0.2.10

DESIGN.md

@@ -748,7 +748,7 @@ Kun is structured as **ports & adapters**:
   `ApprovalGate`, `EventBus`, `WorkspaceInspector`, `Clock`,
   `IdGenerator`. These are intentionally small.
 - `adapters/` — concrete implementations. The default
-  `DeepseekCompatModelClient` speaks the
+  `CompatModelClient` speaks the
   `POST {baseUrl}/v1/chat/completions` shape; the default
   `LocalToolHost` runs tools in-process with approval gating.
 - `services/` — orchestration. `ThreadService` and `TurnService`
@@ -1348,7 +1348,7 @@ If any check fails, the change is not ready.
 | Immutable prefix | `kun/src/cache/immutable-prefix.ts` |
 | HTTP routes | `kun/src/server/routes/` |
 | Tool host | `kun/src/adapters/tool/local-tool-host.ts` |
-| Model client | `kun/src/adapters/model/deepseek-compat-model-client.ts` |
+| Model client | `kun/src/adapters/model/compat-model-client.ts` |
 | Cache doc | `docs/kun-cache-optimization.md` |
 | Architecture doc | `docs/kun-architecture.md` |
 | Contribution doc | `docs/kun-contributing.md` |

DESIGN.zh-CN.md

@@ -748,7 +748,7 @@ Kun is structured as **ports & adapters**:
   `ApprovalGate`, `EventBus`, `WorkspaceInspector`, `Clock`,
   `IdGenerator`. These are intentionally small.
 - `adapters/` — concrete implementations. The default
-  `DeepseekCompatModelClient` speaks the
+  `CompatModelClient` speaks the
   `POST {baseUrl}/v1/chat/completions` shape; the default
   `LocalToolHost` runs tools in-process with approval gating.
 - `services/` — orchestration. `ThreadService` and `TurnService`
@@ -1348,7 +1348,7 @@ If any check fails, the change is not ready.
 | Immutable prefix | `kun/src/cache/immutable-prefix.ts` |
 | HTTP routes | `kun/src/server/routes/` |
 | Tool host | `kun/src/adapters/tool/local-tool-host.ts` |
-| Model client | `kun/src/adapters/model/deepseek-compat-model-client.ts` |
+| Model client | `kun/src/adapters/model/compat-model-client.ts` |
 | Cache doc | `docs/kun-cache-optimization.md` |
 | Architecture doc | `docs/kun-architecture.md` |
 | Contribution doc | `docs/kun-contributing.md` |

docs/kun-cache-optimization.en.md

@@ -110,7 +110,7 @@ The toolset itself is part of the prompt prefix. Kun will do the following befor
 
 Implementation location:
 
-- `kun/src/adapters/model/deepseek-compat-model-client.ts`
+- `kun/src/adapters/model/compat-model-client.ts`
 - `kun/src/cache/tool-catalog-fingerprint.ts`
 - `kun/src/loop/agent-loop.ts`
 
@@ -137,7 +137,7 @@ Kun currently performs a shared model history repair on messages at the model re
 Implementation location:
 
 - `kun/src/domain/model-history-repair.ts`
-- `kun/src/adapters/model/deepseek-compat-model-client.ts`
+- `kun/src/adapters/model/compat-model-client.ts`
 - `kun/src/loop/agent-loop.ts`
 
 Kun will also do a shared layer of history hygiene at model request boundaries:
@@ -204,7 +204,7 @@ Only when the native field is missing will it fall back to the compatible field:
 
 Implementation location:
 
-- `kun/src/adapters/model/deepseek-compat-model-client.ts`
+- `kun/src/adapters/model/compat-model-client.ts`
 
 The hit rate formula uses:
 

docs/kun-cache-optimization.md

@@ -102,7 +102,7 @@ Kun 通过 `ImmutablePrefix` 管理系统 prompt、tools、pinned constraints
 
