From 436f8ade9e7be9ce5116c0976a0991b7e355f533 Mon Sep 17 00:00:00 2001 From: Charlie Doern Date: Fri, 4 Apr 2025 15:53:21 -0400 Subject: [PATCH 1/2] feat: implement provider updating add `v1/providers/` which uses PUT to allow users to change their provider configuration this is a follow up to #1429 and related to #1359 a user can call something like: `llama_stack_client.providers.update(api="inference", provider_id="ollama", provider_type="remote::ollama", config={'url': 'http:/localhost:12345'})` or `llama-stack-client providers update inference ollama remote::ollama "{'url': 'http://localhost:12345'}"` this API works by adding a `RequestMiddleware` to the server which checks requests, and if the user is using PUT /v1/providers, the routes are re-registered with the re-initialized provider configurations/methods for the client, `self.impls` is updated to hold the proper methods+configurations this depends on a client PR, the CI will fail until then but succeeded locally Signed-off-by: Charlie Doern --- docs/_static/llama-stack-spec.html | 103 +++++++++ docs/_static/llama-stack-spec.yaml | 61 +++++ llama_stack/apis/providers/providers.py | 5 + llama_stack/distribution/library_client.py | 17 ++ llama_stack/distribution/providers.py | 86 ++++++- llama_stack/distribution/server/routes.py | 2 + llama_stack/distribution/server/server.py | 214 +++++++++++++----- tests/integration/providers/test_providers.py | 17 ++ 8 files changed, 449 insertions(+), 56 deletions(-) diff --git a/docs/_static/llama-stack-spec.html b/docs/_static/llama-stack-spec.html index ae9ad5d4c..eb31e634a 100644 --- a/docs/_static/llama-stack-spec.html +++ b/docs/_static/llama-stack-spec.html @@ -4946,6 +4946,74 @@ } } }, + "/v1/providers/{api}/{provider_id}/{provider_type}": { + "post": { + "responses": { + "200": { + "description": "OK", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ProviderInfo" + } + } + } + }, + "400": { + "$ref": "#/components/responses/BadRequest400" + }, + "429": { + "$ref": "#/components/responses/TooManyRequests429" + }, + "500": { + "$ref": "#/components/responses/InternalServerError500" + }, + "default": { + "$ref": "#/components/responses/DefaultError" + } + }, + "tags": [ + "Providers" + ], + "description": "", + "parameters": [ + { + "name": "api", + "in": "path", + "required": true, + "schema": { + "type": "string" + } + }, + { + "name": "provider_id", + "in": "path", + "required": true, + "schema": { + "type": "string" + } + }, + { + "name": "provider_type", + "in": "path", + "required": true, + "schema": { + "type": "string" + } + } + ], + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/UpdateProviderRequest" + } + } + }, + "required": true + } + } + }, "/v1/version": { "get": { "responses": { @@ -16101,6 +16169,41 @@ "title": "SyntheticDataGenerationResponse", "description": "Response from the synthetic data generation. Batch of (prompt, response, score) tuples that pass the threshold." }, + "UpdateProviderRequest": { + "type": "object", + "properties": { + "config": { + "type": "object", + "additionalProperties": { + "oneOf": [ + { + "type": "null" + }, + { + "type": "boolean" + }, + { + "type": "number" + }, + { + "type": "string" + }, + { + "type": "array" + }, + { + "type": "object" + } + ] + } + } + }, + "additionalProperties": false, + "required": [ + "config" + ], + "title": "UpdateProviderRequest" + }, "VersionInfo": { "type": "object", "properties": { diff --git a/docs/_static/llama-stack-spec.yaml b/docs/_static/llama-stack-spec.yaml index 48cefe12b..3c4e9c865 100644 --- a/docs/_static/llama-stack-spec.yaml +++ b/docs/_static/llama-stack-spec.yaml @@ -3484,6 +3484,50 @@ paths: schema: $ref: '#/components/schemas/SyntheticDataGenerateRequest' required: true + /v1/providers/{api}/{provider_id}/{provider_type}: + post: + responses: + '200': + description: OK + content: + application/json: + schema: + $ref: '#/components/schemas/ProviderInfo' + '400': + $ref: '#/components/responses/BadRequest400' + '429': + $ref: >- + #/components/responses/TooManyRequests429 + '500': + $ref: >- + #/components/responses/InternalServerError500 + default: + $ref: '#/components/responses/DefaultError' + tags: + - Providers + description: '' + parameters: + - name: api + in: path + required: true + schema: + type: string + - name: provider_id + in: path + required: true + schema: + type: string + - name: provider_type + in: path + required: true + schema: + type: string + requestBody: + content: + application/json: + schema: + $ref: '#/components/schemas/UpdateProviderRequest' + required: true /v1/version: get: responses: @@ -11234,6 +11278,23 @@ components: description: >- Response from the synthetic data generation. Batch of (prompt, response, score) tuples that pass the threshold. + UpdateProviderRequest: + type: object + properties: + config: + type: object + additionalProperties: + oneOf: + - type: 'null' + - type: boolean + - type: number + - type: string + - type: array + - type: object + additionalProperties: false + required: + - config + title: UpdateProviderRequest VersionInfo: type: object properties: diff --git a/llama_stack/apis/providers/providers.py b/llama_stack/apis/providers/providers.py index 4bc977bf1..0f29a6e05 100644 --- a/llama_stack/apis/providers/providers.py +++ b/llama_stack/apis/providers/providers.py @@ -47,3 +47,8 @@ class Providers(Protocol): :returns: A ProviderInfo object containing the provider's details. """ ... + + @webmethod(route="/providers/{api}/{provider_id}/{provider_type}", method="PUT") + async def update_provider( + self, api: str, provider_id: str, provider_type: str, config: dict[str, Any] + ) -> ProviderInfo: ... diff --git a/llama_stack/distribution/library_client.py b/llama_stack/distribution/library_client.py index cebfabba5..5a59706ac 100644 --- a/llama_stack/distribution/library_client.py +++ b/llama_stack/distribution/library_client.py @@ -25,6 +25,7 @@ from llama_stack_client import ( AsyncStream, LlamaStackClient, ) +from llama_stack_client.types import provider_info from pydantic import BaseModel, TypeAdapter from rich.console import Console from termcolor import cprint @@ -293,6 +294,22 @@ class AsyncLlamaStackAsLibraryClient(AsyncLlamaStackClient): cast_to=cast_to, options=options, ) + # Check if response is of a certain type + # this indicates we have done a provider update + if ( + isinstance(response, provider_info.ProviderInfo) + and hasattr(response, "config") + and options.method.lower() == "put" + ): + # patch in the new provider config + for api, providers in self.config.providers.items(): + if api != response.api: + continue + for prov in providers: + if prov.provider_id == response.provider_id: + prov.config = response.config + break + await self.initialize() return response async def _call_non_streaming( diff --git a/llama_stack/distribution/providers.py b/llama_stack/distribution/providers.py index 1d9c1f4e9..3fc61df73 100644 --- a/llama_stack/distribution/providers.py +++ b/llama_stack/distribution/providers.py @@ -5,6 +5,7 @@ # the root directory of this source tree. import asyncio +import copy from typing import Any from pydantic import BaseModel @@ -13,7 +14,7 @@ from llama_stack.apis.providers import ListProvidersResponse, ProviderInfo, Prov from llama_stack.log import get_logger from llama_stack.providers.datatypes import HealthResponse, HealthStatus -from .datatypes import StackRunConfig +from .datatypes import Provider, StackRunConfig from .utils.config import redact_sensitive_fields logger = get_logger(name=__name__, category="core") @@ -129,3 +130,86 @@ class ProviderImpl(Providers): providers_health[api_name] = health_response return providers_health + + async def update_provider( + self, api: str, provider_id: str, provider_type: str, config: dict[str, Any] + ) -> ProviderInfo: + # config = ast.literal_eval(provider_request.config) + prov = Provider( + provider_id=provider_id, + provider_type=provider_type, + config=config, + ) + assert prov.provider_id is not None + existing_provider = None + # if the provider isn't there or the API is invalid, we should not continue + for prov_api, providers in