@mars-sea/dsh-commandcode-provider 0.10.0-alpha.1 → 0.10.0-alpha.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -4,7 +4,34 @@ All notable changes to this project will be documented in this file.
4
4
 
5
5
  The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
6
6
 
7
- ## [Unreleased]
7
+ ## [0.10.0-alpha.3] - 2026-09-01
8
+
9
+ > **Alpha channel for dsh 0.1.2-alpha.2.** This release requires **dsh 0.1.2-alpha.2 or later** and is published under the `alpha` npm tag, so `@latest` stays on 0.9.1 for older Harness releases. Install it explicitly with `@alpha`.
10
+
11
+ ### Added
12
+
13
+ - **New model: DeepSeek V4 Flash Fast (`deepseek/deepseek-v4-flash-fast`, command-code@1.39.0).** A Go-tier 1M-context low-latency DeepSeek V4 Flash deployment: it appears in the picker under the Go tier with `1M` markers, offers selectable reasoning efforts (`low`/`high`/`max`), and carries time-of-day (peak/off-peak) pricing like the rest of the DeepSeek V4 family. It is text-input only, so it stays outside the Vision whitelist.
14
+
15
+ ### Changed
16
+
17
+ - **Synced with the official command-code@1.39.2 CLI** (2026-09-01: 1.39.0 added DeepSeek V4 Flash Fast; 1.39.1 dropped `medium` effort for it; 1.39.2 retired the MiniMax free models). `COMMAND_CODE_CLI_VERSION` is now `1.39.2`.
18
+ - **The MiniMax M3 / M2.7 FREE promo variants were retired** from the official CLI in command-code@1.39.2 ("Retire MiniMax free models"): the browser hides them from the picker and the pricing page no longer lists them as free. The plugin removes their `FREE`/`free: true` entries rather than letting them lapse on their old 2026-09-05 expiry, so users no longer see a retired discount. The paid MiniMax M3 / M2.7 rows keep their own rates.
19
+
20
+ ## [0.10.0-alpha.2] - 2026-08-31
21
+
22
+ > **Alpha channel for dsh 0.1.2-alpha.2.** This release requires **dsh 0.1.2-alpha.2 or later** and is published under the `alpha` npm tag, so `@latest` stays on 0.9.1 for older Harness releases. Install it explicitly with `@alpha`.
23
+
24
+ ### Added
25
+
26
+ - **Route models to accounts (`modelAccountRules`).** A new settings card (and matching `Config.modelAccountRules` YAML) routes catalog models — picked via **multi-select from the live model catalog** (fetched Host-side through a new `commandcode/models` Remote) — to an account slot. Requests whose model is in a rule serve from that account while it is usable, falling back to the normal rotation when the routed account is exhausted or invalid. First matching rule wins. Pairs naturally with the existing multi-account rotation — e.g. keep cheap Go-tier models on the Go account and route GOAT-only models to the GOAT account.
27
+
28
+ ### Changed
29
+
30
+ - **The Models page now configures Command Code entirely through the official 编辑 button.** The plugin card no longer adds its own header, badges, and edit toggle under the row: with the editor closed the row looks like any other provider, and opening 编辑 replaces the old "其余字段在 settings.yaml 中" placeholder (which offered nothing but a disabled apply button) with the real controls — key/route status, the API-key field, official sign-in, and discard/save.
31
+ - **Synced with the official command-code@1.38.2 CLI** (2026-08-28: 1.38.0 added "Let custom agents pin a reasoning effort next to their model" + "Add reasoning effort support for Tencent Hy4 Preview"; 1.38.1/1.38.2 are same-day hotfixes). `COMMAND_CODE_CLI_VERSION` is now `1.38.2`.
32
+ - **Tencent Hy4 Preview now offers selectable reasoning efforts (`low`/`medium`/`high`).** The model gained official effort levels in command-code@1.38.0, so the picker can now offer a reasoning-effort selector for it; it is no longer listed as an auto-thinking model.
33
+ - **Gemini 3.7 Flash's 50% off deal was retired** from the official pricing page's `#deals` section; the snapshot no longer advertises it, so the picker shows the model at full price.
34
+ - **Closed two catalog-coverage gaps**: the hidden free variant `tencent/Hy3` and the deprecated `inclusionai/ling-3.0-flash-free` (its free promo ended 2026-08-03) are now in the Go tier snapshot. Plan tiers read 42/46/59/64.
8
35
 
