llama-stack/llama_stack/distribution/routers/routing_tables.py
Ben Browning 2b2db5fbda
feat: OpenAI-Compatible models, completions, chat/completions (#1894)
# What does this PR do?

This stubs in some OpenAI server-side compatibility with three new
endpoints:

/v1/openai/v1/models
/v1/openai/v1/completions
/v1/openai/v1/chat/completions

This gives common inference apps using OpenAI clients the ability to
talk to Llama Stack using an endpoint like
http://localhost:8321/v1/openai/v1 .

The two "v1" instances in there isn't awesome, but the thinking is that
Llama Stack's API is v1 and then our OpenAI compatibility layer is
compatible with OpenAI V1. And, some OpenAI clients implicitly assume
the URL ends with "v1", so this gives maximum compatibility.

The openai models endpoint is implemented in the routing layer, and just
returns all the models Llama Stack knows about.

The following providers should be working with the new OpenAI
completions and chat/completions API:
* remote::anthropic (untested)
* remote::cerebras-openai-compat (untested)
* remote::fireworks (tested)
* remote::fireworks-openai-compat (untested)
* remote::gemini (untested)
* remote::groq-openai-compat (untested)
* remote::nvidia (tested)
* remote::ollama (tested)
* remote::openai (untested)
* remote::passthrough (untested)
* remote::sambanova-openai-compat (untested)
* remote::together (tested)
* remote::together-openai-compat (untested)
* remote::vllm (tested)

The goal to support this for every inference provider - proxying
directly to the provider's OpenAI endpoint for OpenAI-compatible
providers. For providers that don't have an OpenAI-compatible API, we'll
add a mixin to translate incoming OpenAI requests to Llama Stack
inference requests and translate the Llama Stack inference responses to
OpenAI responses.

This is related to #1817 but is a bit larger in scope than just chat
completions, as I have real use-cases that need the older completions
API as well.

## Test Plan

### vLLM

```
VLLM_URL="http://localhost:8000/v1" INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct" llama stack build --template remote-vllm --image-type venv --run

LLAMA_STACK_CONFIG=http://localhost:8321 INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct" python -m pytest -v tests/integration/inference/test_openai_completion.py --text-model "meta-llama/Llama-3.2-3B-Instruct"
```

### ollama
```
INFERENCE_MODEL="llama3.2:3b-instruct-q8_0" llama stack build --template ollama --image-type venv --run

LLAMA_STACK_CONFIG=http://localhost:8321 INFERENCE_MODEL="llama3.2:3b-instruct-q8_0" python -m pytest -v tests/integration/inference/test_openai_completion.py --text-model "llama3.2:3b-instruct-q8_0"
```



## Documentation

Run a Llama Stack distribution that uses one of the providers mentioned
in the list above. Then, use your favorite OpenAI client to send
completion or chat completion requests with the base_url set to
http://localhost:8321/v1/openai/v1 . Replace "localhost:8321" with the
host and port of your Llama Stack server, if different.

