BIZ-46: Phase3 架构设计 — SidecarContext解耦/Prometheus治理/部署支撑/测试/UX

Co-authored-by: multica-agent <github@multica.ai>
This commit is contained in:
2026-06-24 20:01:25 +08:00
parent 4fd89b038d
commit 8a12ff9693
13 changed files with 3502 additions and 0 deletions
+41
View File
@@ -0,0 +1,41 @@
"""
NVIDIA Sidecar 限流代理 — 核心代理模块。
为 OpenAI Chat Completions 兼容 API 提供四层防护:
1. 请求接收(FastAPI
2. 网关识别 → 非 NVIDIA 直通
3. 优先级排队 → 令牌桶限流
4. httpx 异步转发到 NVIDIA 上游
"""
from __future__ import annotations
from nvidia_sidecar.config import SidecarConfig, load_config
from nvidia_sidecar.rate_limiter import (
Priority,
TokenBucket,
is_nvidia_gateway,
normalize_gateway_name,
)
from nvidia_sidecar.priority_queue import (
PriorityQueueItem,
PriorityRequestQueue,
QueueFullError,
QueueFullPassthrough,
QueueFullPolicy,
)
__version__ = "0.1.0"
__all__ = [
"SidecarConfig",
"load_config",
"Priority",
"TokenBucket",
"is_nvidia_gateway",
"normalize_gateway_name",
"PriorityQueueItem",
"PriorityRequestQueue",
"QueueFullError",
"QueueFullPassthrough",
"QueueFullPolicy",
]