chore!: remove telemetry API usage (#3815)

# What does this PR do?

remove telemetry as a providable API from the codebase. This includes
removing it from generated distributions but also the provider registry,
the router, etc

since `setup_logger` is tied pretty strictly to `Api.telemetry` being in
impls we still need an "instantiated provider" in our implementations.
However it should not be auto-routed or provided. So in
validate_and_prepare_providers (called from resolve_impls) I made it so
that if run_config.telemetry.enabled, we set up the meta-reference
"provider" internally to be used so that log_event will work when
called.

This is the neatest way I think we can remove telemetry from the
provider configs but also not need to rip apart the whole "telemetry is
a provider" logic just yet, but we can do it internally later without
disrupting users.

so telemetry is removed from the registry such that if a user puts
`telemetry:` as an API in their build/run config it will err out, but
can still be used by us internally as we go through this transition.


relates to #3806

Signed-off-by: Charlie Doern <cdoern@redhat.com>

llama_stack/core/configure.py

@@ -64,7 +64,7 @@ def configure_api_providers(config: StackRunConfig, build_spec: DistributionSpec
     if config.apis:
         apis_to_serve = config.apis
     else:
-        apis_to_serve = [a.value for a in Api if a not in (Api.telemetry, Api.inspect, Api.providers)]
+        apis_to_serve = [a.value for a in Api if a not in (Api.inspect, Api.providers)]
 
     for api_str in apis_to_serve:
         api = Api(api_str)

llama_stack/core/datatypes.py

@@ -176,6 +176,10 @@ class DistributionSpec(BaseModel):
     )
 
 
+class TelemetryConfig(BaseModel):
+    enabled: bool = Field(default=False, description="enable or disable telemetry")
+
+
 class LoggingConfig(BaseModel):
     category_levels: dict[str, str] = Field(
         default_factory=dict,
@@ -493,6 +497,8 @@ If not specified, a default SQLite store will be used.""",
 
     logging: LoggingConfig | None = Field(default=None, description="Configuration for Llama Stack Logging")
 
+    telemetry: TelemetryConfig = Field(default_factory=TelemetryConfig, description="Configuration for telemetry")
+
     server: ServerConfig = Field(
         default_factory=ServerConfig,
         description="Configuration for the HTTP(S) server",

llama_stack/core/distribution.py

@@ -25,7 +25,7 @@ from llama_stack.providers.datatypes import (
 logger = get_logger(name=__name__, category="core")
 
 
-INTERNAL_APIS = {Api.inspect, Api.providers, Api.prompts, Api.conversations}
+INTERNAL_APIS = {Api.inspect, Api.providers, Api.prompts, Api.conversations, Api.telemetry}
 
 
 def stack_apis() -> list[Api]:

llama_stack/core/library_client.py

@@ -48,12 +48,7 @@ from llama_stack.core.utils.config import redact_sensitive_fields
 from llama_stack.core.utils.context import preserve_contexts_async_generator
 from llama_stack.core.utils.exec import in_notebook
 from llama_stack.log import get_logger
-from llama_stack.providers.utils.telemetry.tracing import (
-    CURRENT_TRACE_CONTEXT,
-    end_trace,
-    setup_logger,
-    start_trace,
-)
+from llama_stack.providers.utils.telemetry.tracing import CURRENT_TRACE_CONTEXT, end_trace, setup_logger, start_trace
 from llama_stack.strong_typing.inspection import is_unwrapped_body_param
 
 logger = get_logger(name=__name__, category="core")

