Implement WP-26 v1.0 - Phase 2: Enhance edge scoring and aggregation configuration
- Introduced configurable edge scoring with internal and external boosts for intra-note edges. - Added aggregation configuration to support note-level and chunk-level retrieval strategies. - Updated retriever and graph subgraph modules to utilize new scoring and aggregation logic. - Enhanced YAML configuration to include new parameters for edge scoring and aggregation levels. - Added boolean indexing for filtering based on edge properties in the setup script.
This commit is contained in:
1 parent
e86e9f2313
commit
c5215e22e7
5 files changed
+384
-19
No files matched your search
@@ -0,0 +1,240 @@
|
||||
"""
|
||||
FILE: tests/test_wp26_phase2_retriever.py
|
||||
DESCRIPTION: Unit-Tests für WP-26 Phase 2: Retriever-Anpassungen
|
||||
- is_internal-Boost für Intra-Note-Edges
|
||||
- Konfigurierbare Aggregation (Note/Chunk Level)
|
||||
VERSION: 1.0.0
|
||||
"""
|
||||
import pytest
|
||||
from unittest.mock import patch, MagicMock
|
||||
import os
|
||||
|
||||
|
||||
class TestEdgeScoringConfig:
|
||||
"""UT-19: Edge-Scoring-Konfiguration"""
|
||||
|
||||
def test_get_edge_scoring_config_defaults(self):
|
||||
"""Default-Werte werden korrekt geladen"""
|
||||
from app.core.graph.graph_subgraph import get_edge_scoring_config
|
||||
|
||||
# Cache leeren
|
||||
get_edge_scoring_config.cache_clear()
|
||||
|
||||
# Mit nicht-existierender Config-Datei
|
||||
with patch.dict(os.environ, {"MINDNET_RETRIEVER_CONFIG": "/nonexistent/path.yaml"}):
|
||||
get_edge_scoring_config.cache_clear()
|
||||
config = get_edge_scoring_config()
|
||||
|
||||
assert config["internal_edge_boost"] == 1.2
|
||||
assert config["external_edge_boost"] == 1.0
|
||||
|
||||
def test_get_edge_scoring_config_from_yaml(self):
|
||||
"""Werte werden aus YAML geladen"""
|
||||
from app.core.graph.graph_subgraph import get_edge_scoring_config
|
||||
|
||||
# Cache leeren und echte Config laden
|
||||
get_edge_scoring_config.cache_clear()
|
||||
|
||||
# Mit echter Config-Datei
|
||||
config = get_edge_scoring_config()
|
||||
|
||||
# Die Werte sollten den Defaults entsprechen (aus retriever.yaml)
|
||||
assert config["internal_edge_boost"] >= 1.0
|
||||
assert config["external_edge_boost"] >= 1.0
|
||||
|
||||
|
||||
class TestIsInternalBoost:
|
||||
"""UT-20: is_internal-Boost im Subgraph"""
|
||||
|
||||
def test_internal_edge_gets_boost(self):
|
||||
"""Intra-Note-Edges erhalten höheres Gewicht"""
|
||||
from app.core.graph.graph_subgraph import Subgraph, get_edge_scoring_config
|
||||
|
||||
# Cache leeren
|
||||
get_edge_scoring_config.cache_clear()
|
||||
|
||||
sg = Subgraph()
|
||||
|
||||
# Interne Edge (innerhalb derselben Note)
|
||||
sg.add_edge({
|
||||
"source": "note1#c01",
|
||||
"target": "note1#c02",
|
||||
"kind": "derives",
|
||||
"weight": 1.0,
|
||||
"is_internal": True
|
||||
})
|
||||
|
||||
# Prüfe, dass das Gewicht erhöht wurde
|
||||
edges = sg.adj.get("note1#c01", [])
