dsh-peakshift
DeepSeek Harness 错峰插件：高峰时段自动切换备用供应商，低谷切回 DeepSeek / A DeepSeek Harness plugin that shifts expensive models to a cheaper backup provider during peak pricing hours.
AI Analysis
核心用途是根据北京时间的高峰时段自动切换模型供应商以节省成本。适合高频使用 DeepSeek 且希望在拥堵或高价时段自动分流的用户。需在设置中配置备用供应商及 API 密钥。
Install
This plugin has no verified bundle, or compatibility checks failed. Read the repository notes first. Read the full README ↗
README
Read the full README ↗使用
- 「设置 → 模型」：添加一个便宜供应商（Moonshot、GLM、OpenAI 兼容网关……），填 API Key；
- 「设置 → 错峰」：勾选「错峰」，选好备用供应商/模型，贵模型列表默认
deepseek-v4-pro； - 完成。高峰时段请求自动走备用供应商，低谷自动用回 DeepSeek。
配置项（settings.yaml 的 price-router: 分节）
| 字段 | 默认 | 说明 |
|---|---|---|
enabled | false | 错峰总开关 |
primaryProvider | deepseek-official | 主供应商（低谷时段使用） |
fallbackProvider | 空 | 备用供应商（留空 = 自动选第一个已接入的） |
fallbackModel | 空 | 备用模型（留空 = 该供应商第一个模型） |
peakWindows | 09:00–12:00、14:00–18:00 | 高峰时段窗口（北京时间） |
expensiveModels | ["deepseek-v4-pro"] | 贵模型列表（高峰才切走的模型） |
remindEnabled / remindMinutes | true / 5 | 交界前提醒开关与提前分钟数 |
Usage
- Settings → Models: add a cheaper provider and fill in its API key;
- Settings → Peak shift: enable the toggle, pick the backup provider/model (expensive models default to
deepseek-v4-pro); - Done — peak-hour requests go to the backup provider, off-peak requests go back to DeepSeek.
Configuration (price-router: section in settings.yaml)
| Field | Default | Meaning |
|---|---|---|
enabled | false | Master toggle |
primaryProvider | deepseek-official | Primary provider (off-peak) |
fallbackProvider | empty | Backup provider (empty = first connected one) |
fallbackModel | empty | Backup model (empty = that provider's first model) |
peakWindows | 09:00–12:00, 14:00–18:00 | Peak windows (Beijing time) |
expensiveModels | ["deepseek-v4-pro"] | Models to shift away from at peak |
remindEnabled / remindMinutes | true / 5 | Transition reminder toggle / lead minutes |