@reachforstar/dsh-llm-amax

AMAX Token Router model provider for DeepSeek Harness: an OpenAI-compatible gateway route with /v1/models discovery 适合需要通过 AMAX 路由网关接入并调用各类大语言模型的 DSH 用户。

パッケージ

@reachforstar/dsh-llm-amax

互換性

未検証

Harness ピア範囲

^0.1.1-rc.2

Cordis ピア範囲

バージョン

0.1.1-rc.2

ライセンス

MIT

最終更新

2026/08/22

Configuration

Set the credential first: export AMAX_API_KEY (the default credential reference), or store any other reference and point the route at it. Then, in the web Models page, select AMAX Token Router and use fetch available models — the draft key comes from AMAX_API_KEY before the route is even saved.

By hand, the user-settings section is:

llm-amax:
  providers:
    amax:
      # apiKeyEnv: AMAX_API_KEY      # default; any credential ref works
      # baseURL: https://ai.amaxsmp.com/v1
      models:
        - id: deepseek-v4-flash
          contextWindow: 262144

Key	Default	Meaning
`apiKeyEnv`	`AMAX_API_KEY`	Credential reference resolved per request through the credentials service.
`baseURL`	`https://ai.amaxsmp.com/v1`	Endpoint override for the OpenAI-compatible API.
`models`	`[]`	The route's model catalog; each entry takes `id`, `name`, `contextWindow`, `maxTokens`, `input`, `reasoningEfforts`.
`reasoning` / `transport` / `timeoutMs` / `headers` / `retryPolicy`	unset	Provider-neutral knobs passed to the pi-ai adapter, mirroring `dsh-llm-pi-ai`.

@reachforstar/dsh-llm-amax

インストール

ドキュメント

Configuration