---------

Signed-off-by: Ben Browning <bbrownin@redhat.com>
2025-04-11 13:14:17 -07:00

631 lines
25 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.
import logging
import time
import uuid
from typing import Any, Dict, List, Optional
from pydantic import TypeAdapter
from llama_stack.apis.benchmarks import Benchmark, Benchmarks, ListBenchmarksResponse
from llama_stack.apis.common.content_types import URL
from llama_stack.apis.common.type_system import ParamType
from llama_stack.apis.datasets import (
Dataset,
DatasetPurpose,
Datasets,
DatasetType,
DataSource,
ListDatasetsResponse,
RowsDataSource,
URIDataSource,
)
from llama_stack.apis.models import ListModelsResponse, Model, Models, ModelType, OpenAIListModelsResponse, OpenAIModel
from llama_stack.apis.resource import ResourceType
from llama_stack.apis.scoring_functions import (
ListScoringFunctionsResponse,
ScoringFn,
ScoringFnParams,
ScoringFunctions,
)
from llama_stack.apis.shields import ListShieldsResponse, Shield, Shields
from llama_stack.apis.tools import (
ListToolGroupsResponse,
ListToolsResponse,
Tool,
ToolGroup,
ToolGroups,
ToolHost,
)
from llama_stack.apis.vector_dbs import ListVectorDBsResponse, VectorDB, VectorDBs
from llama_stack.distribution.access_control import check_access
from llama_stack.distribution.datatypes import (
AccessAttributes,
BenchmarkWithACL,
DatasetWithACL,
ModelWithACL,
RoutableObject,
RoutableObjectWithProvider,
RoutedProtocol,
ScoringFnWithACL,
ShieldWithACL,
ToolGroupWithACL,
ToolWithACL,
VectorDBWithACL,
)
from llama_stack.distribution.request_headers import get_auth_attributes
from llama_stack.distribution.store import DistributionRegistry
from llama_stack.providers.datatypes import Api, RoutingTable
logger = logging.getLogger(__name__)
def get_impl_api(p: Any) -> Api:
return p.__provider_spec__.api
# TODO: this should return the registered object for all APIs
async def register_object_with_provider(obj: RoutableObject, p: Any) -> RoutableObject:
api = get_impl_api(p)
assert obj.provider_id != "remote", "Remote provider should not be registered"
if api == Api.inference:
return await p.register_model(obj)
elif api == Api.safety:
return await p.register_shield(obj)
elif api == Api.vector_io:
return await p.register_vector_db(obj)
elif api == Api.datasetio:
return await p.register_dataset(obj)
elif api == Api.scoring:
return await p.register_scoring_function(obj)
elif api == Api.eval:
return await p.register_benchmark(obj)
elif api == Api.tool_runtime:
return await p.register_tool(obj)
else:
raise ValueError(f"Unknown API {api} for registering object with provider")
async def unregister_object_from_provider(obj: RoutableObject, p: Any) -> None:
api = get_impl_api(p)
if api == Api.vector_io:
return await p.unregister_vector_db(obj.identifier)
elif api == Api.inference:
return await p.unregister_model(obj.identifier)
elif api == Api.datasetio:
return await p.unregister_dataset(obj.identifier)
elif api == Api.tool_runtime:
return await p.unregister_tool(obj.identifier)
else:
raise ValueError(f"Unregister not supported for {api}")
Registry = Dict[str, List[RoutableObjectWithProvider]]
class CommonRoutingTableImpl(RoutingTable):
def __init__(
self,
impls_by_provider_id: Dict[str, RoutedProtocol],
dist_registry: DistributionRegistry,
) -> None:
self.impls_by_provider_id = impls_by_provider_id
self.dist_registry = dist_registry
async def initialize(self) -> None:
async def add_objects(objs: List[RoutableObjectWithProvider], provider_id: str, cls) -> None:
for obj in objs:
if cls is None:
obj.provider_id = provider_id
else:
# Create a copy of the model data and explicitly set provider_id
model_data = obj.model_dump()
model_data["provider_id"] = provider_id
obj = cls(**model_data)
await self.dist_registry.register(obj)
# Register all objects from providers
for pid, p in self.impls_by_provider_id.items():
api = get_impl_api(p)
if api == Api.inference:
p.model_store = self
elif api == Api.safety:
p.shield_store = self
elif api == Api.vector_io:
p.vector_db_store = self
