models: use provider defaults for extraction tasks

src/main/kotlin/config/PluginConfig.kt

@@ -69,9 +69,6 @@ object PluginConfig : AutoSavePluginConfig("Config") {
     @ValueDescription("画像分析模型。留空时继承聊天模型")
     val profileModel: String by value("")
 
-    @ValueDescription("画像分析模型温度。留空时继承聊天模型温度")
-    val profileModelTemperature: Double? by value(null)
-
     @ValueDescription("画像分析模型额外请求体JSON。留空时继承聊天模型额外请求体")
     val profileModelExtraBody: String by value("")
 
@@ -198,9 +195,6 @@ object PluginConfig : AutoSavePluginConfig("Config") {
     @ValueDescription("网页摘要模型名称")
     val webSummaryModel: String by value("")
 
-    @ValueDescription("网页摘要模型温度，默认为0.1")
-    val webSummaryModelTemperature: Double? by value(0.1)
-
     @ValueDescription("网页摘要模型额外请求体JSON，会合并到请求体中")
     val webSummaryModelExtraBody: String by value("")
 

src/main/kotlin/llm/LargeLanguageModels.kt

@@ -30,13 +30,11 @@ object LargeLanguageModels {
     data class ProfileEndpoint(
         val service: ModelService,
         val model: String,
-        val temperature: Double?,
     )
 
     data class WebSummaryEndpoint(
         val service: ModelService,
         val model: String,
-        val temperature: Double?,
     )
 
     /**
@@ -184,7 +182,6 @@ object LargeLanguageModels {
                         maxConcurrentRequests = PluginConfig.profileMaxConcurrentRequests,
                 ),
                 model = model,
-                    temperature = PluginConfig.profileModelTemperature ?: PluginConfig.chatTemperature,
             )
             }
         }
@@ -204,7 +201,6 @@ object LargeLanguageModels {
                     extraBody = parseExtraBody(PluginConfig.webSummaryModelExtraBody),
                 ),
                 model = PluginConfig.webSummaryModel,
-                temperature = PluginConfig.webSummaryModelTemperature,
             )
         }
 

src/main/kotlin/profile/ProfileModelClient.kt

@@ -78,7 +78,6 @@ class ProfileModelClient(
         val completion = complete(
             ChatCompletionRequest(
                 model = ModelId(endpoint.model),
-                temperature = endpoint.temperature,
                 responseFormat = ChatResponseFormat.JsonObject,
                 streamOptions = StreamOptions(includeUsage = true),
                 messages = listOf(
@@ -111,7 +110,6 @@ class ProfileModelClient(
         val completion = complete(
             ChatCompletionRequest(
                 model = ModelId(endpoint.model),
-                temperature = endpoint.temperature,
                 responseFormat = ChatResponseFormat.JsonObject,
                 streamOptions = StreamOptions(includeUsage = true),
                 messages = listOf(
@@ -142,7 +140,6 @@ class ProfileModelClient(
         val completion = complete(
             ChatCompletionRequest(
                 model = ModelId(endpoint.model),
-                temperature = endpoint.temperature,
                 responseFormat = ChatResponseFormat.JsonObject,
                 streamOptions = StreamOptions(includeUsage = true),
                 messages = listOf(

src/main/kotlin/tools/VisitWeb.kt

@@ -211,7 +211,6 @@ class VisitWeb : BaseAgent(
             endpoint.service.chatCompletions(
                 ChatCompletionRequest(
                     model = ModelId(endpoint.model),
-                    temperature = endpoint.temperature,
                     messages = listOf(
                         ChatMessage.System(WEB_SUMMARY_SYSTEM_PROMPT),
                         ChatMessage.User(prompt),

src/test/kotlin/profile/ConversationProfileLiveExperiment.kt

@@ -51,7 +51,7 @@ class ConversationProfileLiveExperiment {
         )
         try {
             val client = ProfileModelClient(
-                LargeLanguageModels.ProfileEndpoint(service, modelName, temperature = 0.1)
+                LargeLanguageModels.ProfileEndpoint(service, modelName)
             )
             val result = client.analyzeConversation(profiles, batch, eligible)
             val reductions = ConversationProfileReducer.reduce(