llama-stack-mirror/src/llama_stack/apis/models/models.py
Charlie Doern 9df073450f
Some checks failed
Integration Tests (Replay) / generate-matrix (push) Successful in 3s
Test External API and Providers / test-external (venv) (push) Failing after 4s
UI Tests / ui-tests (22) (push) Successful in 55s
SqlStore Integration Tests / test-postgres (3.12) (push) Failing after 1s
Integration Auth Tests / test-matrix (oauth2_token) (push) Failing after 1s
Test External Providers Installed via Module / test-external-providers-from-module (venv) (push) Has been skipped
Python Package Build Test / build (3.12) (push) Failing after 1s
Pre-commit / pre-commit (push) Failing after 2s
Python Package Build Test / build (3.13) (push) Failing after 1s
SqlStore Integration Tests / test-postgres (3.13) (push) Failing after 5s
Vector IO Integration Tests / test-matrix (push) Failing after 5s
API Conformance Tests / check-schema-compatibility (push) Successful in 11s
Unit Tests / unit-tests (3.12) (push) Failing after 4s
Integration Tests (Replay) / Integration Tests (, , , client=, ) (push) Failing after 4s
Unit Tests / unit-tests (3.13) (push) Failing after 5s
feat: remove core.telemetry as a dependency of llama_stack.apis (#4064)
# What does this PR do?

Remove circular dependency by moving tracing from API protocol
definitions
 to router implementation layer.

This gets us closer to having a self contained API package with no other
cross-cutting dependencies to other parts of the llama stack codebase.
To the best of our ability, the llama_stack.api should only be type and
protocol definitions.

  Changes:
- Create apis/common/tracing.py with marker decorator (zero core
dependencies)
- Add the _new_ `@telemetry_traceable` marker decorator to 11 protocol
classes
- Apply actual tracing in core/resolver.py in `instantiate_provider`
based on protocol marker
- Move MetricResponseMixin from core to apis (it's an API response type)
  - APIs package is now self-contained with zero core dependencies

The tracing functionality remains identical - actual trace_protocol from
core
is applied to router implementations at runtime when both telemetry is
enabled
  and the protocol has the `__marked_for_tracing__` marker.

  ## Test Plan

  Manual integration test confirms identical behavior to main branch:

  ```bash
  llama stack list-deps --format uv starter | sh
  export OLLAMA_URL=http://localhost:11434
  llama stack run starter

  curl -X POST http://localhost:8321/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{"model": "ollama/gpt-oss:20b",
         "messages": [{"role": "user", "content": "Say hello"}],
         "max_tokens": 10}'
         
```

  Verified identical between main and this branch:
  - trace_id present in response
  - metrics array with prompt_tokens, completion_tokens, total_tokens
  - Server logs show trace_protocol applied to all routers

  Existing telemetry integration tests (tests/integration/telemetry/) validate
  trace context propagation and span attributes.


relates to #3895

---------

Signed-off-by: Charlie Doern <cdoern@redhat.com>
2025-11-06 10:58:30 -08:00

172 lines
5.1 KiB
Python

# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
from enum import StrEnum
from typing import Any, Literal, Protocol, runtime_checkable
from pydantic import BaseModel, ConfigDict, Field, field_validator
from llama_stack.apis.common.tracing import telemetry_traceable
from llama_stack.apis.resource import Resource, ResourceType
from llama_stack.apis.version import LLAMA_STACK_API_V1
from llama_stack.schema_utils import json_schema_type, webmethod
class CommonModelFields(BaseModel):
metadata: dict[str, Any] = Field(
default_factory=dict,
description="Any additional metadata for this model",
)
@json_schema_type
class ModelType(StrEnum):
"""Enumeration of supported model types in Llama Stack.
:cvar llm: Large language model for text generation and completion
:cvar embedding: Embedding model for converting text to vector representations
:cvar rerank: Reranking model for reordering documents based on their relevance to a query
"""
llm = "llm"
embedding = "embedding"
rerank = "rerank"
@json_schema_type
class Model(CommonModelFields, Resource):
"""A model resource representing an AI model registered in Llama Stack.
:param type: The resource type, always 'model' for model resources
:param model_type: The type of model (LLM or embedding model)
:param metadata: Any additional metadata for this model
:param identifier: Unique identifier for this resource in llama stack
:param provider_resource_id: Unique identifier for this resource in the provider
:param provider_id: ID of the provider that owns this resource
"""
type: Literal[ResourceType.model] = ResourceType.model
@property
def model_id(self) -> str:
return self.identifier
@property
def provider_model_id(self) -> str:
assert self.provider_resource_id is not None, "Provider resource ID must be set"
return self.provider_resource_id
model_config = ConfigDict(protected_namespaces=())
model_type: ModelType = Field(default=ModelType.llm)
@field_validator("provider_resource_id")
@classmethod
def validate_provider_resource_id(cls, v):
if v is None:
raise ValueError("provider_resource_id cannot be None")
return v
class ModelInput(CommonModelFields):
model_id: str
provider_id: str | None = None
provider_model_id: str | None = None
model_type: ModelType | None = ModelType.llm
model_config = ConfigDict(protected_namespaces=())
class ListModelsResponse(BaseModel):
data: list[Model]
@json_schema_type
class OpenAIModel(BaseModel):
"""A model from OpenAI.
:id: The ID of the model
:object: The object type, which will be "model"
:created: The Unix timestamp in seconds when the model was created
:owned_by: The owner of the model
:custom_metadata: Llama Stack-specific metadata including model_type, provider info, and additional metadata
"""
id: str
object: Literal["model"] = "model"
created: int
owned_by: str
custom_metadata: dict[str, Any] | None = None
class OpenAIListModelsResponse(BaseModel):
data: list[OpenAIModel]
@runtime_checkable
@telemetry_traceable
class Models(Protocol):
async def list_models(self) -> ListModelsResponse:
"""List all models.
:returns: A ListModelsResponse.
"""
...
@webmethod(route="/models", method="GET", level=LLAMA_STACK_API_V1)
async def openai_list_models(self) -> OpenAIListModelsResponse:
"""List models using the OpenAI API.
:returns: A OpenAIListModelsResponse.
"""
...
@webmethod(route="/models/{model_id:path}", method="GET", level=LLAMA_STACK_API_V1)
async def get_model(
self,
model_id: str,
) -> Model:
"""Get model.
Get a model by its identifier.
:param model_id: The identifier of the model to get.
:returns: A Model.
"""
...
@webmethod(route="/models", method="POST", level=LLAMA_STACK_API_V1)
async def register_model(
self,
model_id: str,
provider_model_id: str | None = None,
provider_id: str | None = None,
metadata: dict[str, Any] | None = None,
model_type: ModelType | None = None,
) -> Model:
"""Register model.
Register a model.
:param model_id: The identifier of the model to register.
:param provider_model_id: The identifier of the model in the provider.
:param provider_id: The identifier of the provider.
:param metadata: Any additional metadata for this model.
:param model_type: The type of model to register.
:returns: A Model.
"""
...
@webmethod(route="/models/{model_id:path}", method="DELETE", level=LLAMA_STACK_API_V1)
async def unregister_model(
self,
model_id: str,
) -> None:
"""Unregister model.
Unregister a model.
:param model_id: The identifier of the model to unregister.
"""
...