\n    and discussion of `low-rank approximations <https://en.wikipedia.org/wiki/Low-rank_approximation>`_.\n\nBy finetuning with LoRA (as opposed to finetuning all model parameters),\nyou can expect to see memory savings due to a substantial reduction in the\nnumber of parameters with gradients. When using an optimizer with momentum,\nlike `AdamW <https://py\n",
             "type": "text"
           },
           {
-            "text": "Result 3:\nDocument_id:90a49\nContent: ` module, which we swap\n   out for :class:`~torchtune.modules.peft.LoRALinear` when ``use_dora=True``.\n\n.. _glossary_distrib:\n\n\n.. TODO\n\n.. Distributed\n.. -----------\n\n.. .. _glossary_fsdp:\n\n.. Fully Sharded Data Parallel (FSDP)\n.. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n.. All our ``_distributed`` recipes use `FSDP <https://pytorch.org/docs/stable/fsdp.html>`.\n.. .. _glossary_fsdp2:\n\n",
+            "text": "Result 3:\nDocument_id:83901\nContent: ` module, which we swap\n   out for :class:`~torchtune.modules.peft.LoRALinear` when ``use_dora=True``.\n\n.. _glossary_distrib:\n\n\n.. TODO\n\n.. Distributed\n.. -----------\n\n.. .. _glossary_fsdp:\n\n.. Fully Sharded Data Parallel (FSDP)\n.. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n.. All our ``_distributed`` recipes use `FSDP <https://pytorch.org/docs/stable/fsdp.html>`.\n.. .. _glossary_fsdp2:\n\n",
             "type": "text"
           },
           {
-            "text": "Result 4:\nDocument_id:14c27\nContent: 06% of all params are trainable.\n\n.. note::\n    If you are directly using the LoRA recipe (as detailed :ref:`here<lora_recipe_label>`), you need only pass the\n    relevant checkpoint path. Loading model weights and setting trainable parameters will be taken care\n    of in the recipe.\n\n\n.. _lora_recipe_label:\n\nLoRA finetuning recipe in torchtune\n-----------------------------------\n\nFinally, we can put it all together and finetune a model using torchtune's `LoRA recipe <https://github.com/pytorch/torchtune/blob/48626d19d2108f92c749411fbd5f0ff140023a25/recipes/lora_finetune.py>`_.\nMake sure that you have first downloaded the Llama2 weights and tokenizer by following :ref:`these instructions<download_llama_label>`.\nYou can then run the following command to perform a LoRA finetune of Llama2-7B with two GPUs (each having VRAM of at least 16GB):\n\n.. code-block:: bash\n\n    tune run --nnodes 1 --nproc_per_node 2 lora_finetune_distributed --config llama2/7B_lora\n\n.. note::\n    Make sure to point to the location of your Llama2 weights and tokenizer. This can be done\n    either by adding :code:`checkpointer.checkpoint_files=[my_model_checkpoint_path] tokenizer_checkpoint=my_tokenizer_checkpoint_path`\n    or by directly modifying the :code:`7B_lora.yaml` file. See our \"\":ref:`config_tutorial_label`\" recipe\n    for more details on how you can easily clone and modify torchtune configs.\n\n.. note::\n    You can modify the value of :code:`nproc_per_node` depending on (a) the number of GPUs you have available,\n    and (b) the memory constraints of your hardware.\n\nThe preceding command will run a LoRA finetune with torchtune's factory settings, but we may want to experiment a bit.\nLet's take a closer look at some of the :code:`lora_finetune_distributed` config.\n\n.. code-block:: yaml\n\n  # Model Arguments\n  model:\n    _component_: lora_llama2_7b\n    lora_attn_modules: ['q_proj', 'v_proj']\n    lora_rank: 8\n    lora_alpha: 16\n  ...\n\nWe see that the\n",