llama_stack/core/resolver.py

@@ -26,7 +26,6 @@ from llama_stack.apis.safety import Safety
 from llama_stack.apis.scoring import Scoring
 from llama_stack.apis.scoring_functions import ScoringFunctions
 from llama_stack.apis.shields import Shields
-from llama_stack.apis.telemetry import Telemetry
 from llama_stack.apis.tools import ToolGroups, ToolRuntime
 from llama_stack.apis.vector_io import VectorIO
 from llama_stack.apis.version import LLAMA_STACK_API_V1ALPHA
@@ -47,6 +46,7 @@ from llama_stack.providers.datatypes import (
     Api,
     BenchmarksProtocolPrivate,
     DatasetsProtocolPrivate,
+    InlineProviderSpec,
     ModelsProtocolPrivate,
     ProviderSpec,
     RemoteProviderConfig,
@@ -82,7 +82,6 @@ def api_protocol_map(external_apis: dict[Api, ExternalApiSpec] | None = None) ->
         Api.models: Models,
         Api.safety: Safety,
         Api.shields: Shields,
-        Api.telemetry: Telemetry,
         Api.datasetio: DatasetIO,
         Api.datasets: Datasets,
         Api.scoring: Scoring,
@@ -204,9 +203,7 @@ def specs_for_autorouted_apis(apis_to_serve: list[str] | set[str]) -> dict[str,
                     module="llama_stack.core.routers",
                     routing_table_api=info.routing_table_api,
                     api_dependencies=[info.routing_table_api],
-                    # Add telemetry as an optional dependency to all auto-routed providers
-                    optional_api_dependencies=[Api.telemetry],
-                    deps__=([info.routing_table_api.value, Api.telemetry.value]),
+                    deps__=([info.routing_table_api.value]),
                 ),
             )
         }
@@ -239,6 +236,24 @@ def validate_and_prepare_providers(
         key = api_str if api not in router_apis else f"inner-{api_str}"
         providers_with_specs[key] = specs
 
+    # TODO: remove this logic, telemetry should not have providers.
+    # if telemetry has been enabled in the config initialize our internal impl
+    # telemetry is not an external API so it SHOULD NOT be auto-routed.
+    if run_config.telemetry.enabled:
+        specs = {}
+        p = InlineProviderSpec(
+            api=Api.telemetry,
+            provider_type="inline::meta-reference",
+            pip_packages=[],
+            optional_api_dependencies=[Api.datasetio],
+            module="llama_stack.providers.inline.telemetry.meta_reference",
+            config_class="llama_stack.providers.inline.telemetry.meta_reference.config.TelemetryConfig",
+            description="Meta's reference implementation of telemetry and observability using OpenTelemetry.",
+        )
+        spec = ProviderWithSpec(spec=p, provider_type="inline::meta-reference", provider_id="meta-reference")
+        specs["meta-reference"] = spec
+        providers_with_specs["telemetry"] = specs
+
     return providers_with_specs
 
 
@@ -389,6 +404,8 @@ async def instantiate_provider(
         args = [config, deps]
         if "policy" in inspect.signature(getattr(module, method)).parameters:
             args.append(policy)
+        if "telemetry_enabled" in inspect.signature(getattr(module, method)).parameters and run_config.telemetry:
+            args.append(run_config.telemetry.enabled)
 
     fn = getattr(module, method)
     impl = await fn(*args)

llama_stack/core/routers/__init__.py

@@ -63,16 +63,17 @@ async def get_auto_router_impl(
         "eval": EvalRouter,
         "tool_runtime": ToolRuntimeRouter,
     }
-    api_to_deps = {
-        "inference": {"telemetry": Api.telemetry},
-    }
     if api.value not in api_to_routers:
         raise ValueError(f"API {api.value} not found in router map")
 
     api_to_dep_impl = {}
-    for dep_name, dep_api in api_to_deps.get(api.value, {}).items():
-        if dep_api in deps:
-            api_to_dep_impl[dep_name] = deps[dep_api]
+    if run_config.telemetry.enabled:
+        api_to_deps = {
+            "inference": {"telemetry": Api.telemetry},
+        }
+        for dep_name, dep_api in api_to_deps.get(api.value, {}).items():
+            if dep_api in deps:
+                api_to_dep_impl[dep_name] = deps[dep_api]
 
     # TODO: move pass configs to routers instead
     if api == Api.inference and run_config.inference_store:

llama_stack/core/server/server.py

@@ -445,7 +445,7 @@ def create_app() -> StackApp:
         if cors_config:
             app.add_middleware(CORSMiddleware, **cors_config.model_dump())
 
-    if Api.telemetry in impls:
+    if config.telemetry.enabled:
         setup_logger(impls[Api.telemetry])
     else:
         setup_logger(TelemetryAdapter(TelemetryConfig(), {}))