Give visual and reasoning models their own first-chunk timeouts

Vision and reasoning requests reused the chat model's 10s first-chunk
timeout, but both legitimately need longer before the first chunk:
vision must download the image server-side first, and reasoning has a
thinking warmup. Logs showed frequent TimeoutCancellationException at
10s/15s for imageRecognition and reasoning.

Add separate visualFirstChunkTimeout (120s) and reasoningFirstChunkTimeout
(90s) config, and raise each service's socket timeout to at least its
first-chunk budget so the socket layer doesn't sever the connection
before the first-chunk timeout can apply. Chat endpoints are unchanged.

Co-Authored-By: Claude Opus 4.8 <[email protected]>

src/main/kotlin/LargeLanguageModels.kt

@@ -146,22 +146,28 @@ object LargeLanguageModels {
 
         // 初始化推理模型
         if (PluginConfig.reasoningModelApi.isNotBlank() && PluginConfig.reasoningModelToken.isNotBlank()) {
+            // 推理模型出首块前常有思考预热，比对话慢，使用单独放宽的首块超时；
+            // socket 超时（两次读间隔，等首块时也归它管）不能小于首块预算，否则首块超时形同虚设
+            val reasoningFirstChunk = PluginConfig.reasoningFirstChunkTimeout.milliseconds
             reasoning = ModelService(
                 baseUrl = PluginConfig.reasoningModelApi,
                 token = PluginConfig.reasoningModelToken,
-                timeout = timeout,
-                firstChunkTimeout = firstChunkTimeout,
+                timeout = maxOf(timeout, reasoningFirstChunk),
+                firstChunkTimeout = reasoningFirstChunk,
                 extraBody = parseExtraBody(PluginConfig.reasoningModelExtraBody)
             )
         }
 
         // 初始化视觉模型
         if (PluginConfig.visualModelApi.isNotBlank() && PluginConfig.visualModelToken.isNotBlank()) {
+            // 视觉模型需服务端先下载图片再出首块，比对话天然慢，使用单独放宽的首块超时；
+            // socket 超时（两次读间隔，等首块时也归它管）不能小于首块预算，否则首块超时形同虚设
+            val visualFirstChunk = PluginConfig.visualFirstChunkTimeout.milliseconds
             visual = ModelService(
                 baseUrl = PluginConfig.visualModelApi,
                 token = PluginConfig.visualModelToken,
-                timeout = timeout,
-                firstChunkTimeout = firstChunkTimeout,
+                timeout = maxOf(timeout, visualFirstChunk),
+                firstChunkTimeout = visualFirstChunk,
                 extraBody = parseExtraBody(PluginConfig.visualModelExtraBody)
             )
         }

src/main/kotlin/PluginConfig.kt

@@ -105,6 +105,12 @@ object PluginConfig : AutoSavePluginConfig("Config") {
     @ValueDescription("首块响应超时时间，单位毫秒，默认10秒。若连接建立后在此时间内没收到首块data:则中断走重试")
     val firstChunkTimeout: Long by value(10000L)
 
+    @ValueDescription("视觉模型首块响应超时时间，单位毫秒，默认120秒。视觉模型需先下载图片再出首块，比对话天然慢，故单独放宽")
+    val visualFirstChunkTimeout: Long by value(120000L)
+
+    @ValueDescription("推理模型首块响应超时时间，单位毫秒，默认90秒。推理模型出首块前常有思考预热，比对话慢，故单独放宽")
+    val reasoningFirstChunkTimeout: Long by value(90000L)
+
     @Deprecated("使用外部文件而不是在配置文件内保存提示词")
     @ValueDescription("系统提示词，该字段已弃用，使用提示词文件而不是在这里修改")
     var prompt: String by value("你是一个乐于助人的助手")