feat: QA 스위트 — Wikipedia 169건 + GitHub 368건 실제 데이터로 검증

실제 데이터:
- Wikipedia 한국어 기술 문서 169건 (16개 카테고리: 프로그래밍, DB, AI, 보안 등)
- GitHub hive-corp 커밋 168건 (한/영 혼합)
- GitHub vscode 이슈 200건 (영어, 라벨 포함)

QA 테스트 (24건):
- 검색 품질: 8개 쿼리별 relevance 검증, precision@5, resonance 정렬
- Hebbian 학습: reinforce 후 순위 상승, 실패 후 resonance 하락, co-activation edge 생성
- Consolidation: 3회 접근 후 L0→L1 승격 검증
- 성능: P95 < 100ms, batch 50+ nodes/sec, 캐시 효과
- 크로스소스: Wikipedia + GitHub 혼합 검색

버그 수정:
- graph.get() cache hit 시 access_count 미갱신 → consolidation 승격 실패
  → 캐시 히트에도 backend.update_node() 호출하도록 수정

Total: 145 unit+QA + 13 integration = 158 tests passed

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: SonAIengine

pyproject.toml

@@ -40,6 +40,10 @@ packages = ["src/synaptic"]
 [tool.pytest.ini_options]
 asyncio_mode = "auto"
 testpaths = ["tests"]
+markers = [
+    "integration: PostgreSQL integration tests (require running PG server)",
+    "qa: Quality assurance tests with real external data",
+]
 
 [tool.ruff]
 target-version = "py312"

src/synaptic/graph.py

@@ -99,6 +99,10 @@ async def search(
     async def get(self, node_id: str) -> Node | None:
         cached = self._cache.get(node_id)
         if cached is not None:
+            # Still track access in backend for consolidation
+            cached.access_count += 1
+            cached.updated_at = time()
+            await self._backend.update_node(cached)
             return cached
         node = await self._store.get_node(node_id)
         if node is not None:

tests/qa/__init__.py

@@ -0,0 +1 @@
+"""QA tests — real data validation for search quality, learning, and performance."""

tests/qa/conftest.py

@@ -0,0 +1,169 @@
+"""QA test fixtures — real data ingestion + graph setup."""
+
+from __future__ import annotations
+
+import json
+from collections.abc import AsyncGenerator
+from pathlib import Path
+
+import pytest
+
+from synaptic.backends.memory import MemoryBackend
+from synaptic.extensions.tagger_regex import RegexTagExtractor
+from synaptic.graph import SynapticGraph
+from synaptic.models import NodeKind
+
+DATA_DIR = Path(__file__).parent / "data"
+
+
+def _load_json(name: str) -> list[dict[str, object]]:
+    path = DATA_DIR / name
+    if not path.exists():
+        pytest.skip(f"Data file not found: {path}")
+    with open(path) as f:
+        data = json.load(f)
+    if not data:
+        pytest.skip(f"Data file is empty: {path}")
+    return data  # type: ignore[no-any-return]
+
+
+@pytest.fixture
+async def wiki_graph() -> AsyncGenerator[SynapticGraph]:
+    """Graph populated with Korean Wikipedia tech articles."""
+    articles = _load_json("wikipedia_ko_tech.json")
+
+    backend = MemoryBackend()
+    await backend.connect()
+    tagger = RegexTagExtractor()
+    graph = SynapticGraph(backend, tag_extractor=tagger)
+
+    for article in articles:
+        title = str(article.get("title", ""))
+        content = str(article.get("content", ""))
+        if not title or not content:
+            continue
+        # Truncate very long articles for performance
+        if len(content) > 2000:
+            content = content[:2000]
+        cats = article.get("categories", [])
+        tags = list(cats) if isinstance(cats, list) else []
+        await graph.add(
+            title=title,
+            content=content,
+            kind=NodeKind.CONCEPT,
+            tags=[str(t) for t in tags],
+            source="wikipedia:ko",
+        )
+
+    yield graph
+    await backend.close()
+
+
+@pytest.fixture
+async def github_graph() -> AsyncGenerator[SynapticGraph]:
+    """Graph populated with GitHub commits + issues."""
+    backend = MemoryBackend()
+    await backend.connect()
+    tagger = RegexTagExtractor()
+    graph = SynapticGraph(backend, tag_extractor=tagger)
+
+    # Ingest commits
+    commits = _load_json("github_commits.json")
+    for commit in commits:
+        msg = str(commit.get("message", ""))
+        if not msg or len(msg) < 10:
+            continue
+        first_line = msg.split("\n", maxsplit=1)[0]
+        await graph.add(
+            title=first_line[:100],
+            content=msg,
+            kind=NodeKind.ARTIFACT,
+            tags=["commit"],
+            source="github:commit",
+        )
+
+    # Ingest issues
+    try:
+        issues = _load_json("github_issues.json")
+    except Exception:
+        issues = []
+
+    for issue in issues:
+        title = str(issue.get("title", ""))
+        body = str(issue.get("body", "") or "")
+        if not title:
+            continue
+        labels = issue.get("labels", [])
+        tag_list = ["issue"]
+        if isinstance(labels, list):
+            tag_list.extend(str(lb) for lb in labels[:5])
+        content = body[:1500] if body else title
+        await graph.add(
+            title=title[:100],
+            content=content,
+            kind=NodeKind.ENTITY,
+            tags=tag_list,
+            source="github:issue",
+        )
+
+    yield graph
+    await backend.close()
+
+
+@pytest.fixture
+async def combined_graph() -> AsyncGenerator[SynapticGraph]:
+    """Graph with both Wikipedia + GitHub data combined."""
+    backend = MemoryBackend()
+    await backend.connect()
+    tagger = RegexTagExtractor()
+    graph = SynapticGraph(backend, tag_extractor=tagger)
+
+    # Wikipedia
+    try:
+        articles = _load_json("wikipedia_ko_tech.json")
+        for article in articles[:50]:  # Limit for performance
+            title = str(article.get("title", ""))
+            content = str(article.get("content", ""))[:1500]
+            if title and content:
+                await graph.add(
+                    title=title,
+                    content=content,
+                    kind=NodeKind.CONCEPT,
+                    source="wikipedia:ko",
+                )
+    except Exception:  # noqa: S110
+        pass
+
+    # GitHub commits
+    try:
+        commits = _load_json("github_commits.json")
+        for commit in commits[:50]:
+            msg = str(commit.get("message", ""))
+            if msg and len(msg) >= 10:
+                await graph.add(
+                    title=msg.split("\n", maxsplit=1)[0][:100],
+                    content=msg,
+                    kind=NodeKind.ARTIFACT,
+                    source="github:commit",
+                )
+    except Exception:  # noqa: S110
+        pass
+
+    # GitHub issues
+    try:
+        issues = _load_json("github_issues.json")
+        for issue in issues[:50]:
+            title = str(issue.get("title", ""))
+            body = str(issue.get("body", "") or "")[:1500]
+            if title:
+                await graph.add(
+                    title=title[:100],
+                    content=body or title,
+                    kind=NodeKind.ENTITY,
+                    source="github:issue",
+                )
+    except Exception:  # noqa: S110
+        pass
+
+    yield graph
+    await backend.close()