elif api == Api.datasetio:
p.dataset_store = self
elif api == Api.scoring:
p.scoring_function_store = self
scoring_functions = await p.list_scoring_functions()
await add_objects(scoring_functions, pid, ScoringFn)
elif api == Api.eval:
p.benchmark_store = self
elif api == Api.tool_runtime:
p.tool_store = self
async def shutdown(self) -> None:
for p in self.impls_by_provider_id.values():
await p.shutdown()
def get_provider_impl(self, routing_key: str, provider_id: Optional[str] = None) -> Any:
def apiname_object():
if isinstance(self, ModelsRoutingTable):
return ("Inference", "model")
elif isinstance(self, ShieldsRoutingTable):
return ("Safety", "shield")
elif isinstance(self, VectorDBsRoutingTable):
return ("VectorIO", "vector_db")
elif isinstance(self, DatasetsRoutingTable):
return ("DatasetIO", "dataset")
elif isinstance(self, ScoringFunctionsRoutingTable):
return ("Scoring", "scoring_function")
elif isinstance(self, BenchmarksRoutingTable):
return ("Eval", "benchmark")
elif isinstance(self, ToolGroupsRoutingTable):
return ("Tools", "tool")
else:
raise ValueError("Unknown routing table type")
apiname, objtype = apiname_object()
# Get objects from disk registry
obj = self.dist_registry.get_cached(objtype, routing_key)
if not obj:
provider_ids = list(self.impls_by_provider_id.keys())
if len(provider_ids) > 1:
provider_ids_str = f"any of the providers: {', '.join(provider_ids)}"
else:
provider_ids_str = f"provider: `{provider_ids[0]}`"
raise ValueError(
f"{objtype.capitalize()} `{routing_key}` not served by {provider_ids_str}. Make sure there is an {apiname} provider serving this {objtype}."
)
if not provider_id or provider_id == obj.provider_id:
return self.impls_by_provider_id[obj.provider_id]
raise ValueError(f"Provider not found for `{routing_key}`")
async def get_object_by_identifier(self, type: str, identifier: str) -> Optional[RoutableObjectWithProvider]:
# Get from disk registry
obj = await self.dist_registry.get(type, identifier)
if not obj:
return None
# Check if user has permission to access this object
if not check_access(obj.identifier, getattr(obj, "access_attributes", None), get_auth_attributes()):
logger.debug(f"Access denied to {type} '{identifier}' based on attribute mismatch")
return None
return obj
async def unregister_object(self, obj: RoutableObjectWithProvider) -> None:
await self.dist_registry.delete(obj.type, obj.identifier)
await unregister_object_from_provider(obj, self.impls_by_provider_id[obj.provider_id])
async def register_object(self, obj: RoutableObjectWithProvider) -> RoutableObjectWithProvider:
# if provider_id is not specified, pick an arbitrary one from existing entries
if not obj.provider_id and len(self.impls_by_provider_id) > 0:
obj.provider_id = list(self.impls_by_provider_id.keys())[0]
if obj.provider_id not in self.impls_by_provider_id:
raise ValueError(f"Provider `{obj.provider_id}` not found")
p = self.impls_by_provider_id[obj.provider_id]
# If object supports access control but no attributes set, use creator's attributes
if not obj.access_attributes:
creator_attributes = get_auth_attributes()
if creator_attributes:
obj.access_attributes = AccessAttributes(**creator_attributes)
logger.info(f"Setting access attributes for {obj.type} '{obj.identifier}' based on creator's identity")
registered_obj = await register_object_with_provider(obj, p)
# TODO: This needs to be fixed for all APIs once they return the registered object
if obj.type == ResourceType.model.value:
await self.dist_registry.register(registered_obj)
return registered_obj
else:
await self.dist_registry.register(obj)
return obj
async def get_all_with_type(self, type: str) -> List[RoutableObjectWithProvider]:
objs = await self.dist_registry.get_all()
filtered_objs = [obj for obj in objs if obj.type == type]
# Apply attribute-based access control filtering
if filtered_objs:
filtered_objs = [
obj
for obj in filtered_objs
if check_access(obj.identifier, getattr(obj, "access_attributes", None), get_auth_attributes())
]
return filtered_objs
