deps(ai): migrate the Vercel AI SDK from 5 to 6

package.json

@@ -116,11 +116,11 @@
     "rather than in app.asar."
   ],
   "devDependencies": {
-    "@ai-sdk/anthropic": "^2.0.105",
-    "@ai-sdk/google": "^2.0.99",
-    "@ai-sdk/openai": "^2.0.89",
-    "@ai-sdk/openai-compatible": "^1.0.31",
-    "@ai-sdk/provider": "^2.0.5",
+    "@ai-sdk/anthropic": "^3.0.125",
+    "@ai-sdk/google": "^3.0.129",
+    "@ai-sdk/openai": "^3.0.121",
+    "@ai-sdk/openai-compatible": "^2.0.80",
+    "@ai-sdk/provider": "^3.0.18",
     "@dagrejs/dagre": "^1.1.8",
     "@dnd-kit/core": "^6.3.1",
     "@dnd-kit/sortable": "^10.0.0",
@@ -162,7 +162,7 @@
     "@types/turndown": "^5.0.6",
     "@vitejs/plugin-react": "^5.1.2",
     "@xyflow/react": "^12.10.0",
-    "ai": "^5.0.123",
+    "ai": "^6.0.297",
     "autoprefixer": "^10.4.23",
     "cheerio": "^1.2.0",
     "class-variance-authority": "^0.7.1",

src/main/models/ModelClient.ts

@@ -44,9 +44,9 @@ const toTokenUsage = (usage: LanguageModelUsage): ChatTokenUsage => ({
 })
 
 /**
- * 将 APIMessage 转换为 AI SDK 的 CoreMessage 格式
+ * 将 APIMessage 转换为 AI SDK 的 ModelMessage 格式
  */
-function convertToCoreMessages(messages: APIMessage[]) {
+function toModelMessages(messages: APIMessage[]) {
   return messages.map((msg) => ({
     role: msg.role,
     content: msg.content
@@ -90,6 +90,19 @@ export class ModelClient {
     return this.adapter.chatProviderOptions?.(this.connection)
   }
 
+  /**
+   * providerOptions for structured-output calls (streamObject).
+   *
+   * AI SDK 6 flipped the OpenAI-family `strictJsonSchema` default to true, and the
+   * schemas these services generate use `.optional()` fields, which strict mode
+   * rejects because every property has to be listed in `required`. This returns the
+   * protocol's pre-v6 setting so the request body is unchanged; protocols whose
+   * structured output is unaffected return undefined.
+   */
+  structuredOutputProviderOptions(): SharedV2ProviderOptions | undefined {
+    return this.adapter.structuredOutputProviderOptions?.(this.connection)
+  }
+
   /**
    * 流式发送消息
    *
@@ -106,7 +119,7 @@ export class ModelClient {
 
     const result = streamText({
       model: this.getAIModel(),
-      messages: convertToCoreMessages(messages),
+      messages: toModelMessages(messages),
       temperature: DEFAULT_TEMPERATURE,
       // Only when the connection asks for a ceiling. Absent means the model's own
       // default, which is what a reasoning model needs: its thinking shares this

src/main/models/protocols/googleGenerativeAi.ts

@@ -27,7 +27,7 @@ export const googleGenerativeAiAdapter: ProtocolAdapter = {
   },
 
   createEmbeddingModel(connection): EmbeddingModel {
-    return createProvider(connection).textEmbedding(connection.modelId)
+    return createProvider(connection).embedding(connection.modelId)
   },
 
   embeddingProviderOptions(_connection, dimensions) {

src/main/models/protocols/openaiCompletions.ts

@@ -29,7 +29,7 @@ export const openaiCompletionsAdapter: ProtocolAdapter = {
   },
 
   createEmbeddingModel(connection): EmbeddingModel {
-    return createProvider(connection).textEmbeddingModel(connection.modelId)
+    return createProvider(connection).embeddingModel(connection.modelId)
   },
 
   embeddingProviderOptions(_connection, dimensions) {

src/main/models/protocols/openaiResponses.ts

@@ -27,7 +27,7 @@ export const openaiResponsesAdapter: ProtocolAdapter = {
   },
 
   createEmbeddingModel(connection): EmbeddingModel {
-    return createProvider(connection).textEmbeddingModel(connection.modelId)
+    return createProvider(connection).embeddingModel(connection.modelId)
   },
 
   embeddingProviderOptions(_connection, dimensions) {
@@ -41,6 +41,14 @@ export const openaiResponsesAdapter: ProtocolAdapter = {
       : undefined
   },
 
+  /**
+   * AI SDK 6 把 `strictJsonSchema` 的默认值改为 true，而结构化输出服务用的 schema
+   * 带 `.optional()` 字段，strict 模式会拒绝。显式关掉，保持与 v5 相同的请求体。
+   */
+  structuredOutputProviderOptions() {
+    return { [PROVIDER_OPTIONS_KEY]: { strictJsonSchema: false } }
+  },
+
   supportsModelListing: true,
 
   listModels: listOpenAICompatibleModels

src/main/models/protocols/types.ts

@@ -40,6 +40,16 @@ export interface ProtocolAdapter {
    */
   chatProviderOptions?(connection: ModelConnection): SharedV2ProviderOptions | undefined
 
+  /**
+   * 把结构化输出（streamObject）的 providerOptions 翻译成该协议的设置。
+   *
+   * AI SDK 6 把 OpenAI 系的 `strictJsonSchema` 默认值从 false 改成 true。本项目生成
+   * 的结构化输出 schema 大量使用 `.optional()`，而 strict 模式要求每个属性都出现在
+   * `required` 里，服务端会直接拒绝。显式回传 `strictJsonSchema: false` 是为了让线上
+   * 请求体与 v5 完全一致。结构化输出不受该默认值影响的协议返回 undefined。
+   */
+  structuredOutputProviderOptions?(connection: ModelConnection): SharedV2ProviderOptions | undefined
+
   /**
    * 该协议是否支持从端点拉取模型列表（/v1/models 等）。
    */

src/main/services/AnkiCardService.ts

@@ -214,7 +214,8 @@ export class AnkiCardService {
       const { partialObjectStream, object } = streamObject({
         model: model,
         schema: AnkiSchema,
-        prompt: prompt
+        prompt: prompt,
+        providerOptions: client.structuredOutputProviderOptions()
       })
 
       // 监听流式更新 (用于显示进度)