Fix: text manipulation for multimodal messages

bot.go

@@ -119,7 +119,7 @@ func processMessageTag(msg *models.RoleMsg) *models.RoleMsg {
 	}
 	// If KnownTo already set, assume tag already processed (content cleaned).
 	// However, we still check for new tags (maybe added later).
-	knownTo := parseKnownToTag(msg.Content)
+	knownTo := parseKnownToTag(msg.GetText())
 	// If tag found, replace KnownTo with new list (merge with existing?)
 	// For simplicity, if knownTo is not nil, replace.
 	if knownTo == nil {
@@ -1303,12 +1303,9 @@ func removeThinking(chatBody *models.ChatBody) {
 		if msg.Role == cfg.ToolRole {
 			continue
 		}
-		// find thinking and remove it
-		rm := models.RoleMsg{
-			Role:    msg.Role,
-			Content: thinkRE.ReplaceAllString(msg.Content, ""),
-		}
-		msgs = append(msgs, rm)
+		// find thinking and remove it - use SetText to preserve ContentParts
+		msg.SetText(thinkRE.ReplaceAllString(msg.GetText(), ""))
+		msgs = append(msgs, msg)
 	}
 	chatBody.Messages = msgs
 }

helpfuncs.go

@@ -75,15 +75,16 @@ func stripThinkingFromMsg(msg *models.RoleMsg) *models.RoleMsg {
 	if !cfg.StripThinkingFromAPI {
 		return msg
 	}
-	// Skip user, tool, and system messages - they might contain thinking examples
+	// Skip user, tool, they might contain thinking and system messages - examples
 	if msg.Role == cfg.UserRole || msg.Role == cfg.ToolRole || msg.Role == "system" {
 		return msg
 	}
 	// Strip thinking from assistant messages
-	if thinkRE.MatchString(msg.Content) {
-		msg.Content = thinkRE.ReplaceAllString(msg.Content, "")
-		// Clean up any double newlines that might result
-		msg.Content = strings.TrimSpace(msg.Content)
+	msgText := msg.GetText()
+	if thinkRE.MatchString(msgText) {
+		cleanedText := thinkRE.ReplaceAllString(msgText, "")
+		cleanedText = strings.TrimSpace(cleanedText)
+		msg.SetText(cleanedText)
 	}
 	return msg
 }

models/models.go

@@ -329,6 +329,66 @@ func (m *RoleMsg) Copy() RoleMsg {
 	}
 }
 
+// GetText returns the text content of the message, handling both
+// simple Content and multimodal ContentParts formats.
+func (m *RoleMsg) GetText() string {
+	if !m.hasContentParts {
+		return m.Content
+	}
+	var textParts []string
+	for _, part := range m.ContentParts {
+		switch p := part.(type) {
+		case TextContentPart:
+			if p.Type == "text" {
+				textParts = append(textParts, p.Text)
+			}
+		case map[string]any:
+			if partType, exists := p["type"]; exists {
+				if partType == "text" {
+					if textVal, textExists := p["text"]; textExists {
+						if textStr, isStr := textVal.(string); isStr {
+							textParts = append(textParts, textStr)
+						}
+					}
+				}
+			}
+		}
+	}
+	return strings.Join(textParts, " ")
+}
+
+// SetText updates the text content of the message. If the message has
+// ContentParts (multimodal), it updates the text parts while preserving
+// images. If not, it sets the simple Content field.
+func (m *RoleMsg) SetText(text string) {
+	if !m.hasContentParts {
+		m.Content = text
+		return
+	}
+	var newParts []any
+	for _, part := range m.ContentParts {
+		switch p := part.(type) {
+		case TextContentPart:
+			if p.Type == "text" {
+				p.Text = text
+				newParts = append(newParts, p)
+			} else {
+				newParts = append(newParts, p)
+			}
+		case map[string]any:
+			if partType, exists := p["type"]; exists && partType == "text" {
+				p["text"] = text
+				newParts = append(newParts, p)
+			} else {
+				newParts = append(newParts, p)
+			}
+		default:
+			newParts = append(newParts, part)
+		}
+	}
+	m.ContentParts = newParts
+}
+
 // AddTextPart adds a text content part to the message
 func (m *RoleMsg) AddTextPart(text string) {
 	if !m.hasContentParts {

tui.go

@@ -264,7 +264,7 @@ func init() {
 				pages.RemovePage(editMsgPage)
 				return nil
 			}
-			chatBody.Messages[selectedIndex].Content = editedMsg
+			chatBody.Messages[selectedIndex].SetText(editedMsg)
 			// change textarea
 			textView.SetText(chatToText(chatBody.Messages, cfg.ShowSys))
 			pages.RemovePage(editMsgPage)
@@ -352,13 +352,14 @@ func init() {
 			case editMode:
 				hideIndexBar() // Hide overlay first
 				pages.AddPage(editMsgPage, editArea, true, true)
-				editArea.SetText(m.Content, true)
+				editArea.SetText(m.GetText(), true)
 			default:
-				if err := copyToClipboard(m.Content); err != nil {
+				msgText := m.GetText()
+				if err := copyToClipboard(msgText); err != nil {
 					logger.Error("failed to copy to clipboard", "error", err)
 				}
-				previewLen := min(30, len(m.Content))
-				notification := fmt.Sprintf("msg '%s' was copied to the clipboard", m.Content[:previewLen])
+				previewLen := min(30, len(msgText))
+				notification := fmt.Sprintf("msg '%s' was copied to the clipboard", msgText[:previewLen])
 				if err := notifyUser("copied", notification); err != nil {
 					logger.Error("failed to send notification", "error", err)
 				}
@@ -648,11 +649,12 @@ func init() {
 			// copy msg to clipboard
 			editMode = false
 			m := chatBody.Messages[len(chatBody.Messages)-1]
-			if err := copyToClipboard(m.Content); err != nil {
+			msgText := m.GetText()
+			if err := copyToClipboard(msgText); err != nil {
 				logger.Error("failed to copy to clipboard", "error", err)
 			}
-			previewLen := min(30, len(m.Content))
-			notification := fmt.Sprintf("msg '%s' was copied to the clipboard", m.Content[:previewLen])
+			previewLen := min(30, len(msgText))
+			notification := fmt.Sprintf("msg '%s' was copied to the clipboard", msgText[:previewLen])
 			if err := notifyUser("copied", notification); err != nil {
 				logger.Error("failed to send notification", "error", err)
 			}
@@ -847,7 +849,7 @@ func init() {
 				// Stop any currently playing TTS first
 				TTSDoneChan <- true
 				lastMsg := chatBody.Messages[len(chatBody.Messages)-1]
-				cleanedText := models.CleanText(lastMsg.Content)
+				cleanedText := models.CleanText(lastMsg.GetText())
 				if cleanedText != "" {
 					// nolint: errcheck
 					go orator.Speak(cleanedText)