feat(tools): stream shell_run output live into the TUI

internal/tui/app.go

@@ -683,15 +683,27 @@ func (m *model) mergeToolLog(text string) bool {
 	if incoming.output == "" {
 		return false
 	}
+	incomingHeader, _, ok := strings.Cut(text, "\n[output]\n")
+	if !ok {
+		return false
+	}
 	for i := len(m.messages) - 1; i >= 0; i-- {
 		if m.messages[i].role != roleTool {
 			continue
 		}
 		existing := parseToolLog(m.messages[i].content)
-		if existing.name == incoming.name && existing.args == incoming.args && existing.output == "" {
-			m.messages[i].content = text
-			return true
+		if existing.name != incoming.name || existing.args != incoming.args {
+			continue
 		}
+		if !strings.Contains(m.messages[i].content, "\n[output]\n") {
+			// Placeholder card from the tool-start log; first chunk fills it.
+			m.messages[i].content = text
+		} else {
+			// Streaming tools send incremental chunks; append the raw delta
+			// so line boundaries are preserved inside the same card.
+			m.messages[i].content += strings.TrimPrefix(text, incomingHeader+"\n[output]\n")
+		}
+		return true
 	}
 	return false
 }

internal/tui/app_test.go

@@ -395,6 +395,28 @@ func TestToolLogOutputUpdatesExistingToolCard(t *testing.T) {
 	}
 }
 
+func TestToolLogStreamingChunksAppendToSameCard(t *testing.T) {
+	m := newModel(context.Background(), nil, Options{ModelName: "test-model"})
+	m.messages = append(m.messages, message{role: roleTool, content: `shell_run echo hi`})
+
+	if !m.mergeToolLog("shell_run echo hi\n[output]\none\n") {
+		t.Fatal("expected first mergeToolLog to return true")
+	}
+	if !m.mergeToolLog("shell_run echo hi\n[output]\ntwo\n") {
+		t.Fatal("expected second mergeToolLog to return true")
+	}
+	if len(m.messages) != 1 {
+		t.Fatalf("messages count = %d, want 1", len(m.messages))
+	}
+	plain := stripANSI(m.renderMessages())
+	if !strings.Contains(plain, "one") || !strings.Contains(plain, "two") {
+		t.Fatalf("rendered output missing streamed lines:\n%s", plain)
+	}
+	if strings.Contains(plain, "onetwo") {
+		t.Fatalf("streamed lines were concatenated:\n%s", plain)
+	}
+}
+
 type tuiCompactProvider struct{}
 
 func (tuiCompactProvider) ChatStream(ctx context.Context, req llm.ChatRequest, emit func(llm.StreamEvent) error) error {