fix(config): recognize provider model capabilities

.changeset/deepseek-anthropic-capabilities.md

@@ -0,0 +1,5 @@
+---
+"@moonshot-ai/kimi-code": patch
+---
+
+Recognize DeepSeek Anthropic-compatible chat model capabilities.

.changeset/env-model-provider-capabilities.md

@@ -0,0 +1,5 @@
+---
+"@moonshot-ai/kimi-code": patch
+---
+
+Avoid defaulting OpenAI and Anthropic environment models to Kimi-only capabilities.

.changeset/kimi-model-capabilities.md

@@ -0,0 +1,5 @@
+---
+"@moonshot-ai/kimi-code": patch
+---
+
+Recognize current Kimi Open Platform model capabilities.

.changeset/mimo-anthropic-capabilities.md

@@ -0,0 +1,5 @@
+---
+"@moonshot-ai/kimi-code": patch
+---
+
+Recognize MiMo Anthropic-compatible chat model capabilities.

.changeset/minimax-anthropic-capabilities.md

@@ -0,0 +1,5 @@
+---
+"@moonshot-ai/kimi-code": patch
+---
+
+Recognize MiniMax Anthropic-compatible chat model capabilities.

docs/en/configuration/env-vars.md

@@ -105,7 +105,7 @@ Complete variable list:
 | `KIMI_MODEL_PROVIDER_TYPE` | No | Provider type: `kimi`, `anthropic`, `openai` | `kimi` |
 | `KIMI_MODEL_BASE_URL` | No | API base URL | Each type has its own default |
 | `KIMI_MODEL_MAX_CONTEXT_SIZE` | No | Maximum context length (tokens) | `262144` (256 K) |
-| `KIMI_MODEL_CAPABILITIES` | No | Comma-separated capability tags, unioned with auto-detected capabilities | `image_in,thinking` |
+| `KIMI_MODEL_CAPABILITIES` | No | Comma-separated capability tags, unioned with auto-detected capabilities | `kimi` → `image_in,thinking`; `openai` / `anthropic` → unset |
 | `KIMI_MODEL_DISPLAY_NAME` | No | Name shown in `/model` | Falls back to `KIMI_MODEL_NAME` |
 | `KIMI_MODEL_MAX_OUTPUT_SIZE` | No | Per-request output cap (`anthropic` only) | Model default |
 | `KIMI_MODEL_REASONING_KEY` | No | Reasoning field name override (`openai` only) | Auto-detected |

docs/zh/configuration/env-vars.md

@@ -105,7 +105,7 @@ kimi
 | `KIMI_MODEL_PROVIDER_TYPE` | 否 | 供应商类型：`kimi`、`anthropic`、`openai` | `kimi` |
 | `KIMI_MODEL_BASE_URL` | 否 | API 基础 URL | 各类型有各自默认值 |
 | `KIMI_MODEL_MAX_CONTEXT_SIZE` | 否 | 最大上下文长度（token 数） | `262144`（256K） |
-| `KIMI_MODEL_CAPABILITIES` | 否 | 逗号分隔的能力标签，与自动探测的能力取并集 | `image_in,thinking` |
+| `KIMI_MODEL_CAPABILITIES` | 否 | 逗号分隔的能力标签，与自动探测的能力取并集 | `kimi` → `image_in,thinking`；`openai` / `anthropic` → 未设置 |
 | `KIMI_MODEL_DISPLAY_NAME` | 否 | 在 `/model` 中显示的名称 | 回退到 `KIMI_MODEL_NAME` |
 | `KIMI_MODEL_MAX_OUTPUT_SIZE` | 否 | 单次输出上限（仅 `anthropic`） | 模型默认值 |
 | `KIMI_MODEL_REASONING_KEY` | 否 | 推理字段名覆盖（仅 `openai`） | 自动探测 |

packages/agent-core/src/config/env-model.ts

@@ -24,8 +24,10 @@ const DEFAULT_BASE_URL: Partial<Record<ProviderType, string>> = {
 /** Default context window (256K) used when KIMI_MODEL_MAX_CONTEXT_SIZE is unset. */
 const DEFAULT_MAX_CONTEXT_SIZE = 262144;
 
