phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-10-04 20:14:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: introduce API leveling, post_training to v1alpha

Browse source 
Rather than have a single `LLAMA_STACK_VERSION`, we need to have a `_V1`, `_V1ALPHA`, and `_V1BETA` constant.

This also necessitated addition of `level` to the `WebMethod` so that routing can be handeled properly.

For backwards compat, the `v1` routes are being kept around and marked as `deprecated`. When used, the server will log a deprecation warning.

move:

post_training to v1alpha as it is under heavy development and not near its final state
eval: job scheduling is not implemented. Relies heavily on the datasetio API which is under development missing implementations of specific routes indicating the structure of those routes might change. Additionally eval depends on the inference API which is going to be deprecated, eval will likely need a major API surface change to conform to using completions properly

Signed-off-by: Charlie Doern <cdoern@redhat.com>
This commit is contained in:
Charlie Doern 2025-09-12 13:23:57 -04:00
parent a50b63906c
commit 03399cebf3
35 changed files with 1507 additions and 260 deletions
Download patch file Download diff file Expand all files Collapse all files

30
llama_stack/core/server/routes.py
View file

@ -14,7 +14,6 @@ from starlette.routing import Route
from llama_stack.apis.datatypes import Api, ExternalApiSpec
from llama_stack.apis.tools import RAGToolRuntime, SpecialToolGroup
from llama_stack.apis.version import LLAMA_STACK_API_VERSION
from llama_stack.core.resolver import api_protocol_map
from llama_stack.schema_utils import WebMethod
@ -54,22 +53,23 @@ def get_all_api_routes(
protocol_methods.append((f"{tool_group.value}.{name}", method))
for name, method in protocol_methods:
if not hasattr(method, "__webmethod__"):
# Get all webmethods for this method (supports multiple decorators)
webmethods = getattr(method, "__webmethods__", [])
if not webmethods:
continue
# The __webmethod__ attribute is dynamically added by the @webmethod decorator
# mypy doesn't know about this dynamic attribute, so we ignore the attr-defined error
webmethod = method.__webmethod__ # type: ignore[attr-defined]
path = f"/{LLAMA_STACK_API_VERSION}/{webmethod.route.lstrip('/')}"
if webmethod.method == hdrs.METH_GET:
http_method = hdrs.METH_GET
elif webmethod.method == hdrs.METH_DELETE:
http_method = hdrs.METH_DELETE
else:
http_method = hdrs.METH_POST
routes.append(
(Route(path=path, methods=[http_method], name=name, endpoint=None), webmethod)
) # setting endpoint to None since don't use a Router object
# Create routes for each webmethod decorator
for webmethod in webmethods:
path = f"/{webmethod.level}/{webmethod.route.lstrip('/')}"
if webmethod.method == hdrs.METH_GET:
http_method = hdrs.METH_GET
elif webmethod.method == hdrs.METH_DELETE:
http_method = hdrs.METH_DELETE
else:
http_method = hdrs.METH_POST
routes.append(
(Route(path=path, methods=[http_method], name=name, endpoint=None), webmethod)
) # setting endpoint to None since don't use a Router object
apis[api] = routes