|
||||
assert len(edges) == 1
|
||||
|
||||
internal_boost = get_edge_scoring_config()["internal_edge_boost"]
|
||||
assert edges[0]["weight"] == 1.0 * internal_boost
|
||||
assert edges[0]["is_internal"] is True
|
||||
|
||||
def test_external_edge_no_boost(self):
|
||||
"""Inter-Note-Edges erhalten keinen Boost"""
|
||||
from app.core.graph.graph_subgraph import Subgraph, get_edge_scoring_config
|
||||
|
||||
# Cache leeren
|
||||
get_edge_scoring_config.cache_clear()
|
||||
|
||||
sg = Subgraph()
|
||||
|
||||
# Externe Edge (zwischen verschiedenen Notes)
|
||||
sg.add_edge({
|
||||
"source": "note1#c01",
|
||||
"target": "note2#c01",
|
||||
"kind": "references",
|
||||
"weight": 1.0,
|
||||
"is_internal": False
|
||||
})
|
||||
|
||||
edges = sg.adj.get("note1#c01", [])
|
||||
assert len(edges) == 1
|
||||
|
||||
external_boost = get_edge_scoring_config()["external_edge_boost"]
|
||||
assert edges[0]["weight"] == 1.0 * external_boost
|
||||
assert edges[0]["is_internal"] is False
|
||||
|
||||
def test_edge_bonus_aggregation_with_internal(self):
|
||||
"""Edge-Bonus aggregiert korrekt mit is_internal-Boost"""
|
||||
from app.core.graph.graph_subgraph import Subgraph, get_edge_scoring_config
|
||||
|
||||
get_edge_scoring_config.cache_clear()
|
||||
sg = Subgraph()
|
||||
|
||||
# Zwei Edges: eine interne, eine externe
|
||||
sg.add_edge({
|
||||
"source": "note1",
|
||||
"target": "note2",
|
||||
"kind": "solves",
|
||||
"weight": 1.5,
|
||||
"is_internal": True
|
||||
})
|
||||
sg.add_edge({
|
||||
"source": "note1",
|
||||
"target": "note3",
|
||||
"kind": "references",
|
||||
"weight": 0.1,
|
||||
"is_internal": False
|
||||
})
|
||||
|
||||
# Aggregierter Bonus
|
||||
bonus = sg.edge_bonus("note1")
|
||||
|
||||
# Sollte > 0 sein
|
||||
assert bonus > 0
|
||||
|
||||
|
||||
class TestAggregationConfig:
|
||||
"""UT-21: Aggregation-Konfiguration"""
|
||||
|
||||
def test_get_aggregation_config_defaults(self):
|
||||
"""Default-Werte werden korrekt geladen"""
|
||||
from app.core.retrieval.retriever import _get_aggregation_config
|
||||
|
||||
# Mit nicht-existierender Config-Datei
|
||||
with patch.dict(os.environ, {"MINDNET_RETRIEVER_CONFIG": "/nonexistent/path.yaml"}):
|
||||
config = _get_aggregation_config()
|
||||
|
||||
assert config["level"] == "note"
|
||||
assert config["max_chunks_per_note"] == 3
|
||||
|
||||
def test_get_aggregation_config_from_yaml(self):
|
||||
"""Werte werden aus YAML geladen"""
|
||||
from app.core.retrieval.retriever import _get_aggregation_config
|
||||
|
||||
config = _get_aggregation_config()
|
||||
|
||||
# Die Werte sollten aus retriever.yaml kommen
|
||||
assert config["level"] in ["note", "chunk"]
|
||||
assert config["max_chunks_per_note"] >= 1
|
||||
|
||||
|
||||
class TestNoteLevelAggregation:
|
||||
"""UT-22: Note-Level Aggregation mit max_chunks_per_note"""
|
||||
|
||||
def test_note_level_limits_chunks(self):
|
||||
"""Note-Level-Aggregation limitiert Chunks pro Note"""
|
||||
