fix: resolve ruff lint errors in benchmark suite

Remove unused imports (shutil, string, datetime, os, yaml, time,
SCALE_CONFIGS) and unused variable assignments in timing-only calls.

tests/benchmarks/conftest.py

@@ -2,7 +2,6 @@
 
 import json
 import os
-import shutil
 import tempfile
 
 import pytest

tests/benchmarks/data_generator.py

@@ -10,7 +10,6 @@ Planted "needle" drawers enable recall measurement without an LLM judge.
 import hashlib
 import os
 import random
-import string
 from datetime import datetime, timedelta
 from pathlib import Path
 

tests/benchmarks/report.py

@@ -8,7 +8,6 @@ conftest.py pytest_terminal_summary hook writes the collected results.
 import json
 import os
 import tempfile
-from datetime import datetime
 
 
 RESULTS_FILE = os.path.join(tempfile.gettempdir(), "mempalace_bench_results.json")

tests/benchmarks/test_ingest_bench.py

@@ -8,12 +8,10 @@ Measures mining performance at scale:
   - Re-ingest skip overhead (finding #11: file_already_mined check)
 """
 
-import os
 import time
 
 import chromadb
 import pytest
-import yaml
 
 from tests.benchmarks.data_generator import PalaceDataGenerator
 from tests.benchmarks.report import record_metric

tests/benchmarks/test_knowledge_graph_bench.py

@@ -56,7 +56,6 @@ class TestQueryEntityLatency:
         from mempalace.knowledge_graph import KnowledgeGraph
 
         kg = KnowledgeGraph(db_path=str(tmp_path / "kg.sqlite3"))
-        gen = PalaceDataGenerator(seed=42)
 
         # Create a hub entity connected to many others
         kg.add_entity("Hub", "person")
@@ -72,7 +71,7 @@ class TestQueryEntityLatency:
         latencies = []
         for _ in range(20):
             start = time.perf_counter()
-            result = kg.query_entity("Hub")
+            kg.query_entity("Hub")
             elapsed_ms = (time.perf_counter() - start) * 1000
             latencies.append(elapsed_ms)
 
@@ -106,7 +105,7 @@ class TestTimelinePerformance:
         latencies = []
         for _ in range(10):
             start = time.perf_counter()
-            result = kg.timeline()
+            kg.timeline()
             elapsed_ms = (time.perf_counter() - start) * 1000
             latencies.append(elapsed_ms)
 
@@ -143,8 +142,6 @@ class TestTemporalQueryAccuracy:
 
         # Query Alice as of March 2024 — should find ProjectA
         result_march = kg.query_entity("Alice", as_of="2024-03-15")
-        project_names = [r.get("object") or r.get("name", "") for r in result_march] if isinstance(result_march, list) else []
-
         # Query Alice as of September 2024 — should find ProjectB
         result_sept = kg.query_entity("Alice", as_of="2024-09-15")
 
@@ -161,7 +158,6 @@ class TestSQLiteConcurrentAccess:
         from mempalace.knowledge_graph import KnowledgeGraph
 
         kg = KnowledgeGraph(db_path=str(tmp_path / "kg.sqlite3"))
-        gen = PalaceDataGenerator(seed=42)
 
         # Pre-create entities
         for i in range(100):
@@ -276,7 +272,7 @@ class TestKGStats:
         latencies = []
         for _ in range(10):
             start = time.perf_counter()
-            result = kg.stats()
+            kg.stats()
             elapsed_ms = (time.perf_counter() - start) * 1000
             latencies.append(elapsed_ms)
 

tests/benchmarks/test_layers_bench.py

@@ -5,7 +5,6 @@ Tests MemoryStack.wake_up(), Layer1.generate(), and Layer2/L3
 at scale. Layer1 has the same unbounded col.get() as tool_status.
 """
 
-import os
 import time
 
 import pytest
@@ -168,7 +167,7 @@ class TestLayer2Retrieval:
         latencies = []
         for _ in range(10):
             start = time.perf_counter()
-            text = layer.retrieve(wing=wing, n_results=10)
+            layer.retrieve(wing=wing, n_results=10)
             elapsed_ms = (time.perf_counter() - start) * 1000
             latencies.append(elapsed_ms)
 
@@ -198,7 +197,7 @@ class TestLayer3Search:
         latencies = []
         for q in queries:
             start = time.perf_counter()
-            text = stack.search(q, n_results=5)
+            stack.search(q, n_results=5)
             elapsed_ms = (time.perf_counter() - start) * 1000
             latencies.append(elapsed_ms)
 

tests/benchmarks/test_mcp_bench.py

@@ -14,7 +14,7 @@ import time
 import chromadb
 import pytest
 
-from tests.benchmarks.data_generator import PalaceDataGenerator, SCALE_CONFIGS
+from tests.benchmarks.data_generator import PalaceDataGenerator
 from tests.benchmarks.report import record_metric
 
 

tests/benchmarks/test_memory_profile.py

@@ -8,7 +8,6 @@ Targets the highest-risk code paths:
   - Layer1.generate() (fetches all drawers)
 """
 
-import time
 import tracemalloc
 
 import pytest

tests/benchmarks/test_recall_threshold.py

@@ -8,7 +8,6 @@ wing+room to find the actual embedding model limit.
 
 import hashlib
 import os
-import time
 from datetime import datetime
 
 import chromadb

tests/benchmarks/test_search_bench.py

@@ -216,7 +216,7 @@ class TestSearchNResultsScaling:
         latencies = []
         for _ in range(5):
             start = time.perf_counter()
-            result = search_memories(
+            search_memories(
                 "authentication middleware", palace_path=palace_path, n_results=n_results
             )
             latencies.append((time.perf_counter() - start) * 1000)