From 476da53cf1a93bfd43f2198dafe035b69b3ab05c Mon Sep 17 00:00:00 2001
From: lxfight <1686540385@qq.com>
Date: Wed, 2 Sep 2026 14:23:20 +0800
Subject: [PATCH] fix: add missing KB parser routes for mdx/mkd/html/htm/csv
select_parser and the MarkdownChunker whitelist in kb_helper were
inconsistent: .mdx/.mkd were accepted by the chunker but had no parser
route, so uploads failed at the parsing stage with a generic unsupported
format error. .html/.htm/.csv were not routed at all although the bundled
markitdown-no-magika version ships HtmlConverter and CsvConverter.
Route .mdx/.mkd to TextParser (plain markdown variants, no extra
dependency) and .html/.htm/.csv to MarkitdownParser. .rtf intentionally
stays unsupported: markitdown-no-magika 0.1.2 registers no RTF converter
(verified: conversion raises UnsupportedFormatException).
---
astrbot/core/knowledge_base/parsers/util.py | 23 ++++++++++-
tests/test_kb_parser_routing.py | 45 +++++++++++++++++++++
2 files changed, 67 insertions(+), 1 deletion(-)
create mode 100644 tests/test_kb_parser_routing.py
diff --git a/astrbot/core/knowledge_base/parsers/util.py b/astrbot/core/knowledge_base/parsers/util.py
index a98ed60ed2..4d3eb53854 100644
--- a/astrbot/core/knowledge_base/parsers/util.py
+++ b/astrbot/core/knowledge_base/parsers/util.py
@@ -2,7 +2,28 @@
async def select_parser(ext: str) -> BaseParser:
- if ext in {".md", ".txt", ".markdown", ".rst", ".adoc", ".xlsx", ".docx", ".xls"}:
+ # Markdown variants that the MarkdownChunker whitelist in kb_helper
+ # already accepts; parse them as plain text without extra dependencies.
+ if ext in {".mdx", ".mkd"}:
+ from .text_parser import TextParser
+
+ return TextParser()
+ # Formats supported by markitdown-no-magika (verified converters:
+ # HtmlConverter, CsvConverter, PlainTextConverter; note it has no
+ # RTF converter, so .rtf stays unsupported).
+ if ext in {
+ ".csv",
+ ".html",
+ ".htm",
+ ".md",
+ ".txt",
+ ".markdown",
+ ".rst",
+ ".adoc",
+ ".xlsx",
+ ".docx",
+ ".xls",
+ }:
from .markitdown_parser import MarkitdownParser
return MarkitdownParser()
diff --git a/tests/test_kb_parser_routing.py b/tests/test_kb_parser_routing.py
new file mode 100644
index 0000000000..99a21d82a6
--- /dev/null
+++ b/tests/test_kb_parser_routing.py
@@ -0,0 +1,45 @@
+from __future__ import annotations
+
+import pytest
+
+from astrbot.core.knowledge_base.parsers.markitdown_parser import MarkitdownParser
+from astrbot.core.knowledge_base.parsers.text_parser import TextParser
+from astrbot.core.knowledge_base.parsers.util import select_parser
+
+
+@pytest.mark.parametrize("ext", [".mdx", ".mkd"])
+@pytest.mark.asyncio
+async def test_markdown_variants_use_text_parser(ext: str) -> None:
+ # These extensions are accepted by the MarkdownChunker whitelist in
+ # kb_helper but previously had no parser route, so uploads always
+ # failed with "暂时不支持的文件格式".
+ parser = await select_parser(ext)
+
+ assert isinstance(parser, TextParser)
+
+
+@pytest.mark.parametrize("ext", [".html", ".htm", ".csv"])
+@pytest.mark.asyncio
+async def test_html_csv_use_markitdown_parser(ext: str) -> None:
+ parser = await select_parser(ext)
+
+ assert isinstance(parser, MarkitdownParser)
+
+
+@pytest.mark.asyncio
+async def test_markitdown_parses_html_and_csv() -> None:
+ html = b"Title
Hello world
"
+ result = await MarkitdownParser().parse(html, "a.html")
+ assert "Title" in result.text and "Hello **world**" in result.text
+
+ csv = b"name,age\nalice,1\n"
+ result = await MarkitdownParser().parse(csv, "a.csv")
+ assert "alice" in result.text
+
+
+@pytest.mark.asyncio
+async def test_rtf_still_unsupported() -> None:
+ # markitdown-no-magika 0.1.2 has no RTF converter, so .rtf must keep
+ # raising instead of being routed to a parser that cannot handle it.
+ with pytest.raises(ValueError, match="暂时不支持的文件格式"):
+ await select_parser(".rtf")