self.config.run_config.providers.items(): + if prov_api != api: + continue + for p in providers: + if p.provider_id == provider_id: + existing_provider = p + break + if existing_provider is not None: + break + + if existing_provider is None: + raise ValueError(f"Provider {provider_id} not found, you can only update already registered providers.") + + new_config = self.merge_providers(existing_provider, prov) + existing_provider.config = new_config + providers_health = await self.get_providers_health() + # takes a single provider, validates its in the registry + # if it is, merge the provider config with the existing one + ret = ProviderInfo( + api=api, + provider_id=prov.provider_id, + provider_type=prov.provider_type, + config=new_config, + health=providers_health.get(api, {}).get( + p.provider_id, + HealthResponse(status=HealthStatus.NOT_IMPLEMENTED, message="Provider does not implement health check"), + ), + ) + + return ret + + def merge_dicts(self, base: dict[str, Any], overrides: dict[str, Any]) -> dict[str, Any]: + """Recursively merges `overrides` into `base`, replacing only specified keys.""" + + merged = copy.deepcopy(base) # Preserve original dict + for key, value in overrides.items(): + if isinstance(value, dict) and isinstance(merged.get(key), dict): + # Recursively merge if both are dictionaries + merged[key] = self.merge_dicts(merged[key], value) + else: + # Otherwise, directly override + merged[key] = value + + return merged + + def merge_configs( + self, global_config: dict[str, list[Provider]], new_config: dict[str, list[Provider]] + ) -> dict[str, list[Provider]]: + merged_config = copy.deepcopy(global_config) # Preserve original structure + + for key, new_providers in new_config.items(): + if key in merged_config: + existing_providers = {p.provider_id: p for p in merged_config[key]} + + for new_provider in new_providers: + if new_provider.provider_id in existing_providers: + # Override settings of existing provider + existing = existing_providers[new_provider.provider_id] + existing.config = self.merge_dicts(existing.config, new_provider.config) + else: + # Append new provider + merged_config[key].append(new_provider) + else: + # Add new category entirely + merged_config[key] = new_providers + + return merged_config + + def merge_providers(self, current_provider: Provider, new_provider: Provider) -> dict[str, Any]: + return self.merge_dicts(current_provider.config, new_provider.config) diff --git a/llama_stack/distribution/server/routes.py b/llama_stack/distribution/server/routes.py index ea66fec5a..ee059a4a6 100644 --- a/llama_stack/distribution/server/routes.py +++ b/llama_stack/distribution/server/routes.py @@ -62,6 +62,8 @@ def get_all_api_routes() -> dict[Api, list[Route]]: http_method = hdrs.METH_GET elif webmethod.method == hdrs.METH_DELETE: http_method = hdrs.METH_DELETE + elif webmethod.method == hdrs.METH_PUT: + http_method = hdrs.METH_PUT else: http_method = hdrs.METH_POST routes.append( diff --git a/llama_stack/distribution/server/server.py b/llama_stack/distribution/server/server.py index 83407a25f..7a1df8fa1 100644 --- a/llama_stack/distribution/server/server.py +++ b/llama_stack/distribution/server/server.py @@ -27,8 +27,10 @@ from fastapi import Body, FastAPI, HTTPException, Request from fastapi import Path as FastapiPath from fastapi.exceptions import RequestValidationError from fastapi.responses import JSONResponse, StreamingResponse +from fastapi.routing import APIRoute from openai import BadRequestError from pydantic import BaseModel, ValidationError +from starlette.types import Message from llama_stack.apis.common.responses import PaginatedResponse from llama_stack.distribution.datatypes import AuthenticationRequiredError, LoggingConfig, StackRunConfig @@ -269,6 +271,84 @@ def create_dynamic_typed_route(func: Any, method: str, route: str) -> Callable: return route_handler +class RequestMiddleware: + def __init__(self, app, api, stack_run_config): + self.app = app + self.api = api + self.stack_run_config = stack_run_config + + async def __call__(self, scope, receive, send): + import json + + from fastapi import Request + + from llama_stack.apis.providers import ProviderInfo + + # from llama_stack.stack_utils import construct_stack # or wherever you define it + if scope["type"] != "http": + return await self.app(scope, receive, send) + + request = Request(scope, receive) + method = request.method + path = request.url.path + + # Only intercept PUT /v1/providers/update + if method == "PUT" and "/v1/providers" in path: + # Clone the request body so FastAPI doesn't break + + body = await request.body() + request = Request(scope, receive_from_body(body)) + + response_body = b"" + status_code = 500 + headers = [] + + async def send_wrapper(message: Message): + nonlocal response_body, status_code, headers + + if message["type"] == "http.response.start": + status_code = message["status"] + headers = message.get("headers", []) + + elif message["type"] == "http.response.body": + response_body += message.get("body", b"") + + if not message.get("more_body", False): + # Rebuild stack + try: + # Parse the request body (not response) + payload = json.loads(response_body.decode("utf-8")) + new_provider = ProviderInfo(**payload) + for api, providers in self.stack_run_config.providers.items(): + if api != new_provider.api: + continue + for prov in providers: + if prov.provider_id == new_provider.provider_id: + prov.config = new_provider.config + break + + _, impls = await construct(app=self.api, config=self.stack_run_config, reconstruct=True) + self.api.__llama_stack_impls__ = impls + print("✅ Stack rebuilt and updated.") + except Exception as e: + print(f"⚠️ Failed to rebuild stack: {e}") + + await send(message) + + return await self.app(scope, request.receive, send_wrapper) + + # All other requests go through normally + return await self.app(scope, receive, send) + + +# Helper to inject the saved body back into the request +def receive_from_body(body: bytes): + async def receive() -> Message: + return {"type": "http.request", "body": body, "more_body": False} + + return receive + + class TracingMiddleware: def __init__(self, app, impls): self.app = app @@ -482,67 +562,15 @@ def main(args: argparse.Namespace | None = None): window_seconds=window_seconds, ) - try: - impls = asyncio.run(construct_stack(config)) - except InvalidProviderError as e: - logger.error(f"Error: {str(e)}") - sys.exit(1) - - if Api.telemetry in impls: - setup_logger(impls[Api.telemetry]) - else: - setup_logger(TelemetryAdapter(TelemetryConfig(), {})) - - all_routes = get_all_api_routes() - - if config.apis: - apis_to_serve = set(config.apis) - else: - apis_to_serve = set(impls.keys()) - - for inf in builtin_automatically_routed_apis(): - # if we do not serve the corresponding router API, we should not serve the routing table API - if inf.router_api.value not in apis_to_serve: - continue - apis_to_serve.add(inf.routing_table_api.value) - - apis_to_serve.add("inspect") - apis_to_serve.add("providers") - for api_str in apis_to_serve: - api = Api(api_str) - - routes = all_routes[api] - impl = impls[api] - - for route in routes: - if not hasattr(impl, route.name): - # ideally this should be a typing violation already - raise ValueError(f"Could not find method {route.name} on {impl}!") - - impl_method = getattr(impl, route.name) - # Filter out HEAD method since it's automatically handled by FastAPI for GET routes - available_methods = [m for m in route.methods if m != "HEAD"] - if not available_methods: - raise ValueError(f"No methods found for {route.name} on {impl}") - method = available_methods[0] - logger.debug(f"{method} {route.path}") - - with warnings.catch_warnings(): - warnings.filterwarnings("ignore", category=UserWarning, module="pydantic._internal._fields") - getattr(app, method.lower())(route.path, response_model=None)( - create_dynamic_typed_route( - impl_method, - method.lower(), - route.path, - ) - ) - + apis_to_serve, impls = asyncio.run(construct(app=app, config=config)) logger.debug(f"serving APIs: {apis_to_serve}") app.exception_handler(RequestValidationError)(global_exception_handler) app.exception_handler(Exception)(global_exception_handler) app.