class ModelsRoutingTable(CommonRoutingTableImpl, Models):
async def list_models(self) -> ListModelsResponse:
return ListModelsResponse(data=await self.get_all_with_type("model"))
async def openai_list_models(self) -> OpenAIListModelsResponse:
models = await self.get_all_with_type("model")
openai_models = [
OpenAIModel(
id=model.identifier,
object="model",
created=int(time.time()),
owned_by="llama_stack",
)
for model in models
]
return OpenAIListModelsResponse(data=openai_models)
async def get_model(self, model_id: str) -> Model:
model = await self.get_object_by_identifier("model", model_id)
if model is None:
raise ValueError(f"Model '{model_id}' not found")
return model
async def register_model(
self,
model_id: str,
provider_model_id: Optional[str] = None,
provider_id: Optional[str] = None,
metadata: Optional[Dict[str, Any]] = None,
model_type: Optional[ModelType] = None,
) -> Model:
if provider_model_id is None:
provider_model_id = model_id
if provider_id is None:
# If provider_id not specified, use the only provider if it supports this model
if len(self.impls_by_provider_id) == 1:
provider_id = list(self.impls_by_provider_id.keys())[0]
else:
raise ValueError(
f"No provider specified and multiple providers available. Please specify a provider_id. Available providers: {self.impls_by_provider_id.keys()}"
)
if metadata is None:
metadata = {}
if model_type is None:
model_type = ModelType.llm
if "embedding_dimension" not in metadata and model_type == ModelType.embedding:
raise ValueError("Embedding model must have an embedding dimension in its metadata")
model = ModelWithACL(
identifier=model_id,
provider_resource_id=provider_model_id,
provider_id=provider_id,
metadata=metadata,
model_type=model_type,
)
registered_model = await self.register_object(model)
return registered_model
async def unregister_model(self, model_id: str) -> None:
existing_model = await self.get_model(model_id)
if existing_model is None:
raise ValueError(f"Model {model_id} not found")
await self.unregister_object(existing_model)
class ShieldsRoutingTable(CommonRoutingTableImpl, Shields):
async def list_shields(self) -> ListShieldsResponse:
return ListShieldsResponse(data=await self.get_all_with_type(ResourceType.shield.value))
async def get_shield(self, identifier: str) -> Shield:
shield = await self.get_object_by_identifier("shield", identifier)
if shield is None:
raise ValueError(f"Shield '{identifier}' not found")
return shield
async def register_shield(
self,
shield_id: str,
provider_shield_id: Optional[str] = None,
provider_id: Optional[str] = None,
params: Optional[Dict[str, Any]] = None,
) -> Shield:
if provider_shield_id is None:
provider_shield_id = shield_id
if provider_id is None:
# If provider_id not specified, use the only provider if it supports this shield type
if len(self.impls_by_provider_id) == 1:
provider_id = list(self.impls_by_provider_id.keys())[0]
else:
raise ValueError(
"No provider specified and multiple providers available. Please specify a provider_id."
)
if params is None:
params = {}
shield = ShieldWithACL(
identifier=shield_id,
provider_resource_id=provider_shield_id,
provider_id=provider_id,
params=params,
)
await self.register_object(shield)
return shield
class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
async def list_vector_dbs(self) -> ListVectorDBsResponse:
return ListVectorDBsResponse(data=await self.get_all_with_type("vector_db"))
async def get_vector_db(self, vector_db_id: str) -> VectorDB:
vector_db = await self.get_object_by_identifier("vector_db", vector_db_id)
if vector_db is None:
raise ValueError(f"Vector DB '{vector_db_id}' not found")
return vector_db
async def register_vector_db(
self,
vector_db_id: str,
embedding_model: str,
embedding_dimension: Optional[int] = 384,
provider_id: Optional[str] = None,
provider_vector_db_id: Optional[str] = None,
) -> VectorDB:
if provider_vector_db_id is None:
provider_vector_db_id = vector_db_id
if provider_id is None:
if len(self.impls_by_provider_id) > 0:
