llm-fallback

Multi-level provider fallback for same-named models in the DeepSeek Harness

AI 分析

核心用途是为核心模型配置多服务商备用链。适合对 API 稳定性要求极高、需要在主服务商故障时自动无缝切换的用户。

other

パッケージ: llm-fallback
バージョン: 0.1.0
最終更新: 2026/08/16

インストール

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:CanGeng/llm-fallback

ドキュメント

README 全文を読む ↗

Configuration

See cordis.patch.yml. The pools list maps a model id to its ordered provider chain. The config is validated when the plugin loads: an unknown alwaysPolicy, a pool without a non-empty model, or a pool whose providers is not an array throws immediately instead of failing silently later.

config:
  alwaysPolicy: reject
  alwaysMaxRetries: 2
  pools:
    - model: deepseek-v4-pro
      providers: [deepseek-official, ark-coding, ark-agent]