dsh-llm-mimo
Xiaomi MiMo v2.5 adapter for DeepSeek Harness LLM seam with thinking chain support
安装
$
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:dfhxxc666/dsh-llm-mimo说明文档
阅读完整 README ↗3. 配置 cordis.patch.yml 激活块
dsh-llm-mimo 是非 bundle 型插件(package.json 无
dsh.bundle),不能加进dsh.profile.bundles(会报declares no dsh.bundle),只能通过 profile 的cordis.patch.yml的insert块注册进 LLM seam。
在 profile 的 cordis.patch.yml 追加:
- insert:
- id: llm-mimo
name: 'dsh-llm-mimo'
config:
apiKeyEnv: MIMO_API_KEY
baseURL: https://api.xiaomimimo.com/v1
enableThinking: true
maxTokens: 131072
defaultContextWindow: 1000000
models:
- id: mimo-v2.5-pro
name: MiMo-V2.5-Pro
contextWindow: 1000000
- id: mimo-v2.5
name: MiMo-V2.5
contextWindow: 1000000
streamIdleTimeoutMs: 300000
4. 配置 API Key(密钥,绝不提交到仓库)
MiMo API Key 通过环境变量名引用(apiKeyEnv: MIMO_API_KEY),插件在请求时从
DSH 的凭据服务解析,不会硬编码、不会写入仓库。
方式 A:环境变量(命令行启动 dsh 时)
# 1. 配置树包含 llm-mimo
dsh --profile web --dump-config | grep -A 3 llm-mimo
## Configuration Options
| Option | Type | Default | Description |
|--------|------|---------|-------------|
| `apiKeyEnv` | `string` | `MIMO_API_KEY` | Credential reference (environment-variable name) for the API key |
| `baseURL` | `string` | `https://api.xiaomimimo.com/v1` | API endpoint base URL |
| `enableThinking` | `boolean` | `true` | Whether thinking mode is enabled by default |
| `maxTokens` | `number` | `131072` | Default per-request output token cap (MiMo hard limit) |
| `defaultContextWindow` | `number` | `1000000` | Default context window size |
| `models` | `MiMoCatalogModel[]` | V2.5, V2.5-Pro | Advisory model catalog |
| `streamIdleTimeoutMs` | `number` | `300000` | Stream idle timeout (ms) |
| `retryPolicy` | `RetryPolicyConfig` | — | Custom retry policy |