+            "text": "Result 4:\nDocument_id:15b86\nContent: 06% of all params are trainable.\n\n.. note::\n    If you are directly using the LoRA recipe (as detailed :ref:`here<lora_recipe_label>`), you need only pass the\n    relevant checkpoint path. Loading model weights and setting trainable parameters will be taken care\n    of in the recipe.\n\n\n.. _lora_recipe_label:\n\nLoRA finetuning recipe in torchtune\n-----------------------------------\n\nFinally, we can put it all together and finetune a model using torchtune's `LoRA recipe <https://github.com/pytorch/torchtune/blob/48626d19d2108f92c749411fbd5f0ff140023a25/recipes/lora_finetune.py>`_.\nMake sure that you have first downloaded the Llama2 weights and tokenizer by following :ref:`these instructions<download_llama_label>`.\nYou can then run the following command to perform a LoRA finetune of Llama2-7B with two GPUs (each having VRAM of at least 16GB):\n\n.. code-block:: bash\n\n    tune run --nnodes 1 --nproc_per_node 2 lora_finetune_distributed --config llama2/7B_lora\n\n.. note::\n    Make sure to point to the location of your Llama2 weights and tokenizer. This can be done\n    either by adding :code:`checkpointer.checkpoint_files=[my_model_checkpoint_path] tokenizer_checkpoint=my_tokenizer_checkpoint_path`\n    or by directly modifying the :code:`7B_lora.yaml` file. See our \"\":ref:`config_tutorial_label`\" recipe\n    for more details on how you can easily clone and modify torchtune configs.\n\n.. note::\n    You can modify the value of :code:`nproc_per_node` depending on (a) the number of GPUs you have available,\n    and (b) the memory constraints of your hardware.\n\nThe preceding command will run a LoRA finetune with torchtune's factory settings, but we may want to experiment a bit.\nLet's take a closer look at some of the :code:`lora_finetune_distributed` config.\n\n.. code-block:: yaml\n\n  # Model Arguments\n  model:\n    _component_: lora_llama2_7b\n    lora_attn_modules: ['q_proj', 'v_proj']\n    lora_rank: 8\n    lora_alpha: 16\n  ...\n\nWe see that the\n",
             "type": "text"
           },
           {
-            "text": "Result 5:\nDocument_id:90a49\nContent: etune\n:func:`torchtune.models.llama3.llama3_8b` with DoRA, you would use :func:`torchtune.models.llama3.lora_llama3_8b` with ``use_dora=True``:\n\n.. code-block:: bash\n\n  tune run lora_finetune_single_device --config llama3/8B_lora_single_device \\\n  model.use_dora=True\n\n.. code-block:: yaml\n\n  model:\n    _component_: torchtune.models.lora_llama3_8b\n    use_dora: True\n\nSince DoRA extends LoRA, the parameters for :ref:`customizing LoRA <glossary_lora>` are identical. You can also quantize the base model weights like in :ref:`glossary_qlora` by using ``quantize=True`` to reap\neven more memory savings!\n\n.. code-block:: bash\n\n  tune run lora_finetune_single_device --config llama3/8B_lora_single_device \\\n  model.apply_lora_to_mlp=True \\\n  model.lora_attn_modules=[\"q_proj\",\"k_proj\",\"v_proj\"] \\\n  model.lora_rank=16 \\\n  model.lora_alpha=32 \\\n  model.use_dora=True \\\n  model.quantize_base=True\n\n.. code-block:: yaml\n\n  model:\n    _component_: torchtune.models.lora_llama3_8b\n    apply_lora_to_mlp: True\n    lora_attn_modules: [\"q_proj\", \"k_proj\", \"v_proj\"]\n    lora_rank: 16\n    lora_alpha: 32\n    use_dora: True\n    quantize_base: True\n\n\n.. note::\n\n   Under the hood, we've enabled DoRA by adding the :class:`~torchtune.modules.peft.DoRALinear` module, which we swap\n   out for :class:`~torchtune.modules.peft.LoRALinear` when ``use_dora=True``.\n\n.. _glossary_distrib:\n\n\n.. TODO\n\n.. Distributed\n.. -----------\n\n.. .. _glossary_fsdp:\n\n.. Fully Sharded Data Parallel (FSDP)\n.. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n.. All our ``_distributed`` recipes use `FSDP <https://pytorch.org/docs/stable/fsdp.html>`.\n.. .. _glossary_fsdp2:\n\n",
