From f56940139855c183167cd7cd1397ac767a30f1cb Mon Sep 17 00:00:00 2001
From: Stephen Hu <stephenhu@seismic.com>
Date: Fri, 11 Jul 2025 18:21:39 +0800
Subject: [PATCH] Fix: better_handle_different_types (#8775)

### What problem does this PR solve?


https://github.com/infiniflow/ragflow/issues/8719#issuecomment-3055883271

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
---
 rag/nlp/__init__.py | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/rag/nlp/__init__.py b/rag/nlp/__init__.py
index 9a24d48b2..fd302fac3 100644
--- a/rag/nlp/__init__.py
+++ b/rag/nlp/__init__.py
@@ -589,17 +589,21 @@ def naive_merge_with_images(texts, images, chunk_token_num=128, delimiter="\n。
 
     dels = get_delimiters(delimiter)
     for text, image in zip(texts, images):
-        splited_sec = re.split(r"(%s)" % dels, text)
-        for sub_sec in splited_sec:
-            if re.match(f"^{dels}$", sub_sec):
-                continue
         # if text is tuple, unpack it
         if isinstance(text, tuple):
             text_str = text[0]
             text_pos = text[1] if len(text) > 1 else ""
-            add_chunk(text_str, image, text_pos)
+            splited_sec = re.split(r"(%s)" % dels, text_str)
+            for sub_sec in splited_sec:
+                if re.match(f"^{dels}$", sub_sec):
+                    continue
+                add_chunk(sub_sec, image, text_pos)
         else:
-            add_chunk(text, image)
+            splited_sec = re.split(r"(%s)" % dels, text)
+            for sub_sec in splited_sec:
+                if re.match(f"^{dels}$", sub_sec):
+                    continue
+                add_chunk(sub_sec, image)
 
     return cks, result_images