-/** Default capabilities when KIMI_MODEL_CAPABILITIES is unset (kimi models support both). */
-const DEFAULT_CAPABILITIES = ['image_in', 'thinking'];
+/** Default capabilities when KIMI_MODEL_CAPABILITIES is unset. */
+const DEFAULT_CAPABILITIES_BY_TYPE: Partial<Record<ProviderType, readonly string[]>> = {
+  kimi: ['image_in', 'thinking'],
+};
 
 type Env = Readonly<Record<string, string | undefined>>;
 
@@ -119,7 +121,9 @@ export function applyEnvModelConfig(config: KimiConfig, env: Env = process.env):
     maxOutputRaw !== undefined
       ? parsePositiveInt(maxOutputRaw, 'KIMI_MODEL_MAX_OUTPUT_SIZE')
       : undefined;
-  const capabilities = parseCapabilities(env['KIMI_MODEL_CAPABILITIES']) ?? DEFAULT_CAPABILITIES;
+  const capabilities =
+    parseCapabilities(env['KIMI_MODEL_CAPABILITIES']) ??
+    DEFAULT_CAPABILITIES_BY_TYPE[type]?.slice();
   const displayName = trimmed(env['KIMI_MODEL_DISPLAY_NAME']);
   const reasoningKey = trimmed(env['KIMI_MODEL_REASONING_KEY']);
   const adaptiveThinking = parseBooleanVar(

packages/agent-core/test/config/env-model.test.ts

@@ -88,6 +88,17 @@ describe('applyEnvModelConfig', () => {
     expect(anthropic?.baseUrl).toBeUndefined();
   });
 
+  it('does not apply Kimi default capabilities to non-Kimi providers', () => {
+    expect(
+      apply({ ...MIN, KIMI_MODEL_PROVIDER_TYPE: 'openai' })
+        .models?.[ENV_MODEL_ALIAS_KEY]?.capabilities,
+    ).toBeUndefined();
+    expect(
+      apply({ ...MIN, KIMI_MODEL_PROVIDER_TYPE: 'anthropic' })
+        .models?.[ENV_MODEL_ALIAS_KEY]?.capabilities,
+    ).toBeUndefined();
+  });
+
   it('rejects unsupported provider types', () => {
     expectConfigInvalid(() =>
       apply({ ...MIN, KIMI_MODEL_PROVIDER_TYPE: 'google-genai' }),

packages/kosong/src/providers/capability-registry.ts

@@ -45,6 +45,28 @@ const GEMINI_CATALOGUED_PREFIXES = [
   'gemini-2.5-flash',
 ] as const;
 
+const MIMO_THINKING_TOOL_MODELS = [
+  'mimo-v2.5-pro',
+  'mimo-v2-pro',
+  'mimo-v2-flash',
+] as const;
+
+const MIMO_THINKING_VISION_TOOL_MODELS = ['mimo-v2.5', 'mimo-v2-omni'] as const;
+
+const MINIMAX_THINKING_VISION_TOOL_MODELS = ['minimax-m3'] as const;
+
+const MINIMAX_TEXT_TOOL_MODELS = [
+  'minimax-m2.7',
+  'minimax-m2.7-highspeed',
+  'minimax-m2.5',
+  'minimax-m2.5-highspeed',
+  'minimax-m2.1',
+  'minimax-m2.1-highspeed',
+  'minimax-m2',
+] as const;
+
+const DEEPSEEK_THINKING_TOOL_MODELS = ['deepseek-v4-pro', 'deepseek-v4-flash'] as const;
+
 const OPENAI_REASONING_CAPABILITY: ModelCapability = Object.freeze({
   image_in: false,
   video_in: false,
@@ -90,6 +112,51 @@ const ANTHROPIC_THINKING_VISION_TOOL_CAPABILITY: ModelCapability = Object.freeze
   max_context_tokens: 0,
 });
 