9
36
  ## [0.10.0-alpha.1] - 2026-09-01
10
37
 
package/README.md CHANGED
@@ -93,6 +93,7 @@ With several Command Code subscriptions, the plugin **switches to the next accou
93
93
 
94
94
  - **Setup** — use the **Account rotation** card at Settings → **Command Code** to add accounts with a label and API key; the top-level key always serves first as the `default` account.
95
95
  - **Manual switching** — the **Active account** dropdown pins a preferred account; if it is exhausted, requests fall back to other accounts and return once its window resets.
96
+ - **Route models to accounts** — the **Route models to accounts** card picks catalog models (multi-select, fetched from the live catalog) and routes them to an account. A request whose model is in a rule serves from that account while it is usable; an exhausted or invalid routed account falls back to the normal rotation. Rules match in list order — the first hit wins.
96
97
  - **Status** — the **Account usage** card and `/commandcode` report per-account state.
97
98
 
98
99
  The equivalent YAML (`$DSH_HOME/settings.yaml` or composition config):
@@ -106,6 +107,14 @@ llm-commandcode:
106
107
  apiKeyEnv: COMMANDCODE_API_KEY_2
107
108
  - label: Go #3
108
109
  apiKeyEnv: COMMANDCODE_API_KEY_3
110
+ modelAccountRules: # optional: route models to accounts (first match wins)
111
+ - models: # catalog model ids (multi-select)
112
+ - deepseek/deepseek-v4-pro
113
+ - deepseek/deepseek-v4-flash-vision-exp
114
+ account: COMMANDCODE_API_KEY_2
115
+ - models:
116
+ - tencent/hy4-preview
117
+ account: default
109
118
  ```
110
119
 
111
120
  ## Configure
package/README.zh-CN.md CHANGED
@@ -91,6 +91,7 @@ cmd login # macOS/Linux;Windows 原生版:cmdc login
91
91
 
92
92
  - **配置**:在 **设置 → Command Code** 的「多账户轮换」卡片添加账户并填写备注名与 API key;顶层 key 始终是第一顺位的 `default` 账户。
93
93
  - **手动切换**:卡片上的「当前使用账户」下拉框可指定优先账户;所选账户耗尽时自动回落到其他账户,窗口重置后自动恢复。
94
+ - **按模型切换账户**:在「按模型切换账户」卡片从实时模型目录**多选**模型并固定到某个账户。请求的模型在规则列表中且该账户可用时使用该账户;账户耗尽或密钥失效时自动回落到常规轮换。规则按列表顺序匹配,第一条命中生效。
94
95
  - **状态展示**:「账户用量」卡片与 `/commandcode` 均按账户分别显示状态。
95
96
 
96
97
  等价的 YAML(`$DSH_HOME/settings.yaml` 或组合配置):
@@ -104,6 +105,14 @@ llm-commandcode:
104
105
  apiKeyEnv: COMMANDCODE_API_KEY_2
105
106
  - label: Go #3
106
107
  apiKeyEnv: COMMANDCODE_API_KEY_3
108
+ modelAccountRules: # 可选:按模型路由到账户(第一条命中生效)
109
+ - models: # 目录模型 id(可多选)
110
+ - deepseek/deepseek-v4-pro
111
+ - deepseek/deepseek-v4-flash-vision-exp
112
+ account: COMMANDCODE_API_KEY_2
113
+ - models:
114
+ - tencent/hy4-preview
115
+ account: default
107
116
  ```
108
117
 
109
118
  ## 配置