Enhance chunking strategies and graph utilities for section-type transitions and block ID extraction
- Implemented WP-26 v1.1: Section-Type-Wechsel erzwingt Split auch in SMART MODE (Schritt 2) zur Verbesserung der Chunking-Logik. - Updated `parse_link_target` to extract block IDs from section strings, ensuring accurate handling of links with block references. - Added unit tests to validate section-type change behavior and block ID extraction functionality, enhancing overall reliability.
This commit is contained in:
1 parent
af3cc0a254
commit
52fdc425f7
3 files changed
+121
No files matched your search
@@ -477,6 +477,81 @@ Meine positive Einstellung hat mir geholfen.
|
||||
# Der insight-Chunk sollte den Lektions-Inhalt enthalten
|
||||
insight_text = insight_chunks[0].text
|
||||
assert "durchdenken" in insight_text.lower() or "positive" in insight_text.lower()
|
||||
|
||||
def test_section_type_change_in_smart_mode_forces_split(self):
|
||||
"""WP-26 v1.1 Fix: Section-Type-Wechsel erzwingt Split auch in SMART MODE (Schritt 2)"""
|
||||
md = """
|
||||
## Section A ohne Typ
|
||||
|
||||
Inhalt A ohne section_type.
|
||||
|
||||
## Section B ohne Typ
|
||||
|
||||
Inhalt B ohne section_type.
|
||||
|
||||
## Section C mit Typ
|
||||
> [!section] insight
|
||||
|
||||
Inhalt C mit section_type "insight".
|
||||
"""
|
||||
blocks, _ = parse_blocks(md)
|
||||
|
||||
# SMART MODE: strict=False, smart_edge=True
|
||||
# Token-Limit hoch genug, dass alles zusammengefasst werden KÖNNTE
|
||||
config = {
|
||||
"target": 2000,
|
||||
"max": 4000,
|
||||
"split_level": 2,
|
||||
"strict_heading_split": False,
|
||||
"enable_smart_edge_allocation": True
|
||||
}
|
||||
|
||||
chunks = strategy_by_heading(blocks, config, "test-note")
|
||||
|
||||
# Trotz hohem Token-Limit sollte Section C ein separater Chunk sein
|
||||
# wegen Section-Type-Wechsel (None -> insight)
|
||||
assert len(chunks) >= 2, f"Erwartet mindestens 2 Chunks, bekommen: {len(chunks)}"
|
||||
|
||||
# Der letzte Chunk sollte section_type "insight" haben
|
||||
insight_chunks = [c for c in chunks if c.section_type == "insight"]
|
||||
assert len(insight_chunks) >= 1, "Kein Chunk mit section_type 'insight' gefunden"
|
||||
|
||||
|
||||
class TestBlockIdParsing:
|
||||
"""UT-18: Block-ID-Extraktion aus Section-Referenzen"""
|
||||
|
||||
def test_block_id_extraction_from_section(self):
|
||||
"""Block-ID wird aus Section-String extrahiert"""
|
||||
from app.core.graph.graph_utils import parse_link_target
|
||||
|
||||
# Test: Überschrift mit Block-ID
|
||||
target, section = parse_link_target("#📖 Diagnose: Glioblastom ^kontext", "note1")
|
||||
assert target == "note1" # Self-Link
|
||||
assert section == "kontext", f"Erwartet 'kontext', bekommen: {section}"
|
||||
|
||||
def test_block_id_extraction_only_caret(self):
|
||||
"""Nur Block-ID mit ^"""
|
||||
from app.core.graph.graph_utils import parse_link_target
|
||||
|
||||
target, section = parse_link_target("#^learning", "note1")
|
||||
assert target == "note1"
|
||||
assert section == "learning"
|
||||
|
||||
def test_block_id_extraction_with_spaces(self):
|
||||
"""Block-ID mit Text davor"""
|
||||
from app.core.graph.graph_utils import parse_link_target
|
||||
|
||||
target, section = parse_link_target("OtherNote#🎭 Emotions-Check ^emotionen", None)
|
||||
assert target == "OtherNote"
|
||||
assert section == "emotionen"
|
||||
|
||||
def test_section_without_block_id(self):
|
||||
"""Section ohne Block-ID bleibt unverändert"""
|
||||
from app.core.graph.graph_utils import parse_link_target
|
||||
|
||||
target, section = parse_link_target("Note#Normale Überschrift", None)
|
||||
assert target == "Note"
|
||||
assert section == "Normale Überschrift"
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
Reference in new issue
Block a user