Enha: add (loaded) suffix if model is loaded

bot.go

@@ -403,6 +403,23 @@ func fetchLCPModels() ([]string, error) {
 	return localModels, nil
 }
 
+// fetchLCPModelsWithLoadStatus returns models with "(loaded)" indicator for loaded models
+func fetchLCPModelsWithLoadStatus() ([]string, error) {
+	models, err := fetchLCPModelsWithStatus()
+	if err != nil {
+		return nil, err
+	}
+	result := make([]string, 0, len(models.Data))
+	for _, m := range models.Data {
+		modelName := m.ID
+		if m.Status.Value == "loaded" {
+			modelName = modelName + " (loaded)"
+		}
+		result = append(result, modelName)
+	}
+	return result, nil
+}
+
 // fetchLCPModelsWithStatus returns the full LCPModels struct including status information.
 func fetchLCPModelsWithStatus() (*models.LCPModels, error) {
 	resp, err := http.Get(cfg.FetchModelNameAPI)
@@ -832,6 +849,7 @@ func chatRound(r *models.ChatRoundReq) error {
 	// Variables for handling thinking blocks during streaming
 	inThinkingBlock := false
 	thinkingBuffer := strings.Builder{}
+	justExitedThinkingCollapsed := false
 out:
 	for {
 		select {
@@ -859,6 +877,7 @@ out:
 					if thinkingCollapsed {
 						// Thinking already displayed as placeholder, just update respText
 						respText.WriteString(chunk)
+						justExitedThinkingCollapsed = true
 						if scrollToEndEnabled {
 							textView.ScrollToEnd()
 						}
@@ -872,6 +891,11 @@ out:
 				}
 				// If not collapsed, fall through to normal display
 			}
+			// Add spacing after collapsed thinking block before real response
+			if justExitedThinkingCollapsed {
+				chunk = "\n\n" + chunk
+				justExitedThinkingCollapsed = false
+			}
 			fmt.Fprint(textView, chunk)
 			respText.WriteString(chunk)
 			// Update the message in chatBody.Messages so it persists during Alt+T
@@ -1197,6 +1221,16 @@ func chatToText(messages []models.RoleMsg, showSys bool) string {
 			}
 			return "[yellow::i][thinking... (press Alt+T to expand)][-:-:-]"
 		})
+		// Handle incomplete thinking blocks (during streaming when </think> hasn't arrived yet)
+		if strings.Contains(text, "<think>") && !strings.Contains(text, "</think>") {
+			// Find the incomplete thinking block and replace it
+			startIdx := strings.Index(text, "<think>")
+			if startIdx != -1 {
+				content := text[startIdx+len("<think>"):]
+				placeholder := fmt.Sprintf("[yellow::i][thinking... (%d chars) (press Alt+T to expand)][-:-:-]", len(content))
+				text = text[:startIdx] + placeholder
+			}
+		}
 	}
 
 	return text

popups.go

@@ -17,10 +17,14 @@ func showModelSelectionPopup() {
 		} else if strings.Contains(api, "openrouter.ai") {
 			return ORFreeModels
 		}
-		// Assume local llama.cpp
-		updateModelLists()
+		// Assume local llama.cpp - fetch with load status
+		models, err := fetchLCPModelsWithLoadStatus()
+		if err != nil {
+			logger.Error("failed to fetch models with load status", "error", err)
 			return LocalModels
 		}
+		return models
+	}
 	// Get the current model list based on the API
 	modelList := getModelListForAPI(cfg.CurrentAPI)
 	// Check for empty options list
@@ -57,8 +61,10 @@ func showModelSelectionPopup() {
 		modelListWidget.SetCurrentItem(currentModelIndex)
 	}
 	modelListWidget.SetSelectedFunc(func(index int, mainText string, secondaryText string, shortcut rune) {
+		// Strip "(loaded)" suffix if present for local llama.cpp models
+		modelName := strings.TrimSuffix(mainText, " (loaded)")
 		// Update the model in both chatBody and config
-		chatBody.Model = mainText
+		chatBody.Model = modelName
 		cfg.CurrentModel = chatBody.Model
 		// Remove the popup page
 		pages.RemovePage("modelSelectionPopup")