feat(llm): implement getUsedMaxToken method

Implement getUsedMaxToken method to return ModelLimits based on current model configuration. Method prioritizes GitHub Copilot model limits when available, falls back to settings max token length for default and custom models.

Author: phodal

core/src/main/kotlin/cc/unitmesh/devti/settings/model/LLMModelManager.kt

@@ -3,6 +3,7 @@ package cc.unitmesh.devti.settings.model
 import cc.unitmesh.devti.llm2.GithubCopilotManager
 import cc.unitmesh.devti.llm2.LLMProvider2
 import cc.unitmesh.devti.llm2.model.LlmConfig
+import cc.unitmesh.devti.llm2.model.ModelLimits
 import cc.unitmesh.devti.llms.custom.Message
 import cc.unitmesh.devti.settings.AutoDevSettingsState
 import cc.unitmesh.devti.settings.dialog.LLMDialog
@@ -73,7 +74,38 @@ class LLMModelManager(
 
         return models
     }
-    
+
+    fun getUsedMaxToken(): ModelLimits {
+        val settingsState = AutoDevSettingsState.getInstance()
+        val modelId = settingsState.defaultModelId
+        
+        // If no model ID is set, use the default max token length
+        if (modelId.isEmpty() || modelId == "Default") {
+            return ModelLimits(
+                maxContextWindowTokens = settingsState.fetchMaxTokenLength(),
+                maxPromptTokens = settingsState.fetchMaxTokenLength(),
+                maxOutputTokens = settingsState.fetchMaxTokenLength()
+            )
+        }
+        
+        // Check if it's a GitHub Copilot model
+        val manager = service<GithubCopilotManager>()
+        if (manager.isInitialized()) {
+            val githubModels = manager.getSupportedModels(forceRefresh = false)
+            val githubModel = githubModels?.find { it.id == modelId }
+            if (githubModel != null && githubModel.capabilities.limits != null) {
+                return githubModel.capabilities.limits
+            }
+        }
+        
+        // For custom models, use the settings max token length
+        return ModelLimits(
+            maxContextWindowTokens = settingsState.fetchMaxTokenLength(),
+            maxPromptTokens = settingsState.fetchMaxTokenLength(),
+            maxOutputTokens = settingsState.fetchMaxTokenLength()
+        )
+    }
+
     /**
      * Get provider name from model ID
      * Used by AutoDevInputSection when model selection changes