__llama_stack_impls__ = impls + # Add the custom middleware + app.add_middleware(RequestMiddleware, api=app, stack_run_config=config) app.add_middleware(TracingMiddleware, impls=impls) import uvicorn @@ -592,5 +620,81 @@ def extract_path_params(route: str) -> list[str]: return params +async def construct( + app: FastAPI, config: StackRunConfig, reconstruct: bool = False +) -> tuple[set[str] | set[Api], dict[Api, Any]]: + try: + impls = await construct_stack(config) + except InvalidProviderError as e: + logger.error(f"Error: {str(e)}") + sys.exit(1) + + if Api.telemetry in impls: + setup_logger(impls[Api.telemetry]) + else: + setup_logger(TelemetryAdapter(TelemetryConfig(), {})) + + all_routes = get_all_api_routes() + + if config.apis: + apis_to_serve = set(config.apis) + else: + apis_to_serve = set(impls.keys()) + + for inf in builtin_automatically_routed_apis(): + # if we do not serve the corresponding router API, we should not serve the routing table API + if inf.router_api.value not in apis_to_serve: + continue + apis_to_serve.add(inf.routing_table_api.value) + + apis_to_serve.add("inspect") + apis_to_serve.add("providers") + + for api_str in apis_to_serve: + api = Api(api_str) + + routes = all_routes[api] + impl = impls[api] + for route in routes: + if not hasattr(impl, route.name): + # ideally this should be a typing violation already + raise ValueError(f"Could not find method {route.name} on {impl}!!") + + impl_method = getattr(impl, route.name) + + # Filter out HEAD method since it's automatically handled by FastAPI for GET routes + available_methods = [m for m in route.methods if m != "HEAD"] + if not available_methods: + raise ValueError(f"No methods found for {route.name} on {impl}") + method = available_methods[0] + logger.debug(f"{method} {route.path}") + + with warnings.catch_warnings(): + warnings.filterwarnings("ignore", category=UserWarning, module="pydantic._internal._fields") + # Remove old route + if reconstruct: + app.router.routes = [ + r for r in app.router.routes if not (r.path == route.path and method.lower() in r.methods) + ] + new_endpoint = create_dynamic_typed_route( + impl_method, + method.lower(), + route.path, # route.path + ) + getattr(app, method.lower())(route.path, response_model=None)(new_endpoint) + if reconstruct: + new_route = APIRoute( + response_model=None, + path=route.path, + endpoint=new_endpoint, + methods=[method.lower()], + name=impl_method.__name__, + ) + # Add new route + app.router.routes.append(new_route) + + return apis_to_serve, impls + + if __name__ == "__main__": main() diff --git a/tests/integration/providers/test_providers.py b/tests/integration/providers/test_providers.py index 8b153411c..2069d4b61 100644 --- a/tests/integration/providers/test_providers.py +++ b/tests/integration/providers/test_providers.py @@ -21,3 +21,20 @@ class TestProviders: pid = provider.provider_id provider = llama_stack_client.providers.retrieve(pid) assert provider is not None + + @pytest.mark.asyncio + def test_providers_update(self, llama_stack_client: LlamaStackAsLibraryClient | LlamaStackClient): + new_cfg = {"url": "http://localhost:12345"} + + _ = llama_stack_client.providers.retrieve("ollama") + + llama_stack_client.providers.update( + api="inference", + provider_id="ollama", + provider_type="remote::ollama", + config=new_cfg, + ) + + new_provider = llama_stack_client.providers.retrieve("ollama") + + assert new_provider.config == new_cfg From e4b040d5cc503a04a6bf0a68863969f6c4dffccd Mon Sep 17 00:00:00 2001 From: Charlie Doern Date: Mon, 30 Jun 2025 15:07:09 -0400 Subject: [PATCH 2/2] feat: allow for provider immutability being able to update providers means that admins should have the ability to turn this feature off. introduce `immutable` as a field in the Provider class. Defauling to false means all providers can be updated by default, but an admin at runtime can choose to set this to True to disable provider updating Signed-off-by: Charlie Doern --- llama_stack/distribution/datatypes.py | 1 + llama_stack/distribution/providers.py | 3 ++ llama_stack/templates/bedrock/run.yaml | 15 ++++++++++ llama_stack/templates/cerebras/run.yaml | 15 ++++++++++ llama_stack/templates/ci-tests/run.yaml | 16 ++++++++++ .../templates/dell/run-with-safety.yaml | 16 ++++++++++ llama_stack/templates/dell/run.yaml | 15 ++++++++++ .../templates/fireworks/run-with-safety.yaml | 20 +++++++++++++ llama_stack/templates/fireworks/run.yaml | 18 ++++++++++++ llama_stack/templates/groq/run.yaml | 15 ++++++++++ .../hf-endpoint/run-with-safety.yaml | 17 +++++++++++ llama_stack/templates/hf-endpoint/run.yaml | 16 ++++++++++ .../hf-serverless/run-with-safety.yaml | 17 +++++++++++ llama_stack/templates/hf-serverless/run.yaml | 16 ++++++++++ llama_stack/templates/llama_api/run.yaml | 18 ++++++++++++ .../meta-reference-gpu/run-with-safety.yaml | 17 +++++++++++ .../templates/meta-reference-gpu/run.yaml | 16 ++++++++++ .../templates/nvidia/run-with-safety.yaml | 12 ++++++++ llama_stack/templates/nvidia/run.yaml | 10 +++++++ .../templates/ollama/run-with-safety.yaml | 19 ++++++++++++ llama_stack/templates/ollama/run.yaml | 18 ++++++++++++ llama_stack/templates/open-benchmark/run.yaml | 21 ++++++++++++++ .../passthrough/run-with-safety.yaml | 19 ++++++++++++ llama_stack/templates/passthrough/run.yaml | 17 +++++++++++ llama_stack/templates/postgres-demo/run.yaml | 10 +++++++ .../remote-vllm/run-with-safety.yaml | 18 ++++++++++++ llama_stack/templates/remote-vllm/run.yaml | 17 +++++++++++ llama_stack/templates/sambanova/run.yaml | 13 +++++++++ llama_stack/templates/starter/run.yaml | 29 +++++++++++++++++++ .../templates/tgi/run-with-safety.yaml | 16 ++++++++++ llama_stack/templates/tgi/run.yaml | 16 ++++++++++ .../templates/together/run-with-safety.yaml | 19 ++++++++++++ llama_stack/templates/together/run.yaml | 17 +++++++++++ llama_stack/templates/vllm-gpu/run.yaml | 16 ++++++++++ llama_stack/templates/watsonx/run.yaml | 16 ++++++++++ 35 files changed, 554 insertions(+) diff --git a/llama_stack/distribution/datatypes.py b/llama_stack/distribution/datatypes.py index 5e48ac0ad..1118bf82f 100644 --- a/llama_stack/distribution/datatypes.py +++ b/llama_stack/distribution/datatypes.py @@ -151,6 +151,7 @@ class Provider(BaseModel): provider_id: str | None provider_type: str config: dict[str, Any] + immutable: bool = False class LoggingConfig(BaseModel): diff --git a/llama_stack/distribution/providers.py b/llama_stack/distribution/providers.py index 3fc61df73..75599f9ed 100644 --- a/llama_stack/distribution/providers.py +++ b/llama_stack/distribution/providers.py @@ -147,7 +147,10 @@ class ProviderImpl(Providers): if prov_api != api: continue for p in providers: + # the provider needs to be mutable for us to update its config if p.provider_id == provider_id: + if p.immutable: + raise ValueError(f"Provider {provider_id} is immutable, you can only update mutable providers.") existing_provider = p break if existing_provider is not None: diff --git a/llama_stack/templates/bedrock/run.yaml b/llama_stack/templates/bedrock/run.yaml index f12c5bec5..afb4a1e38 100644 --- a/llama_stack/templates/bedrock/run.yaml +++ b/llama_stack/templates/bedrock/run.yaml @@ -15,6 +15,7 @@ providers: - provider_id: bedrock provider_type: remote::bedrock config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -23,10 +24,12 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/bedrock}/faiss_store.db + immutable: false safety: - provider_id: bedrock provider_type: remote::bedrock config: {} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -38,6 +41,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/bedrock}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -45,6 +49,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/bedrock}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -53,6 +58,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/bedrock}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -61,6 +67,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/bedrock}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -68,34 +75,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/bedrock}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/bedrock}/registry.db diff --git a/llama_stack/templates/cerebras/run.yaml b/llama_stack/templates/cerebras/run.yaml index c3877ddce..b34e51540 100644 --- a/llama_stack/templates/cerebras/run.yaml +++ b/llama_stack/templates/cerebras/run.yaml @@ -17,14 +17,17 @@ providers: config: base_url: https://api.cerebras.ai api_key: ${env.CEREBRAS_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -33,6 +36,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/cerebras}/faiss_store.db + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +48,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/cerebras}/responses_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -52,6 +57,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/cerebras}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -60,6 +66,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/cerebras}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -67,17 +74,21 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/cerebras}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -85,20 +96,24 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/cerebras}/trace_store.db + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/cerebras}/registry.db diff --git a/llama_stack/templates/ci-tests/run.yaml b/llama_stack/templates/ci-tests/run.yaml index a38d09324..1fbd3a329 100644 --- a/llama_stack/templates/ci-tests/run.yaml +++ b/llama_stack/templates/ci-tests/run.yaml @@ -17,19 +17,23 @@ providers: config: url: https://api.fireworks.ai/inference/v1 api_key: ${env.FIREWORKS_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: sqlite-vec provider_type: inline::sqlite-vec config: db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/sqlite_vec.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -41,6 +45,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -48,6 +53,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -56,6 +62,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -64,6 +71,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -71,34 +79,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/registry.db diff --git a/llama_stack/templates/dell/run-with-safety.yaml b/llama_stack/templates/dell/run-with-safety.yaml index 48639c772..b170f0d10 100644 --- a/llama_stack/templates/dell/run-with-safety.yaml +++ b/llama_stack/templates/dell/run-with-safety.yaml @@ -16,23 +16,28 @@ providers: provider_type: remote::tgi config: url: ${env.DEH_URL} + immutable: false - provider_id: tgi1 provider_type: remote::tgi config: url: ${env.DEH_SAFETY_URL} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: chromadb provider_type: remote::chromadb config: url: ${env.CHROMA_URL} + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +49,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,6 +57,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +66,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -67,6 +75,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -74,31 +83,38 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/registry.db diff --git a/llama_stack/templates/dell/run.yaml b/llama_stack/templates/dell/run.yaml index 13d43530b..6c847280e 100644 --- a/llama_stack/templates/dell/run.yaml +++ b/llama_stack/templates/dell/run.yaml @@ -16,19 +16,23 @@ providers: provider_type: remote::tgi config: url: ${env.DEH_URL} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: chromadb provider_type: remote::chromadb config: url: ${env.CHROMA_URL} + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -40,6 +44,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -47,6 +52,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -55,6 +61,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -63,6 +70,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -70,31 +78,38 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/registry.db diff --git a/llama_stack/templates/fireworks/run-with-safety.yaml b/llama_stack/templates/fireworks/run-with-safety.yaml index ecb53a18d..d9739ed03 100644 --- a/llama_stack/templates/fireworks/run-with-safety.yaml +++ b/llama_stack/templates/fireworks/run-with-safety.yaml @@ -18,9 +18,11 @@ providers: config: url: https://api.fireworks.ai/inference/v1 api_key: ${env.FIREWORKS_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -29,16 +31,20 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: {} + immutable: false - provider_id: llama-guard-vision provider_type: inline::llama-guard config: {} + immutable: false - provider_id: code-scanner provider_type: inline::code-scanner config: {} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -50,6 +56,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -57,6 +64,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -65,6 +73,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -73,6 +82,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -80,17 +90,21 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false files: - provider_id: meta-reference-files provider_type: inline::localfs @@ -99,27 +113,33 @@ providers: metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/files_metadata.db + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/registry.db diff --git a/llama_stack/templates/fireworks/run.yaml b/llama_stack/templates/fireworks/run.yaml index 298d28d52..4859730f8 100644 --- a/llama_stack/templates/fireworks/run.yaml +++ b/llama_stack/templates/fireworks/run.yaml @@ -18,9 +18,11 @@ providers: config: url: https://api.fireworks.ai/inference/v1 api_key: ${env.FIREWORKS_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -29,11 +31,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -45,6 +49,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -52,6 +57,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -60,6 +66,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -68,6 +75,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -75,17 +83,21 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false files: - provider_id: meta-reference-files provider_type: inline::localfs @@ -94,27 +106,33 @@ providers: metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/files_metadata.db + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/fireworks}/registry.db diff --git a/llama_stack/templates/groq/run.yaml b/llama_stack/templates/groq/run.yaml index 13bb65ed2..2826af613 100644 --- a/llama_stack/templates/groq/run.yaml +++ b/llama_stack/templates/groq/run.yaml @@ -17,9 +17,11 @@ providers: config: url: https://api.groq.com api_key: ${env.GROQ_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,11 +30,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/groq}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +48,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/groq}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,6 +56,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/groq}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +65,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/groq}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -67,6 +74,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/groq}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -74,31 +82,38 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/groq}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/groq}/registry.db diff --git a/llama_stack/templates/hf-endpoint/run-with-safety.yaml b/llama_stack/templates/hf-endpoint/run-with-safety.yaml index b2bc6a8e9..7b6a38340 100644 --- a/llama_stack/templates/hf-endpoint/run-with-safety.yaml +++ b/llama_stack/templates/hf-endpoint/run-with-safety.yaml @@ -17,14 +17,17 @@ providers: config: endpoint_name: ${env.INFERENCE_ENDPOINT_NAME} api_token: ${env.HF_API_TOKEN} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false - provider_id: hf-endpoint-safety provider_type: remote::hf::endpoint config: endpoint_name: ${env.SAFETY_INFERENCE_ENDPOINT_NAME} api_token: ${env.HF_API_TOKEN} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -33,11 +36,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -49,6 +54,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -56,6 +62,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -64,6 +71,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -72,6 +80,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -79,34 +88,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/registry.db diff --git a/llama_stack/templates/hf-endpoint/run.yaml b/llama_stack/templates/hf-endpoint/run.yaml index d62921ccc..08b55e9b0 100644 --- a/llama_stack/templates/hf-endpoint/run.yaml +++ b/llama_stack/templates/hf-endpoint/run.yaml @@ -17,9 +17,11 @@ providers: config: endpoint_name: ${env.INFERENCE_ENDPOINT_NAME} api_token: ${env.HF_API_TOKEN} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,11 +30,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +48,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,6 +56,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +65,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -67,6 +74,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -74,34 +82,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-endpoint}/registry.db diff --git a/llama_stack/templates/hf-serverless/run-with-safety.yaml b/llama_stack/templates/hf-serverless/run-with-safety.yaml index d7ff4f446..92fcc24ad 100644 --- a/llama_stack/templates/hf-serverless/run-with-safety.yaml +++ b/llama_stack/templates/hf-serverless/run-with-safety.yaml @@ -17,14 +17,17 @@ providers: config: huggingface_repo: ${env.INFERENCE_MODEL} api_token: ${env.HF_API_TOKEN} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false - provider_id: hf-serverless-safety provider_type: remote::hf::serverless config: huggingface_repo: ${env.SAFETY_MODEL} api_token: ${env.HF_API_TOKEN} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -33,11 +36,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -49,6 +54,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -56,6 +62,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -64,6 +71,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -72,6 +80,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -79,34 +88,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/registry.db diff --git a/llama_stack/templates/hf-serverless/run.yaml b/llama_stack/templates/hf-serverless/run.yaml index 19484cba6..5a1b3c87b 100644 --- a/llama_stack/templates/hf-serverless/run.yaml +++ b/llama_stack/templates/hf-serverless/run.yaml @@ -17,9 +17,11 @@ providers: config: huggingface_repo: ${env.INFERENCE_MODEL} api_token: ${env.HF_API_TOKEN} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,11 +30,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +48,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,6 +56,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +65,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -67,6 +74,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -74,34 +82,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/hf-serverless}/registry.db diff --git a/llama_stack/templates/llama_api/run.yaml b/llama_stack/templates/llama_api/run.yaml index 3bfb284a3..1a10cf6b1 100644 --- a/llama_stack/templates/llama_api/run.yaml +++ b/llama_stack/templates/llama_api/run.yaml @@ -17,18 +17,22 @@ providers: config: openai_compat_api_base: https://api.llama.com/compat/v1/ api_key: ${env.LLAMA_API_KEY:=} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: sqlite-vec provider_type: inline::sqlite-vec config: db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/llama_api}/sqlite_vec.db + immutable: false - provider_id: ${env.ENABLE_CHROMADB:+chromadb} provider_type: remote::chromadb config: url: ${env.CHROMADB_URL:=} + immutable: false - provider_id: ${env.ENABLE_PGVECTOR:+pgvector} provider_type: remote::pgvector config: @@ -37,11 +41,13 @@ providers: db: ${env.PGVECTOR_DB:=} user: ${env.PGVECTOR_USER:=} password: ${env.PGVECTOR_PASSWORD:=} + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -53,6 +59,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/llama_api}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -60,6 +67,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/llama_api}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -68,6 +76,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/llama_api}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -76,6 +85,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/llama_api}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -83,34 +93,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/llama_api}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/llama_api}/registry.db diff --git a/llama_stack/templates/meta-reference-gpu/run-with-safety.yaml b/llama_stack/templates/meta-reference-gpu/run-with-safety.yaml index 46b3a33a6..03decd27c 100644 --- a/llama_stack/templates/meta-reference-gpu/run-with-safety.yaml +++ b/llama_stack/templates/meta-reference-gpu/run-with-safety.yaml @@ -22,9 +22,11 @@ providers: model_parallel_size: ${env.MODEL_PARALLEL_SIZE:=0} max_batch_size: ${env.MAX_BATCH_SIZE:=1} max_seq_len: ${env.MAX_SEQ_LEN:=4096} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false - provider_id: meta-reference-safety provider_type: inline::meta-reference config: @@ -35,6 +37,7 @@ providers: model_parallel_size: ${env.MODEL_PARALLEL_SIZE:=0} max_batch_size: ${env.MAX_BATCH_SIZE:=1} max_seq_len: ${env.MAX_SEQ_LEN:=4096} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -43,11 +46,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +64,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -66,6 +72,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -74,6 +81,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -82,6 +90,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -89,34 +98,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/registry.db diff --git a/llama_stack/templates/meta-reference-gpu/run.yaml b/llama_stack/templates/meta-reference-gpu/run.yaml index 033ec245a..7862b10a9 100644 --- a/llama_stack/templates/meta-reference-gpu/run.yaml +++ b/llama_stack/templates/meta-reference-gpu/run.yaml @@ -22,9 +22,11 @@ providers: model_parallel_size: ${env.MODEL_PARALLEL_SIZE:=0} max_batch_size: ${env.MAX_BATCH_SIZE:=1} max_seq_len: ${env.MAX_SEQ_LEN:=4096} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -33,11 +35,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -49,6 +53,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -56,6 +61,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -64,6 +70,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -72,6 +79,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -79,34 +87,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/registry.db diff --git a/llama_stack/templates/nvidia/run-with-safety.yaml b/llama_stack/templates/nvidia/run-with-safety.yaml index 73783be98..5655e7691 100644 --- a/llama_stack/templates/nvidia/run-with-safety.yaml +++ b/llama_stack/templates/nvidia/run-with-safety.yaml @@ -19,11 +19,13 @@ providers: url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com} api_key: ${env.NVIDIA_API_KEY:=} append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True} + immutable: false - provider_id: nvidia provider_type: remote::nvidia config: guardrails_service_url: ${env.GUARDRAILS_SERVICE_URL:=http://localhost:7331} config_id: ${env.NVIDIA_GUARDRAILS_CONFIG_ID:=self-check} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -32,12 +34,14 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/faiss_store.db + immutable: false safety: - provider_id: nvidia provider_type: remote::nvidia config: guardrails_service_url: ${env.GUARDRAILS_SERVICE_URL:=http://localhost:7331} config_id: ${env.NVIDIA_GUARDRAILS_CONFIG_ID:=self-check} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -49,6 +53,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -56,11 +61,13 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/trace_store.db + immutable: false eval: - provider_id: nvidia provider_type: remote::nvidia config: evaluator_url: ${env.NVIDIA_EVALUATOR_URL:=http://localhost:7331} + immutable: false post_training: - provider_id: nvidia provider_type: remote::nvidia @@ -69,6 +76,7 @@ providers: dataset_namespace: ${env.NVIDIA_DATASET_NAMESPACE:=default} project_id: ${env.NVIDIA_PROJECT_ID:=test-project} customizer_url: ${env.NVIDIA_CUSTOMIZER_URL:=http://nemo.test} + immutable: false datasetio: - provider_id: localfs provider_type: inline::localfs @@ -77,6 +85,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/localfs_datasetio.db + immutable: false - provider_id: nvidia provider_type: remote::nvidia config: @@ -84,14 +93,17 @@ providers: dataset_namespace: ${env.NVIDIA_DATASET_NAMESPACE:=default} project_id: ${env.NVIDIA_PROJECT_ID:=test-project} datasets_url: ${env.NVIDIA_DATASETS_URL:=http://nemo.test} + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false tool_runtime: - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/registry.db diff --git a/llama_stack/templates/nvidia/run.yaml b/llama_stack/templates/nvidia/run.yaml index af9d5904a..e9699f714 100644 --- a/llama_stack/templates/nvidia/run.yaml +++ b/llama_stack/templates/nvidia/run.yaml @@ -19,6 +19,7 @@ providers: url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com} api_key: ${env.NVIDIA_API_KEY:=} append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -27,12 +28,14 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/faiss_store.db + immutable: false safety: - provider_id: nvidia provider_type: remote::nvidia config: guardrails_service_url: ${env.GUARDRAILS_SERVICE_URL:=http://localhost:7331} config_id: ${env.NVIDIA_GUARDRAILS_CONFIG_ID:=self-check} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +47,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,11 +55,13 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/trace_store.db + immutable: false eval: - provider_id: nvidia provider_type: remote::nvidia config: evaluator_url: ${env.NVIDIA_EVALUATOR_URL:=http://localhost:7331} + immutable: false post_training: - provider_id: nvidia provider_type: remote::nvidia @@ -64,6 +70,7 @@ providers: dataset_namespace: ${env.NVIDIA_DATASET_NAMESPACE:=default} project_id: ${env.NVIDIA_PROJECT_ID:=test-project} customizer_url: ${env.NVIDIA_CUSTOMIZER_URL:=http://nemo.test} + immutable: false datasetio: - provider_id: nvidia provider_type: remote::nvidia @@ -72,14 +79,17 @@ providers: dataset_namespace: ${env.NVIDIA_DATASET_NAMESPACE:=default} project_id: ${env.NVIDIA_PROJECT_ID:=test-project} datasets_url: ${env.NVIDIA_DATASETS_URL:=http://nemo.test} + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false tool_runtime: - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/nvidia}/registry.db diff --git a/llama_stack/templates/ollama/run-with-safety.yaml b/llama_stack/templates/ollama/run-with-safety.yaml index bad51de09..ac3fe02ca 100644 --- a/llama_stack/templates/ollama/run-with-safety.yaml +++ b/llama_stack/templates/ollama/run-with-safety.yaml @@ -19,6 +19,7 @@ providers: config: url: ${env.OLLAMA_URL:=http://localhost:11434} raise_on_connect_error: true + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -27,13 +28,16 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: {} + immutable: false - provider_id: code-scanner provider_type: inline::code-scanner config: {} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -45,6 +49,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -52,6 +57,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -60,6 +66,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -68,6 +75,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -75,17 +83,21 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false files: - provider_id: meta-reference-files provider_type: inline::localfs @@ -94,6 +106,7 @@ providers: metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/files_metadata.db + immutable: false post_training: - provider_id: huggingface provider_type: inline::huggingface @@ -101,27 +114,33 @@ providers: checkpoint_format: huggingface distributed_backend: null device: cpu + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/registry.db diff --git a/llama_stack/templates/ollama/run.yaml b/llama_stack/templates/ollama/run.yaml index e1dea730e..ce27cefc5 100644 --- a/llama_stack/templates/ollama/run.yaml +++ b/llama_stack/templates/ollama/run.yaml @@ -19,6 +19,7 @@ providers: config: url: ${env.OLLAMA_URL:=http://localhost:11434} raise_on_connect_error: true + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -27,11 +28,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -43,6 +46,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -50,6 +54,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -58,6 +63,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -66,6 +72,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -73,17 +80,21 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false files: - provider_id: meta-reference-files provider_type: inline::localfs @@ -92,6 +103,7 @@ providers: metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/files_metadata.db + immutable: false post_training: - provider_id: huggingface provider_type: inline::huggingface @@ -99,27 +111,33 @@ providers: checkpoint_format: huggingface distributed_backend: null device: cpu + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/registry.db diff --git a/llama_stack/templates/open-benchmark/run.yaml b/llama_stack/templates/open-benchmark/run.yaml index 57ae6b9be..5268b7627 100644 --- a/llama_stack/templates/open-benchmark/run.yaml +++ b/llama_stack/templates/open-benchmark/run.yaml @@ -16,33 +16,40 @@ providers: provider_type: remote::openai config: api_key: ${env.OPENAI_API_KEY:=} + immutable: false - provider_id: anthropic provider_type: remote::anthropic config: api_key: ${env.ANTHROPIC_API_KEY:=} + immutable: false - provider_id: gemini provider_type: remote::gemini config: api_key: ${env.GEMINI_API_KEY:=} + immutable: false - provider_id: groq provider_type: remote::groq config: url: https://api.groq.com api_key: ${env.GROQ_API_KEY:=} + immutable: false - provider_id: together provider_type: remote::together config: url: https://api.together.xyz/v1 api_key: ${env.TOGETHER_API_KEY:=} + immutable: false vector_io: - provider_id: sqlite-vec provider_type: inline::sqlite-vec config: db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/open-benchmark}/sqlite_vec.db + immutable: false - provider_id: ${env.ENABLE_CHROMADB:+chromadb} provider_type: remote::chromadb config: url: ${env.CHROMADB_URL:=} + immutable: false - provider_id: ${env.ENABLE_PGVECTOR:+pgvector} provider_type: remote::pgvector config: @@ -51,11 +58,13 @@ providers: db: ${env.PGVECTOR_DB:=} user: ${env.PGVECTOR_USER:=} password: ${env.PGVECTOR_PASSWORD:=} + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -67,6 +76,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/open-benchmark}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -74,6 +84,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/open-benchmark}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -82,6 +93,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/open-benchmark}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -90,6 +102,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/open-benchmark}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -97,34 +110,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/open-benchmark}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/open-benchmark}/registry.db diff --git a/llama_stack/templates/passthrough/run-with-safety.yaml b/llama_stack/templates/passthrough/run-with-safety.yaml index 7a30f665c..2e09bb808 100644 --- a/llama_stack/templates/passthrough/run-with-safety.yaml +++ b/llama_stack/templates/passthrough/run-with-safety.yaml @@ -17,9 +17,11 @@ providers: config: url: ${env.PASSTHROUGH_URL} api_key: ${env.PASSTHROUGH_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,16 +30,20 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: {} + immutable: false - provider_id: llama-guard-vision provider_type: inline::llama-guard config: {} + immutable: false - provider_id: code-scanner provider_type: inline::code-scanner config: {} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -49,6 +55,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -56,6 +63,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -64,6 +72,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -72,6 +81,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -79,38 +89,47 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/registry.db diff --git a/llama_stack/templates/passthrough/run.yaml b/llama_stack/templates/passthrough/run.yaml index dc751ea20..c5675c7ed 100644 --- a/llama_stack/templates/passthrough/run.yaml +++ b/llama_stack/templates/passthrough/run.yaml @@ -17,9 +17,11 @@ providers: config: url: ${env.PASSTHROUGH_URL} api_key: ${env.PASSTHROUGH_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,11 +30,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +48,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,6 +56,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +65,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -67,6 +74,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -74,38 +82,47 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/passthrough}/registry.db diff --git a/llama_stack/templates/postgres-demo/run.yaml b/llama_stack/templates/postgres-demo/run.yaml index dd20cc6ac..4071ed5d8 100644 --- a/llama_stack/templates/postgres-demo/run.yaml +++ b/llama_stack/templates/postgres-demo/run.yaml @@ -16,19 +16,23 @@ providers: max_tokens: ${env.VLLM_MAX_TOKENS:=4096} api_token: ${env.VLLM_API_TOKEN:=fake} tls_verify: ${env.VLLM_TLS_VERIFY:=true} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: ${env.ENABLE_CHROMADB:+chromadb} provider_type: remote::chromadb config: url: ${env.CHROMADB_URL:=} + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -47,6 +51,7 @@ providers: db: ${env.POSTGRES_DB:=llamastack} user: ${env.POSTGRES_USER:=llamastack} password: ${env.POSTGRES_PASSWORD:=llamastack} + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -54,23 +59,28 @@ providers: service_name: ${env.OTEL_SERVICE_NAME:=} sinks: ${env.TELEMETRY_SINKS:=console,otel_trace} otel_trace_endpoint: ${env.OTEL_TRACE_ENDPOINT:=http://localhost:4318/v1/traces} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: postgres host: ${env.POSTGRES_HOST:=localhost} diff --git a/llama_stack/templates/remote-vllm/run-with-safety.yaml b/llama_stack/templates/remote-vllm/run-with-safety.yaml index 78fb22d38..122672847 100644 --- a/llama_stack/templates/remote-vllm/run-with-safety.yaml +++ b/llama_stack/templates/remote-vllm/run-with-safety.yaml @@ -19,6 +19,7 @@ providers: max_tokens: ${env.VLLM_MAX_TOKENS:=4096} api_token: ${env.VLLM_API_TOKEN:=fake} tls_verify: ${env.VLLM_TLS_VERIFY:=true} + immutable: false - provider_id: vllm-safety provider_type: remote::vllm config: @@ -26,9 +27,11 @@ providers: max_tokens: ${env.VLLM_MAX_TOKENS:=4096} api_token: ${env.VLLM_API_TOKEN:=fake} tls_verify: ${env.VLLM_TLS_VERIFY:=true} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -37,11 +40,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -53,6 +58,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/responses_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -61,6 +67,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -69,6 +76,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -76,17 +84,21 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -94,27 +106,33 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/trace_store.db + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/registry.db diff --git a/llama_stack/templates/remote-vllm/run.yaml b/llama_stack/templates/remote-vllm/run.yaml index 1cc4596f3..d60cec984 100644 --- a/llama_stack/templates/remote-vllm/run.yaml +++ b/llama_stack/templates/remote-vllm/run.yaml @@ -19,9 +19,11 @@ providers: max_tokens: ${env.VLLM_MAX_TOKENS:=4096} api_token: ${env.VLLM_API_TOKEN:=fake} tls_verify: ${env.VLLM_TLS_VERIFY:=true} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -30,11 +32,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -46,6 +50,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/responses_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -54,6 +59,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -62,6 +68,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -69,17 +76,21 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -87,27 +98,33 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/trace_store.db + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/remote-vllm}/registry.db diff --git a/llama_stack/templates/sambanova/run.yaml b/llama_stack/templates/sambanova/run.yaml index 6163a58b3..44a2dff01 100644 --- a/llama_stack/templates/sambanova/run.yaml +++ b/llama_stack/templates/sambanova/run.yaml @@ -14,9 +14,11 @@ providers: config: url: https://api.sambanova.ai/v1 api_key: ${env.SAMBANOVA_API_KEY} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -25,10 +27,12 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/sambanova}/faiss_store.db + immutable: false - provider_id: ${env.ENABLE_CHROMADB:+chromadb} provider_type: remote::chromadb config: url: ${env.CHROMADB_URL:=} + immutable: false - provider_id: ${env.ENABLE_PGVECTOR:+pgvector} provider_type: remote::pgvector config: @@ -37,12 +41,14 @@ providers: db: ${env.PGVECTOR_DB:=} user: ${env.PGVECTOR_USER:=} password: ${env.PGVECTOR_PASSWORD:=} + immutable: false safety: - provider_id: sambanova provider_type: remote::sambanova config: url: https://api.sambanova.ai/v1 api_key: ${env.SAMBANOVA_API_KEY} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -54,6 +60,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/sambanova}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -61,27 +68,33 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/sambanova}/trace_store.db + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/sambanova}/registry.db diff --git a/llama_stack/templates/starter/run.yaml b/llama_stack/templates/starter/run.yaml index 190030690..7a3f5f88f 100644 --- a/llama_stack/templates/starter/run.yaml +++ b/llama_stack/templates/starter/run.yaml @@ -17,39 +17,47 @@ providers: provider_type: remote::openai config: api_key: ${env.OPENAI_API_KEY:=} + immutable: false - provider_id: fireworks provider_type: remote::fireworks config: url: https://api.fireworks.ai/inference/v1 api_key: ${env.FIREWORKS_API_KEY:=} + immutable: false - provider_id: together provider_type: remote::together config: url: https://api.together.xyz/v1 api_key: ${env.TOGETHER_API_KEY:=} + immutable: false - provider_id: ollama provider_type: remote::ollama config: url: ${env.OLLAMA_URL:=http://localhost:11434} raise_on_connect_error: false + immutable: false - provider_id: anthropic provider_type: remote::anthropic config: api_key: ${env.ANTHROPIC_API_KEY:=} + immutable: false - provider_id: gemini provider_type: remote::gemini config: api_key: ${env.GEMINI_API_KEY:=} + immutable: false - provider_id: groq provider_type: remote::groq config: url: https://api.groq.com api_key: ${env.GROQ_API_KEY:=} + immutable: false - provider_id: sambanova provider_type: remote::sambanova config: url: https://api.sambanova.ai/v1 api_key: ${env.SAMBANOVA_API_KEY:=} + immutable: false - provider_id: vllm provider_type: remote::vllm config: @@ -57,9 +65,11 @@ providers: max_tokens: ${env.VLLM_MAX_TOKENS:=4096} api_token: ${env.VLLM_API_TOKEN:=fake} tls_verify: ${env.VLLM_TLS_VERIFY:=true} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -68,10 +78,12 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/faiss_store.db + immutable: false - provider_id: ${env.ENABLE_SQLITE_VEC:+sqlite-vec} provider_type: inline::sqlite-vec config: db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/sqlite_vec.db + immutable: false - provider_id: ${env.ENABLE_MILVUS:+milvus} provider_type: inline::milvus config: @@ -80,10 +92,12 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/milvus_registry.db + immutable: false - provider_id: ${env.ENABLE_CHROMADB:+chromadb} provider_type: remote::chromadb config: url: ${env.CHROMADB_URL:=} + immutable: false - provider_id: ${env.ENABLE_PGVECTOR:+pgvector} provider_type: remote::pgvector config: @@ -92,6 +106,7 @@ providers: db: ${env.PGVECTOR_DB:=} user: ${env.PGVECTOR_USER:=} password: ${env.PGVECTOR_PASSWORD:=} + immutable: false files: - provider_id: meta-reference-files provider_type: inline::localfs @@ -100,11 +115,13 @@ providers: metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/files_metadata.