+            "text": "Result 5:\nDocument_id:83901\nContent: etune\n:func:`torchtune.models.llama3.llama3_8b` with DoRA, you would use :func:`torchtune.models.llama3.lora_llama3_8b` with ``use_dora=True``:\n\n.. code-block:: bash\n\n  tune run lora_finetune_single_device --config llama3/8B_lora_single_device \\\n  model.use_dora=True\n\n.. code-block:: yaml\n\n  model:\n    _component_: torchtune.models.lora_llama3_8b\n    use_dora: True\n\nSince DoRA extends LoRA, the parameters for :ref:`customizing LoRA <glossary_lora>` are identical. You can also quantize the base model weights like in :ref:`glossary_qlora` by using ``quantize=True`` to reap\neven more memory savings!\n\n.. code-block:: bash\n\n  tune run lora_finetune_single_device --config llama3/8B_lora_single_device \\\n  model.apply_lora_to_mlp=True \\\n  model.lora_attn_modules=[\"q_proj\",\"k_proj\",\"v_proj\"] \\\n  model.lora_rank=16 \\\n  model.lora_alpha=32 \\\n  model.use_dora=True \\\n  model.quantize_base=True\n\n.. code-block:: yaml\n\n  model:\n    _component_: torchtune.models.lora_llama3_8b\n    apply_lora_to_mlp: True\n    lora_attn_modules: [\"q_proj\", \"k_proj\", \"v_proj\"]\n    lora_rank: 16\n    lora_alpha: 32\n    use_dora: True\n    quantize_base: True\n\n\n.. note::\n\n   Under the hood, we've enabled DoRA by adding the :class:`~torchtune.modules.peft.DoRALinear` module, which we swap\n   out for :class:`~torchtune.modules.peft.LoRALinear` when ``use_dora=True``.\n\n.. _glossary_distrib:\n\n\n.. TODO\n\n.. Distributed\n.. -----------\n\n.. .. _glossary_fsdp:\n\n.. Fully Sharded Data Parallel (FSDP)\n.. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n.. All our ``_distributed`` recipes use `FSDP <https://pytorch.org/docs/stable/fsdp.html>`.\n.. .. _glossary_fsdp2:\n\n",
             "type": "text"
           },
           {
@@ -441,11 +441,11 @@
         "error_message": null,
         "metadata": {
           "document_ids": [
-            "3e31d46a-5568-49d4-978e-0ab3b6598a7f",
+            "bbddbe62-508d-4c8d-9455-3b60bc2825a5",
-            "14c2766f-d176-468d-9124-f4a46c5166cf",
+            "15b8638f-b1b6-4f58-adfa-eb6644c47de3",
-            "90a49271-c246-4bdc-9db9-e3bfa9be5ff1",
+            "83901b53-33d4-4f5e-8145-b94c783e9f61",
-            "14c2766f-d176-468d-9124-f4a46c5166cf",
+            "15b8638f-b1b6-4f58-adfa-eb6644c47de3",
-            "90a49271-c246-4bdc-9db9-e3bfa9be5ff1"
+            "83901b53-33d4-4f5e-8145-b94c783e9f61"
           ]
         }
       }

tests/unit/providers/test_configs.py

@@ -0,0 +1,50 @@
+# 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.
+
+import pytest
+from pydantic import BaseModel
+
+from llama_stack.distribution.distribution import get_provider_registry, providable_apis
+from llama_stack.distribution.utils.dynamic import instantiate_class_type
+
+
+class TestProviderConfigurations:
+    """Test suite for testing provider configurations across all API types."""
+
+    def test_all_api_providers_exist(self):
+        provider_registry = get_provider_registry()
+        for api in providable_apis():
+            providers = provider_registry.get(api, {})
+            assert providers, f"No providers found for API type: {api}"
+
+    @pytest.mark.parametrize("api", providable_apis())
+    def test_api_providers(self, api):
+        provider_registry = get_provider_registry()
+        providers = provider_registry.get(api, {})
+        assert providers, f"No providers found for API type: {api}"
+
+        failures = []
+        for provider_type, provider_spec in providers.items():
+            try:
+                self._verify_provider_config(provider_type, provider_spec)
+            except Exception as e:
+                failures.append(f"Failed to verify {provider_type} config: {str(e)}")
+
+        if failures:
+            pytest.fail("\n".join(failures))
+
+    def _verify_provider_config(self, provider_type, provider_spec):
+        """Helper method to verify a single provider configuration."""
+        # Get the config class
+        config_class_name = provider_spec.config_class
+        config_type = instantiate_class_type(config_class_name)
+
+        assert issubclass(config_type, BaseModel), f"{config_class_name} is not a subclass of BaseModel"
+
+        assert hasattr(config_type, "sample_run_config"), f"{config_class_name} does not have sample_run_config method"
+
+        sample_config = config_type.sample_run_config(__distro_dir__="foobarbaz")
+        assert isinstance(sample_config, dict), f"{config_class_name}.sample_run_config() did not return a dict"