From 107fe6cf90452e54e53116d6fa8d17ec512de0a2 Mon Sep 17 00:00:00 2001
From: Magicbook1108 <newyorkupperbay@gmail.com>
Date: Thu, 9 Apr 2026 16:40:42 +0800
Subject: [PATCH] Feat: support doc for pipeline parser in word (#14005)

### What problem does this PR solve?

Feat: support doc for pipeline parser in word

### Type of change

- [x] New Feature (non-breaking change which adds functionality)



<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **New Features**
* Added support for processing legacy Word `.doc` file formats,
extending document compatibility.

* **Bug Fixes**
* Enhanced error handling during document parsing to improve reliability
and prevent processing failures.
---
 rag/flow/parser/parser.py | 37 ++++++++++++++++++++++++++++++++++++-
 1 file changed, 36 insertions(+), 1 deletion(-)

diff --git a/rag/flow/parser/parser.py b/rag/flow/parser/parser.py
index bbce185c39..81562781d8 100644
--- a/rag/flow/parser/parser.py
+++ b/rag/flow/parser/parser.py
@@ -741,6 +741,41 @@ class Parser(ProcessBase):
         self.callback(random.randint(1, 5) / 100.0, "Start to work on a Word Processor Document")
         conf = self._param.setups["word"]
         self.set_output("output_format", conf["output_format"])
+        
+        if re.search(r"\.doc$", name, re.IGNORECASE):
+            self.set_output("file", {**kwargs.get("file", {}), "outlines": []})
+            try:
+                from tika import parser as tika_parser
+            except Exception as e:
+                msg = f"tika not available: {e}. Unsupported .doc parsing."
+                self.callback(0.8, msg)
+                logging.warning(f"{msg} for {name}.")
+                return
+
+            doc_parsed = tika_parser.from_buffer(io.BytesIO(blob))
+            content = doc_parsed.get("content")
+            if content is None:
+                msg = f"tika.parser got empty content from {name}."
+                self.callback(0.8, msg)
+                logging.warning(msg)
+                return
+
+            sections = [line.strip() for line in content.splitlines() if line and line.strip()]
+            if conf.get("remove_toc"):
+                sections = remove_toc_word(sections, outlines)
+
+            if conf.get("output_format") == "json":
+                self.set_output(
+                    "json",
+                    [{"text": line, "image": None, "doc_type_kwd": "text"} for line in sections],
+                )
+            elif conf.get("output_format") == "markdown":
+                # Tika gives us plain text lines, so join with blank lines to preserve paragraph boundaries in markdown.
+                self.set_output("markdown", "\n\n".join(sections))
+
+            self.callback(0.8, "Finish parsing.")
+            return
+
         docx_parser = Docx()
 
         # Extract heading-based outlines for metadata and TOC removal.
@@ -784,7 +819,7 @@ class Parser(ProcessBase):
             markdown_text = docx_parser.to_markdown(name, binary=blob)
             if conf.get("remove_toc"):
                 markdown_text = "\n".join(remove_toc_word(markdown_text.split("\n"), outlines))
-                 
+
             self.set_output("markdown", markdown_text)
 
     def _slides(self, name, blob, **kwargs):