+const MIMO_THINKING_TOOL_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: false,
+  video_in: false,
+  audio_in: false,
+  thinking: true,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
+const MIMO_THINKING_VISION_TOOL_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: true,
+  video_in: false,
+  audio_in: false,
+  thinking: true,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
+const MINIMAX_THINKING_VISION_TOOL_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: true,
+  video_in: false,
+  audio_in: false,
+  thinking: true,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
+const MINIMAX_TEXT_TOOL_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: false,
+  video_in: false,
+  audio_in: false,
+  thinking: false,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
+const DEEPSEEK_THINKING_TOOL_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: false,
+  video_in: false,
+  audio_in: false,
+  thinking: true,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
 const GEMINI_MULTIMODAL_TOOL_CAPABILITY: ModelCapability = Object.freeze({
   image_in: true,
   video_in: true,
@@ -135,6 +202,26 @@ const OPENAI_RESPONSES_CAPABILITY_CATALOG: readonly CapabilityCatalogEntry[] = [
 ];
 
 const ANTHROPIC_CAPABILITY_CATALOG: readonly CapabilityCatalogEntry[] = [
+  {
+    matches: (name) => hasExactModel(name, MIMO_THINKING_TOOL_MODELS),
+    capability: MIMO_THINKING_TOOL_CAPABILITY,
+  },
+  {
+    matches: (name) => hasExactModel(name, MIMO_THINKING_VISION_TOOL_MODELS),
+    capability: MIMO_THINKING_VISION_TOOL_CAPABILITY,
+  },
+  {
+    matches: (name) => hasExactModel(name, MINIMAX_THINKING_VISION_TOOL_MODELS),
+    capability: MINIMAX_THINKING_VISION_TOOL_CAPABILITY,
+  },
+  {
+    matches: (name) => hasExactModel(name, MINIMAX_TEXT_TOOL_MODELS),
+    capability: MINIMAX_TEXT_TOOL_CAPABILITY,
+  },
+  {
+    matches: (name) => hasExactModel(name, DEEPSEEK_THINKING_TOOL_MODELS),
+    capability: DEEPSEEK_THINKING_TOOL_CAPABILITY,
+  },
   {
     matches: (name) => hasPrefix(name, CLAUDE_3_PREFIXES),
     capability: ANTHROPIC_VISION_TOOL_CAPABILITY,
@@ -153,6 +240,10 @@ function hasPrefix(modelName: string, prefixes: readonly string[]): boolean {
   return prefixes.some((prefix) => modelName.startsWith(prefix));
 }
 
+function hasExactModel(modelName: string, models: readonly string[]): boolean {
+  return models.some((model) => modelName === model);
+}
+
 function isOpenAIReasoningModel(modelName: string): boolean {
   return /^o\d/.test(modelName);
 }

packages/kosong/src/providers/kimi.ts

@@ -40,6 +40,43 @@ import {
   sanitizeToolCallId,
   type ToolCallIdPolicy,
 } from './tool-call-id';
+
+const KIMI_K2_6_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: true,
+  video_in: true,
+  audio_in: false,
+  thinking: true,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
+const KIMI_K2_5_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: true,
+  video_in: false,
+  audio_in: false,
+  thinking: true,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
+const KIMI_VISION_TOOL_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: true,
+  video_in: false,
+  audio_in: false,
+  thinking: false,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
+const KIMI_TEXT_TOOL_CAPABILITY: ModelCapability = Object.freeze({
+  image_in: false,
+  video_in: false,
+  audio_in: false,
+  thinking: false,
+  tool_use: true,
+  max_context_tokens: 0,
+});
+
 export interface KimiOptions {
   apiKey?: string | undefined;
   baseUrl?: string | undefined;
@@ -499,6 +536,15 @@ export class KimiChatProvider implements ChatProvider {
   }
 
   getCapability(_model?: string): ModelCapability {
+    const model = (_model ?? this._model).toLowerCase();
+    if (model === 'kimi-k2.6') return KIMI_K2_6_CAPABILITY;
+    if (model === 'kimi-k2.5') return KIMI_K2_5_CAPABILITY;
+    if (/^moonshot-v1-(?:8k|32k|128k)-vision-preview$/.test(model)) {
+      return KIMI_VISION_TOOL_CAPABILITY;
+    }
+    if (/^moonshot-v1-(?:8k|32k|128k)$/.test(model)) {
+      return KIMI_TEXT_TOOL_CAPABILITY;
+    }
     return UNKNOWN_CAPABILITY;
   }