Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 23 additions & 0 deletions src/vllm_router/dynamic_config.py
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,11 @@ class DynamicRouterConfig:
static_backends: Optional[str] = None
static_models: Optional[str] = None
static_aliases: Optional[str] = None
static_model_labels: Optional[str] = None
static_model_types: Optional[str] = None
static_backend_health_checks: Optional[bool] = False
prefill_model_labels: Optional[str] = None
decode_model_labels: Optional[str] = None
k8s_port: Optional[int] = None
k8s_namespace: Optional[str] = None
k8s_label_selector: Optional[str] = None
Expand Down Expand Up @@ -137,15 +142,33 @@ def reconfigure_service_discovery(self, config: DynamicRouterConfig):
if config.service_discovery == "static":
reconfigure_service_discovery(
ServiceDiscoveryType.STATIC,
app=self.app,
urls=parse_static_urls(config.static_backends),
models=parse_comma_separated_args(config.static_models),
aliases=parse_comma_separated_args(config.static_aliases),
model_labels=parse_comma_separated_args(config.static_model_labels),
model_types=parse_comma_separated_args(config.static_model_types),
static_backend_health_checks=config.static_backend_health_checks,
prefill_model_labels=parse_comma_separated_args(
config.prefill_model_labels
),
decode_model_labels=parse_comma_separated_args(
config.decode_model_labels
),
)
elif config.service_discovery == "k8s":
reconfigure_service_discovery(
ServiceDiscoveryType.K8S,
app=self.app,
namespace=config.k8s_namespace,
port=config.k8s_port,
label_selector=config.k8s_label_selector,
prefill_model_labels=parse_comma_separated_args(
config.prefill_model_labels
),
decode_model_labels=parse_comma_separated_args(
config.decode_model_labels
),
)
else:
raise ValueError(
Expand Down
12 changes: 6 additions & 6 deletions src/vllm_router/service_discovery.py
Original file line number Diff line number Diff line change
Expand Up @@ -206,12 +206,12 @@ def __init__(
app,
urls: List[str],
models: List[str],
aliases: List[str] | None,
model_labels: List[str] | None,
model_types: List[str] | None,
static_backend_health_checks: bool,
prefill_model_labels: List[str] | None,
decode_model_labels: List[str] | None,
aliases: List[str] | None = None,
model_labels: List[str] | None = None,
model_types: List[str] | None = None,
static_backend_health_checks: bool = False,
prefill_model_labels: List[str] | None = None,
decode_model_labels: List[str] | None = None,
):
self.app = app
assert len(urls) == len(models), "URLs and models should have the same length"
Expand Down
5 changes: 4 additions & 1 deletion src/vllm_router/utils.py
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
import json
import re
import resource
from typing import Optional

import requests
from fastapi.requests import Request
Expand Down Expand Up @@ -132,7 +133,9 @@ def parse_static_urls(static_backends: str):
return backend_urls


def parse_comma_separated_args(comma_separated_string: str):
def parse_comma_separated_args(comma_separated_string: Optional[str]):
if comma_separated_string is None:
return None
return comma_separated_string.split(",")


Expand Down