fix(backend): 部署韧性——模型自愈/就绪门禁/全局代理/启动诊断

- whisper: model.bin 截断/损坏时删目录重下重试一次,修「Unable to
  open file model.bin」死循环;mlx 同样按 config.json 判完整性
- /generate_note 加就绪门禁:本地转写引擎模型没下好直接拦截,返回
  reason=transcriber_model_not_ready,不让任务静默卡在首次下载
- 全局代理:新增 ProxyConfigManager(JSON 配置 + HTTP_PROXY env 兜底)
  + build_openai_client,统一注入代理到 LLM/Groq 客户端;yt-dlp 与
  youtube-transcript-api 也走代理
- build_openai_client 校验 api_key 非空,空 key 给「xxx 的 API Key
  未配置」而不是天书般的 Illegal header value b'Bearer '
- universal_gpt: 模型拒绝自定义 temperature(o1/o3/gpt-5 系列)时
  就地去掉参数重试,不消耗重试预算
- connect_test 改用真实 chat completion 而非 /v1/models 探测
- main.py: lifespan 拆 [startup 1/5..5/5] 分段日志 + 异常清晰定位
- /sys_health 重构为结构化返回 {backend,ffmpeg,db,whisper_model}

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
huangjianwu
2026-05-14 19:01:14 +08:00
co-authored by Claude Opus 4.7
parent 88d25f8cc1
commit 41f17592c2
16 changed files with 534 additions and 92 deletions
@@ -1,13 +1,13 @@
from typing import Optional, Union
from openai import OpenAI
from app.utils.logger import get_logger
from app.utils.openai_client import build_openai_client
logging= get_logger(__name__)
class OpenAICompatibleProvider:
def __init__(self, api_key: str, base_url: str, model: Union[str, None]=None):
self.client = OpenAI(api_key=api_key, base_url=base_url)
# build_openai_client:注入全局代理 + 校验 api_key 非空
self.client = build_openai_client(api_key, base_url, key_label="模型供应商的 API Key")
self.model = model
@property
@@ -15,17 +15,27 @@ class OpenAICompatibleProvider:
return self.client
@staticmethod
def test_connection(api_key: str, base_url: str) -> bool:
def test_connection(api_key: str, base_url: str, model: str) -> bool:
"""发一条最小化 chat completion 验证 key / base_url / model 三方都通。
为什么不用 client.models.list()
- 部分代理 / 自建供应商不实现 /v1/models(如某些 OpenAI 兼容网关)
- 部分供应商 key 在没有 inference 权限时 /v1/models 仍返回 200
最终用户跑的就是 chat.completions.create,所以直接测它最忠实。
max_tokens=1 + temperature=0 让请求开销 < 0.0001 美元、延迟 < 2s。
"""
try:
client = OpenAI(api_key=api_key, base_url=base_url)
model = client.models.list()
# for segment in model:
# print(segment)
# print(model)
logging.info("连通性测试成功")
client = build_openai_client(
api_key, base_url, key_label="模型供应商的 API Key", timeout=15.0,
)
client.chat.completions.create(
model=model,
messages=[{"role": "user", "content": "ping"}],
max_tokens=1,
temperature=0,
)
logging.info(f"连通性测试成功(model={model}")
return True
except Exception as e:
logging.info(f"连通性测试失败:{e}")
# print(f"Error connecting to OpenAI API: {e}")
logging.warning(f"连通性测试失败model={model}{e}")
return False