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
@@ -61,3 +61,54 @@ class TranscriberConfigManager:
def get_whisper_model_size(self) -> str:
return self.get_config()["whisper_model_size"]
def is_model_ready(self) -> Dict[str, Any]:
"""当前转写器是否就绪可用。
返回 {ready, transcriber_type, model_size, downloading, reason}
- 在线引擎 (groq/bcut/kuaishou):永远 ready(不需要本地模型)
- fast-whisper:检查 whisper-{size}/model.bin 落盘
- mlx-whisper:检查 {repo_id}/config.json 落盘
给 /generate_note 入口做「开始视频前先确认模型下载好」的门禁用。
"""
cfg = self.get_config()
ttype = cfg["transcriber_type"]
size = cfg["whisper_model_size"]
result = {
"ready": True,
"transcriber_type": ttype,
"model_size": size,
"downloading": False,
"reason": "",
}
if ttype not in ("fast-whisper", "mlx-whisper"):
return result # 在线引擎无需本地模型
# 延迟 import 避免与 routers.config 的循环依赖;只取纯函数,不触发路由副作用
try:
from app.routers.config import (
_check_whisper_model_exists,
_check_mlx_whisper_model_exists,
_downloading,
)
except Exception as e:
# 拿不到检查函数时保守放行,不要把用户卡死
result["reason"] = f"无法检查模型状态: {e}"
return result
if ttype == "fast-whisper":
downloaded = _check_whisper_model_exists(size, "whisper")
downloading = _downloading.get(size) == "downloading"
else: # mlx-whisper
downloaded = _check_mlx_whisper_model_exists(size)
downloading = _downloading.get(f"mlx-{size}") == "downloading"
result["downloading"] = downloading
if downloaded:
return result
result["ready"] = False
result["reason"] = (
f"转写模型 {ttype} / {size} 尚未下载就绪"
+ (",正在下载中,请稍候" if downloading else ",请先在「设置 → 音频转写配置」页下载")
)
return result