From b31589064aaa4cebfe24e19514eb3b7218cce00e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Deniz=20=C5=9Eafak?= <denizsafak98@gmail.com>
Date: Wed, 18 Mar 2026 01:28:26 +0300
Subject: [PATCH] Fix jobs preview fallback priority when text and images
 coexist

---
 comfy_execution/jobs.py      | 18 ++++++++++++------
 tests/execution/test_jobs.py | 31 +++++++++++++++++++++++++++++++
 2 files changed, 43 insertions(+), 6 deletions(-)

diff --git a/comfy_execution/jobs.py b/comfy_execution/jobs.py
index fcd7ef735..b2e691470 100644
--- a/comfy_execution/jobs.py
+++ b/comfy_execution/jobs.py
@@ -223,11 +223,13 @@ def get_outputs_summary(outputs: dict) -> tuple[int, Optional[dict]]:
 
     Preview priority (matching frontend):
     1. type="output" with previewable media
-    2. Any previewable media
+    2. Any non-text previewable media
+    3. Text preview fallback
     """
     count = 0
     preview_output = None
     fallback_preview = None
+    text_fallback_preview = None
 
     for node_id, node_outputs in outputs.items():
         if not isinstance(node_outputs, dict):
@@ -256,8 +258,8 @@ def get_outputs_summary(outputs: dict) -> tuple[int, Optional[dict]]:
                                     'nodeId': node_id,
                                     'mediaType': media_type
                                 }
-                                if fallback_preview is None:
-                                    fallback_preview = enriched
+                                if text_fallback_preview is None:
+                                    text_fallback_preview = enriched
                         continue
                     # normalize_output_item returned a dict (e.g. 3D file)
                     item = normalized
@@ -276,10 +278,14 @@ def get_outputs_summary(outputs: dict) -> tuple[int, Optional[dict]]:
                         enriched['mediaType'] = media_type
                     if item.get('type') == 'output':
                         preview_output = enriched
-                    elif fallback_preview is None:
-                        fallback_preview = enriched
+                    elif media_type != 'text':
+                        # Prefer non-text media (images/video/audio/3d) over text fallback.
+                        if fallback_preview is None:
+                            fallback_preview = enriched
+                    elif text_fallback_preview is None:
+                        text_fallback_preview = enriched
 
-    return count, preview_output or fallback_preview
+    return count, preview_output or fallback_preview or text_fallback_preview
 
 
 def apply_sorting(jobs: list[dict], sort_by: str, sort_order: str) -> list[dict]:
diff --git a/tests/execution/test_jobs.py b/tests/execution/test_jobs.py
index 814af5c13..895c0006e 100644
--- a/tests/execution/test_jobs.py
+++ b/tests/execution/test_jobs.py
@@ -136,6 +136,37 @@ class TestGetOutputsSummary:
         count, preview = get_outputs_summary(outputs)
         assert preview['filename'] == 'temp1.png'
 
+    def test_text_preview_does_not_override_image_fallback(self):
+        """Text output should not take precedence over image fallback previews."""
+        outputs = {
+            'node1': {
+                'text': ['debug text from PreviewAny']
+            },
+            'node2': {
+                'images': [
+                    {'filename': 'temp1.png', 'type': 'temp'}
+                ]
+            }
+        }
+        count, preview = get_outputs_summary(outputs)
+        assert count == 2
+        assert preview is not None
+        assert preview['mediaType'] == 'images'
+        assert preview['filename'] == 'temp1.png'
+
+    def test_text_fallback_used_when_only_text_exists(self):
+        """Text preview should still be used when no non-text previewable outputs exist."""
+        outputs = {
+            'node1': {
+                'text': ['debug text from PreviewAny']
+            }
+        }
+        count, preview = get_outputs_summary(outputs)
+        assert count == 1
+        assert preview is not None
+        assert preview['mediaType'] == 'text'
+        assert preview['content'] == 'debug text from PreviewAny'
+
     def test_non_previewable_media_types_counted_but_no_preview(self):
         """Non-previewable media types should be counted but not used as preview."""
         outputs = {