provider_id = list(self.impls_by_provider_id.keys())[0]
if len(self.impls_by_provider_id) > 1:
logger.warning(
f"No provider specified and multiple providers available. Arbitrarily selected the first provider {provider_id}."
)
else:
raise ValueError("No provider available. Please configure a vector_io provider.")
model = await self.get_object_by_identifier("model", embedding_model)
if model is None:
raise ValueError(f"Model {embedding_model} not found")
if model.model_type != ModelType.embedding:
raise ValueError(f"Model {embedding_model} is not an embedding model")
if "embedding_dimension" not in model.metadata:
raise ValueError(f"Model {embedding_model} does not have an embedding dimension")
vector_db_data = {
"identifier": vector_db_id,
"type": ResourceType.vector_db.value,
"provider_id": provider_id,
"provider_resource_id": provider_vector_db_id,
"embedding_model": embedding_model,
"embedding_dimension": model.metadata["embedding_dimension"],
}
vector_db = TypeAdapter(VectorDBWithACL).validate_python(vector_db_data)
await self.register_object(vector_db)
return vector_db
async def unregister_vector_db(self, vector_db_id: str) -> None:
existing_vector_db = await self.get_vector_db(vector_db_id)
if existing_vector_db is None:
raise ValueError(f"Vector DB {vector_db_id} not found")
await self.unregister_object(existing_vector_db)
class DatasetsRoutingTable(CommonRoutingTableImpl, Datasets):
async def list_datasets(self) -> ListDatasetsResponse:
return ListDatasetsResponse(data=await self.get_all_with_type(ResourceType.dataset.value))
async def get_dataset(self, dataset_id: str) -> Dataset:
dataset = await self.get_object_by_identifier("dataset", dataset_id)
if dataset is None:
raise ValueError(f"Dataset '{dataset_id}' not found")
return dataset
async def register_dataset(
self,
purpose: DatasetPurpose,
source: DataSource,
metadata: Optional[Dict[str, Any]] = None,
dataset_id: Optional[str] = None,
) -> Dataset:
if isinstance(source, dict):
if source["type"] == "uri":
source = URIDataSource.parse_obj(source)
elif source["type"] == "rows":
source = RowsDataSource.parse_obj(source)
if not dataset_id:
dataset_id = f"dataset-{str(uuid.uuid4())}"
provider_dataset_id = dataset_id
# infer provider from source
if source.type == DatasetType.rows.value:
provider_id = "localfs"
elif source.type == DatasetType.uri.value:
# infer provider from uri
if source.uri.startswith("huggingface"):
provider_id = "huggingface"
else:
provider_id = "localfs"
else:
raise ValueError(f"Unknown data source type: {source.type}")
if metadata is None:
metadata = {}
dataset = DatasetWithACL(
identifier=dataset_id,
provider_resource_id=provider_dataset_id,
provider_id=provider_id,
purpose=purpose,
source=source,
metadata=metadata,
)
await self.register_object(dataset)
return dataset
async def unregister_dataset(self, dataset_id: str) -> None:
dataset = await self.get_dataset(dataset_id)
if dataset is None:
raise ValueError(f"Dataset {dataset_id} not found")
await self.unregister_object(dataset)
class ScoringFunctionsRoutingTable(CommonRoutingTableImpl, ScoringFunctions):
async def list_scoring_functions(self) -> ListScoringFunctionsResponse:
return ListScoringFunctionsResponse(data=await self.get_all_with_type(ResourceType.scoring_function.value))
async def get_scoring_function(self, scoring_fn_id: str) -> ScoringFn:
scoring_fn = await self.get_object_by_identifier("scoring_function", scoring_fn_id)
if scoring_fn is None:
raise ValueError(f"Scoring function '{scoring_fn_id}' not found")
return scoring_fn
async def register_scoring_function(
self,
scoring_fn_id: str,
description: str,
return_type: ParamType,
provider_scoring_fn_id: Optional[str] = None,
provider_id: Optional[str] = None,
params: Optional[ScoringFnParams] = None,
) -> None:
if provider_scoring_fn_id is None:
provider_scoring_fn_id = scoring_fn_id
if provider_id is None:
if len(self.impls_by_provider_id) == 1:
provider_id = list(self.impls_by_provider_id.keys())[0]
else:
raise ValueError(