 实现位置：
 
-- `kun/src/adapters/model/deepseek-compat-model-client.ts`
+- `kun/src/adapters/model/compat-model-client.ts`
 - `kun/src/cache/tool-catalog-fingerprint.ts`
 - `kun/src/loop/agent-loop.ts`
 
@@ -130,7 +130,7 @@ Kun 当前在模型请求边界对消息做一层共享的 model history repair
 实现位置：
 
 - `kun/src/domain/model-history-repair.ts`
-- `kun/src/adapters/model/deepseek-compat-model-client.ts`
+- `kun/src/adapters/model/compat-model-client.ts`
 - `kun/src/loop/agent-loop.ts`
 
 Kun 也会在模型请求边界做一层共享的 history hygiene：
@@ -198,7 +198,7 @@ Kun 的缓存命中统计优先使用 provider 原生 usage 字段：
 
 实现位置：
 
-- `kun/src/adapters/model/deepseek-compat-model-client.ts`
+- `kun/src/adapters/model/compat-model-client.ts`
 
 命中率公式采用：
 

docs/kun-contributing.en.md

@@ -62,7 +62,7 @@ Kun's directory itself is a hexagonal physical layout:
             ▼                         ▼
    ┌──────────────────────────────────────────────────────────────────────────┐
    │ adapters/  -- concrete implementations of ports                         │
-   │   model/deepseek-compat-model-client.ts                                │
+   │   model/compat-model-client.ts                                │
    │   tool/local-tool-host.ts                                              │
    │   in-memory-event-bus / in-memory-approval-gate / in-memory-thread-…  │
    │   file/file-thread-store / file-session-store                         │
@@ -137,7 +137,7 @@ export interface ModelClient {
 
 ```
 
-Implementation (`kun/src/adapters/model/deepseek-compat-model-client.ts`)
+Implementation (`kun/src/adapters/model/compat-model-client.ts`)
 Parse HTTP+SSE into a sequence of `ModelStreamChunk`.
 
 Tests (`kun/tests/ports.test.ts`) are injected directly with `makeFakeModel`
@@ -502,7 +502,7 @@ Does not involve `loop/`, `server/`, `services/`. Changes are isolated to a smal
 
 1. **ports**:`ModelClient` is already an abstract interface, **no need to change**.
 2. **adapters**:`kun/src/adapters/model/openai-compat-model-client.ts`
-   Create a new one (refer to the SSE parsing logic of `deepseek-compat-model-client.ts`).
+   Create a new one (refer to the SSE parsing logic of `compat-model-client.ts`).
 3. **adapters barrel**:`kun/src/adapters/index.ts` exports new classes.
 4. **settings**: in `kun/src/contracts/...` or the caller
    `ServeOptionsSchema` adds `modelProvider` field (if you need to switch).

docs/kun-contributing.md

@@ -62,7 +62,7 @@ Kun 的目录本身就是六边形的物理布局:
             ▼                         ▼
    ┌──────────────────────────────────────────────────────────────────────────┐
    │ adapters/  ──  ports 的具体实现                                        │
-   │   model/deepseek-compat-model-client.ts                                │
+   │   model/compat-model-client.ts                                │
    │   tool/local-tool-host.ts                                              │
    │   in-memory-event-bus / in-memory-approval-gate / in-memory-thread-…  │
    │   file/file-thread-store / file-session-store                         │
@@ -134,7 +134,7 @@ export interface ModelClient {
 }
 ```
 
-实现 (`kun/src/adapters/model/deepseek-compat-model-client.ts`)
+实现 (`kun/src/adapters/model/compat-model-client.ts`)
 把 HTTP+SSE 解析为 `ModelStreamChunk` 序列。
 
 测试 (`kun/tests/ports.test.ts`) 直接 `makeFakeModel` 注入
@@ -489,7 +489,7 @@ cache 路由"。
 
