Verbesserung des Chunking-Parsers zur Unterstützung von H1-Überschriften und Anpassung der Metadatenlogik. Implementierung einer atomaren Sektions-Chunking-Strategie, die Überschriften und deren Inhalte zusammenhält.
This commit is contained in:
1 parent
8f5eb36b5f
commit
838083b909
2 files changed
+61
-48
No files matched your search
@@ -17,53 +17,58 @@ def split_sentences(text: str) -> list[str]:
|
||||
return [p.strip() for p in _SENT_SPLIT.split(text) if p.strip()]
|
||||
|
||||
def parse_blocks(md_text: str) -> Tuple[List[RawBlock], str]:
|
||||
"""Zerlegt Text in logische Einheiten."""
|
||||
"""Zerlegt Text in logische Einheiten, inklusive H1."""
|
||||
blocks = []
|
||||
h1_title = "Dokument"; section_path = "/"; current_h2 = None
|
||||
h1_title = "Dokument"; section_path = "/"; current_section_title = None
|
||||
fm, text_without_fm = extract_frontmatter_from_text(md_text)
|
||||
|
||||
# H1 für Note-Metadaten extrahieren
|
||||
h1_match = re.search(r'^#\s+(.*)', text_without_fm, re.MULTILINE)
|
||||
if h1_match: h1_title = h1_match.group(1).strip()
|
||||
|
||||
lines = text_without_fm.split('\n')
|
||||
buffer = []
|
||||
|
||||
for line in lines:
|
||||
stripped = line.strip()
|
||||
|
||||
# H1 ignorieren (ist Doc Title)
|
||||
if stripped.startswith('# '):
|
||||
continue
|
||||
|
||||
# Generische Heading-Erkennung (H2 bis H6) für flexible Split-Levels
|
||||
heading_match = re.match(r'^(#{2,6})\s+(.*)', stripped)
|
||||
# Heading-Erkennung (H1 bis H6)
|
||||
heading_match = re.match(r'^(#{1,6})\s+(.*)', stripped)
|
||||
if heading_match:
|
||||
# Buffer leeren (vorherigen Text abschließen)
|
||||
if buffer:
|
||||
content = "\n".join(buffer).strip()
|
||||
if content: blocks.append(RawBlock("paragraph", content, None, section_path, current_h2))
|
||||
if content:
|
||||
blocks.append(RawBlock("paragraph", content, None, section_path, current_section_title))
|
||||
buffer = []
|
||||
|
||||
level = len(heading_match.group(1))
|
||||
title = heading_match.group(2).strip()
|
||||
|
||||
# Pfad-Logik: H2 setzt den Haupt-Pfad
|
||||
if level == 2:
|
||||
current_h2 = title
|
||||
section_path = f"/{current_h2}"
|
||||
# Bei H3+ bleibt der section_path beim Parent, aber das Level wird korrekt gesetzt
|
||||
# Metadaten-Update
|
||||
if level == 1:
|
||||
current_section_title = title
|
||||
section_path = "/"
|
||||
elif level == 2:
|
||||
current_section_title = title
|
||||
section_path = f"/{current_section_title}"
|
||||
|
||||
blocks.append(RawBlock("heading", stripped, level, section_path, current_h2))
|
||||
|
||||
elif not stripped:
|
||||
blocks.append(RawBlock("heading", stripped, level, section_path, current_section_title))
|
||||
continue
|
||||
|
||||
if not stripped:
|
||||
if buffer:
|
||||
content = "\n".join(buffer).strip()
|
||||
if content: blocks.append(RawBlock("paragraph", content, None, section_path, current_h2))
|
||||
if content:
|
||||
blocks.append(RawBlock("paragraph", content, None, section_path, current_section_title))
|
||||
buffer = []
|
||||
else:
|
||||
buffer.append(line)
|
||||
|
||||
if buffer:
|
||||
content = "\n".join(buffer).strip()
|
||||
if content: blocks.append(RawBlock("paragraph", content, None, section_path, current_h2))
|
||||
if content:
|
||||
blocks.append(RawBlock("paragraph", content, None, section_path, current_section_title))
|
||||
|
||||
return blocks, h1_title
|
||||
|
||||
def parse_edges_robust(text: str) -> Set[str]:
|
||||
|
||||
Reference in new issue
Block a user