"No provider specified and multiple providers available. Please specify a provider_id."
)
scoring_fn = ScoringFnWithACL(
identifier=scoring_fn_id,
description=description,
return_type=return_type,
provider_resource_id=provider_scoring_fn_id,
provider_id=provider_id,
params=params,
)
scoring_fn.provider_id = provider_id
await self.register_object(scoring_fn)
class BenchmarksRoutingTable(CommonRoutingTableImpl, Benchmarks):
async def list_benchmarks(self) -> ListBenchmarksResponse:
return ListBenchmarksResponse(data=await self.get_all_with_type("benchmark"))
async def get_benchmark(self, benchmark_id: str) -> Benchmark:
benchmark = await self.get_object_by_identifier("benchmark", benchmark_id)
if benchmark is None:
raise ValueError(f"Benchmark '{benchmark_id}' not found")
return benchmark
async def register_benchmark(
self,
benchmark_id: str,
dataset_id: str,
scoring_functions: List[str],
metadata: Optional[Dict[str, Any]] = None,
provider_benchmark_id: Optional[str] = None,
provider_id: Optional[str] = None,
) -> None:
if metadata is None:
metadata = {}
if provider_id is None:
if len(self.impls_by_provider_id) == 1:
provider_id = list(self.impls_by_provider_id.keys())[0]
else:
raise ValueError(
"No provider specified and multiple providers available. Please specify a provider_id."
)
if provider_benchmark_id is None:
provider_benchmark_id = benchmark_id
benchmark = BenchmarkWithACL(
identifier=benchmark_id,
dataset_id=dataset_id,
scoring_functions=scoring_functions,
metadata=metadata,
provider_id=provider_id,
provider_resource_id=provider_benchmark_id,
)
await self.register_object(benchmark)
class ToolGroupsRoutingTable(CommonRoutingTableImpl, ToolGroups):
async def list_tools(self, toolgroup_id: Optional[str] = None) -> ListToolsResponse:
tools = await self.get_all_with_type("tool")
if toolgroup_id:
tools = [tool for tool in tools if tool.toolgroup_id == toolgroup_id]
return ListToolsResponse(data=tools)
async def list_tool_groups(self) -> ListToolGroupsResponse:
return ListToolGroupsResponse(data=await self.get_all_with_type("tool_group"))
async def get_tool_group(self, toolgroup_id: str) -> ToolGroup:
tool_group = await self.get_object_by_identifier("tool_group", toolgroup_id)
if tool_group is None:
raise ValueError(f"Tool group '{toolgroup_id}' not found")
return tool_group
async def get_tool(self, tool_name: str) -> Tool:
return await self.get_object_by_identifier("tool", tool_name)
async def register_tool_group(
self,
toolgroup_id: str,
provider_id: str,
mcp_endpoint: Optional[URL] = None,
args: Optional[Dict[str, Any]] = None,
) -> None:
tools = []
tool_defs = await self.impls_by_provider_id[provider_id].list_runtime_tools(toolgroup_id, mcp_endpoint)
tool_host = ToolHost.model_context_protocol if mcp_endpoint else ToolHost.distribution
for tool_def in tool_defs.data:
tools.append(
ToolWithACL(
identifier=tool_def.name,
toolgroup_id=toolgroup_id,
description=tool_def.description or "",
parameters=tool_def.parameters or [],
provider_id=provider_id,
provider_resource_id=tool_def.name,
metadata=tool_def.metadata,
tool_host=tool_host,
)
)
for tool in tools:
existing_tool = await self.get_tool(tool.identifier)
# Compare existing and new object if one exists
if existing_tool:
existing_dict = existing_tool.model_dump()
new_dict = tool.model_dump()
if existing_dict != new_dict:
raise ValueError(
f"Object {tool.identifier} already exists in registry. Please use a different identifier."
)
await self.register_object(tool)
await self.dist_registry.register(
ToolGroupWithACL(
identifier=toolgroup_id,
provider_id=provider_id,
provider_resource_id=toolgroup_id,
mcp_endpoint=mcp_endpoint,
args=args,
)
)
async def unregister_toolgroup(self, toolgroup_id: str) -> None:
tool_group = await self.get_tool_group(toolgroup_id)
if tool_group is None:
raise ValueError(f"Tool group {toolgroup_id} not found")
tools = await self.list_tools(toolgroup_id)
for tool in getattr(tools, "data", []):
await self.unregister_object(tool)
await self.unregister_object(tool_group)
async def shutdown(self) -> None:
pass