# Mock-Daten: 5 Chunks von Note1, 3 Chunks von Note2
|
||||
mock_hits = [
|
||||
("c1", 0.9, {"note_id": "note1", "chunk_id": "c1"}),
|
||||
("c2", 0.85, {"note_id": "note1", "chunk_id": "c2"}),
|
||||
("c3", 0.8, {"note_id": "note2", "chunk_id": "c3"}),
|
||||
("c4", 0.75, {"note_id": "note1", "chunk_id": "c4"}),
|
||||
("c5", 0.7, {"note_id": "note2", "chunk_id": "c5"}),
|
||||
("c6", 0.65, {"note_id": "note1", "chunk_id": "c6"}),
|
||||
("c7", 0.6, {"note_id": "note1", "chunk_id": "c7"}),
|
||||
]
|
||||
|
||||
# Simuliere Note-Level-Aggregation mit max_chunks_per_note=2
|
||||
max_chunks_per_note = 2
|
||||
pooled = []
|
||||
note_count = {}
|
||||
|
||||
for pid, score, payload in sorted(mock_hits, key=lambda x: x[1], reverse=True):
|
||||
note_id = payload["note_id"]
|
||||
if note_count.get(note_id, 0) < max_chunks_per_note:
|
||||
pooled.append((pid, score, payload))
|
||||
note_count[note_id] = note_count.get(note_id, 0) + 1
|
||||
|
||||
# Erwartung: 2 von note1, 2 von note2 = 4 Chunks
|
||||
assert len(pooled) == 4
|
||||
|
||||
# Prüfe, dass jede Note maximal 2 Chunks hat
|
||||
note1_chunks = [p for p in pooled if p[2]["note_id"] == "note1"]
|
||||
note2_chunks = [p for p in pooled if p[2]["note_id"] == "note2"]
|
||||
assert len(note1_chunks) == 2
|
||||
assert len(note2_chunks) == 2
|
||||
|
||||
|
||||
class TestChunkLevelAggregation:
|
||||
"""UT-23: Chunk-Level Aggregation (keine Deduplizierung)"""
|
||||
|
||||
def test_chunk_level_no_dedup(self):
|
||||
"""Chunk-Level-Aggregation gibt alle Chunks zurück"""
|
||||
mock_hits = [
|
||||
("c1", 0.9, {"note_id": "note1"}),
|
||||
("c2", 0.85, {"note_id": "note1"}),
|
||||
("c3", 0.8, {"note_id": "note1"}),
|
||||
("c4", 0.75, {"note_id": "note1"}),
|
||||
("c5", 0.7, {"note_id": "note1"}),
|
||||
]
|
||||
|
||||
# Chunk-Level: Keine Deduplizierung
|
||||
aggregation_level = "chunk"
|
||||
|
||||
if aggregation_level == "chunk":
|
||||
pooled = mock_hits
|
||||
else:
|
||||
pooled = [] # Note-Level würde nur 1 behalten
|
||||
|
||||
# Alle 5 Chunks sollten erhalten bleiben
|
||||
assert len(pooled) == 5
|
||||
|
||||
|
||||
class TestQdrantIndexSetup:
|
||||
"""UT-24: Qdrant-Index-Setup"""
|
||||
|
||||
def test_bool_index_method_exists(self):
|
||||
"""create_bool_index Methode existiert"""
|
||||
from scripts.setup_mindnet_collections import QdrantHTTP
|
||||
|
||||
q = QdrantHTTP("http://localhost:6333")
|
||||
assert hasattr(q, "create_bool_index")
|
||||
|
||||
def test_setup_includes_is_internal_index(self):
|
||||
"""Setup-Funktion enthält is_internal Index"""
|
||||
import inspect
|
||||
from scripts.setup_mindnet_collections import setup_mindnet_collections
|
||||
|
||||
# Prüfe den Quellcode der Funktion
|
||||
source = inspect.getsource(setup_mindnet_collections)
|
||||
|
||||
assert "is_internal" in source
|
||||
assert "create_bool_index" in source
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
pytest.main([__file__, "-v"])
|
||||
Reference in new issue
Block a user