 1. **ports**:`ModelClient` 已经是抽象接口,**不需要改**。
 2. **adapters**:`kun/src/adapters/model/openai-compat-model-client.ts`
-   新建(参考 `deepseek-compat-model-client.ts` 的 SSE 解析逻辑)。
+   新建(参考 `compat-model-client.ts` 的 SSE 解析逻辑)。
 3. **adapters barrel**:`kun/src/adapters/index.ts` 导出新类。
 4. **settings**:在 `kun/src/contracts/...` 或调用方
    `ServeOptionsSchema` 加 `modelProvider` 字段(若需要切换)。

docs/model-provider-presets.md

@@ -64,6 +64,60 @@ MiniMax:
 - image base URL: `https://api.minimaxi.com`
 - image models: `image-01`
 
+Zhipu Coding Plan:
+
+- id: `zhipu-coding-plan`
+- base URL: `https://open.bigmodel.cn/api/coding/paas/v4`
+- endpoint format: OpenAI Chat Completions
+- initial models: `glm-5.2`, `glm-5.1`, `glm-5-turbo`, `glm-4.7`,
+  `glm-4.5-air`
+- role: coding subscription provider added from Settings > Providers only
+
+Z.ai Coding Plan:
+
+- id: `zai-coding-plan`
+- base URL: `https://api.z.ai/api/coding/paas/v4`
+- endpoint format: OpenAI Chat Completions
+- initial models: `glm-5.1`, `glm-5`, `glm-5-turbo`, `glm-4.7`,
+  `glm-4.5-air`
+- role: international coding subscription provider added from Settings >
+  Providers only
+
+Kimi Code:
+
+- id: `kimi-code`
+- base URL: `https://api.kimi.com/coding/v1`
+- endpoint format: OpenAI Chat Completions
+- initial model: `kimi-for-coding`
+- role: Kimi coding subscription provider added from Settings > Providers only
+
+Moonshot CN:
+
+- id: `moonshot-cn`
+- base URL: `https://api.moonshot.cn/v1`
+- endpoint format: OpenAI Chat Completions
+- initial models: `kimi-k2.7-code`, `kimi-k2.6`, `kimi-k2.5`,
+  `moonshot-v1-128k`, `moonshot-v1-32k`, `moonshot-v1-8k`
+- model profile note: Kimi K2 models are marked as text+image chat models;
+  video input is not represented in the current provider schema
+- role: Moonshot open-platform provider added from Settings > Providers only
+
+Moonshot Global:
+
+- id: `moonshot-global`
+- base URL: `https://api.moonshot.ai/v1`
+- endpoint format: OpenAI Chat Completions
+- initial models: `kimi-k2.7-code`, `kimi-k2.6`, `kimi-k2.5`,
+  `moonshot-v1-128k`, `moonshot-v1-32k`, `moonshot-v1-8k`
+- model profile note: Kimi K2 models are marked as text+image chat models;
+  video input is not represented in the current provider schema
+- role: international Moonshot open-platform provider added from Settings >
+  Providers only
+
 The defaults are not locked. Users can edit base URLs, protocols, and model IDs
 if provider endpoints change, and they can add custom compatible providers at
 any time.
+
+First-run setup intentionally remains focused on the default stack. It only
+shows DeepSeek plus the Xiaomi and MiniMax presets; LiteLLM, Zhipu, Z.ai, Kimi
+Code, and Moonshot presets are opt-in from Settings > Providers.

electron-builder.config.cjs

@@ -156,7 +156,9 @@ module.exports = {
     sign: hasExplicitMacSigningIdentity
   },
   win: {
-    icon: './src/asset/img/kun.png',
+    // Windows does not mask app icons for us; use the rounded asset so
+    // desktop/start-menu/taskbar shortcuts do not show a hard square edge.
+    icon: './src/asset/img/kun_mac.png',
     target: [{ target: 'nsis', arch: ['x64'] }]
   },
   nsis: {

kun/src/adapters/index.ts

@@ -1,4 +1,4 @@
-export * from './model/deepseek-compat-model-client.js'
+export * from './model/compat-model-client.js'
 export * from './tool/index.js'
 export * from './in-memory-event-bus.js'
 export * from './in-memory-approval-gate.js'