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -116,6 +133,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -123,6 +141,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -131,6 +150,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -139,6 +159,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -146,34 +167,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/registry.db diff --git a/llama_stack/templates/tgi/run-with-safety.yaml b/llama_stack/templates/tgi/run-with-safety.yaml index c4f9ae7ef..856d32fa9 100644 --- a/llama_stack/templates/tgi/run-with-safety.yaml +++ b/llama_stack/templates/tgi/run-with-safety.yaml @@ -16,10 +16,12 @@ providers: provider_type: remote::tgi config: url: ${env.TGI_URL} + immutable: false - provider_id: tgi-safety provider_type: remote::tgi config: url: ${env.TGI_SAFETY_URL} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,11 +30,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +48,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,6 +56,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +65,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -67,6 +74,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -74,34 +82,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/registry.db diff --git a/llama_stack/templates/tgi/run.yaml b/llama_stack/templates/tgi/run.yaml index 70e5872b3..bef7e8469 100644 --- a/llama_stack/templates/tgi/run.yaml +++ b/llama_stack/templates/tgi/run.yaml @@ -16,9 +16,11 @@ providers: provider_type: remote::tgi config: url: ${env.TGI_URL} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -27,11 +29,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -43,6 +47,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -50,6 +55,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -58,6 +64,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -66,6 +73,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -73,34 +81,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/tgi}/registry.db diff --git a/llama_stack/templates/together/run-with-safety.yaml b/llama_stack/templates/together/run-with-safety.yaml index 14f423855..0b4800cb0 100644 --- a/llama_stack/templates/together/run-with-safety.yaml +++ b/llama_stack/templates/together/run-with-safety.yaml @@ -17,9 +17,11 @@ providers: config: url: https://api.together.xyz/v1 api_key: ${env.TOGETHER_API_KEY:=} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,16 +30,20 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: {} + immutable: false - provider_id: llama-guard-vision provider_type: inline::llama-guard config: {} + immutable: false - provider_id: code-scanner provider_type: inline::code-scanner config: {} + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -49,6 +55,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -56,6 +63,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -64,6 +72,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -72,6 +81,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -79,38 +89,47 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/registry.db diff --git a/llama_stack/templates/together/run.yaml b/llama_stack/templates/together/run.yaml index 38f1922c0..a349f30a1 100644 --- a/llama_stack/templates/together/run.yaml +++ b/llama_stack/templates/together/run.yaml @@ -17,9 +17,11 @@ providers: config: url: https://api.together.xyz/v1 api_key: ${env.TOGETHER_API_KEY:=} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -28,11 +30,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -44,6 +48,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -51,6 +56,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -59,6 +65,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -67,6 +74,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -74,38 +82,47 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false - provider_id: wolfram-alpha provider_type: remote::wolfram-alpha config: api_key: ${env.WOLFRAM_ALPHA_API_KEY:=} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/together}/registry.db diff --git a/llama_stack/templates/vllm-gpu/run.yaml b/llama_stack/templates/vllm-gpu/run.yaml index 6854ad05c..32cae89ca 100644 --- a/llama_stack/templates/vllm-gpu/run.yaml +++ b/llama_stack/templates/vllm-gpu/run.yaml @@ -21,9 +21,11 @@ providers: max_num_seqs: ${env.MAX_NUM_SEQS:=4} enforce_eager: ${env.ENFORCE_EAGER:=False} gpu_memory_utilization: ${env.GPU_MEMORY_UTILIZATION:=0.3} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -32,11 +34,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/vllm-gpu}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -48,6 +52,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/vllm-gpu}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -55,6 +60,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/vllm-gpu}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -63,6 +69,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/vllm-gpu}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -71,6 +78,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/vllm-gpu}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -78,34 +86,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/vllm-gpu}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/vllm-gpu}/registry.db diff --git a/llama_stack/templates/watsonx/run.yaml b/llama_stack/templates/watsonx/run.yaml index 8b8fc09c4..7d5ea23d5 100644 --- a/llama_stack/templates/watsonx/run.yaml +++ b/llama_stack/templates/watsonx/run.yaml @@ -18,9 +18,11 @@ providers: url: ${env.WATSONX_BASE_URL:=https://us-south.ml.cloud.ibm.com} api_key: ${env.WATSONX_API_KEY:=} project_id: ${env.WATSONX_PROJECT_ID:=} + immutable: false - provider_id: sentence-transformers provider_type: inline::sentence-transformers config: {} + immutable: false vector_io: - provider_id: faiss provider_type: inline::faiss @@ -29,11 +31,13 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/watsonx}/faiss_store.db + immutable: false safety: - provider_id: llama-guard provider_type: inline::llama-guard config: excluded_categories: [] + immutable: false agents: - provider_id: meta-reference provider_type: inline::meta-reference @@ -45,6 +49,7 @@ providers: responses_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/watsonx}/responses_store.db + immutable: false telemetry: - provider_id: meta-reference provider_type: inline::meta-reference @@ -52,6 +57,7 @@ providers: service_name: "${env.OTEL_SERVICE_NAME:=\u200B}" sinks: ${env.TELEMETRY_SINKS:=console,sqlite} sqlite_db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/watsonx}/trace_store.db + immutable: false eval: - provider_id: meta-reference provider_type: inline::meta-reference @@ -60,6 +66,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/watsonx}/meta_reference_eval.db + immutable: false datasetio: - provider_id: huggingface provider_type: remote::huggingface @@ -68,6 +75,7 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/watsonx}/huggingface_datasetio.db + immutable: false - provider_id: localfs provider_type: inline::localfs config: @@ -75,34 +83,42 @@ providers: type: sqlite namespace: null db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/watsonx}/localfs_datasetio.db + immutable: false scoring: - provider_id: basic provider_type: inline::basic config: {} + immutable: false - provider_id: llm-as-judge provider_type: inline::llm-as-judge config: {} + immutable: false - provider_id: braintrust provider_type: inline::braintrust config: openai_api_key: ${env.OPENAI_API_KEY:=} + immutable: false tool_runtime: - provider_id: brave-search provider_type: remote::brave-search config: api_key: ${env.BRAVE_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: tavily-search provider_type: remote::tavily-search config: api_key: ${env.TAVILY_SEARCH_API_KEY:=} max_results: 3 + immutable: false - provider_id: rag-runtime provider_type: inline::rag-runtime config: {} + immutable: false - provider_id: model-context-protocol provider_type: remote::model-context-protocol config: {} + immutable: false metadata_store: type: sqlite db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/watsonx}/registry.db