神经风暴图 ⚡https://github.com/lichuang9890-star/neural-storm-graph

核心特性（Features）

目录结构 (Structure)

<span style="background-color:#151b23"><span style="color:#f0f6fc"><span style="color:#f0f6fc"><span style="background-color:#151b23"><code>neural-storm-graph/
├── core/
│   ├── __init__.py          # 包入口
│   ├── storm_skeleton.py    # 风暴骨架引擎（调度、推理、骨架扫描）
│   └── realtime_graph.py    # 实时网络流与关系节点提取算法
├── test_storm.py            # 完整测试套件 (45 tests)
├── requirements.txt
└── README.md
</code></span></span></span></span>

功能矩阵（特征矩阵）

Storm Skeleton — 骨架引擎

实时图谱 — 实时图谱 + 定位

快速入门

from core import StormSkeleton, RealtimeGraph
from core.realtime_graph import CursorPosition

# 1. 骨架扫描
skeleton = StormSkeleton("/path/to/your/project")
report = skeleton.skeleton_scan()
print(f"Scanned {report['files_scanned']} files, "
      f"{report['total_symbols']} symbols in {report['scan_time_ms']}ms")

# 2. 风暴中心定位 — 找到项目中最相关的核心节点
center = skeleton.storm_center("修复限流中间件")
for node in center["top_nodes"][:5]:
    print(f"  {node['fqn']} (score: {node['total']})")

# 3. DAG 依赖推理
dag = skeleton.decompose("重构用户服务")
print(f"Decomposed into {dag['total_nodes']} nodes, cycles: {dag['cycles']}")

# 4. 实时图谱 + 光标定位
graph = RealtimeGraph(skeleton)
ctx = graph.locate(CursorPosition("src/middleware.py", 42))
print(f"Symbol: {ctx.symbol.name if ctx.symbol else 'N/A'}")
print(f"Callers: {len(ctx.callers)}, Callees: {len(ctx.callees)}")
print(f"Breadcrumb: {' > '.join(ctx.breadcrumb)}")

# 5. 增量更新 + 事件监听
graph.on_change(lambda c: print(f"[CHANGE] {c.file}: +{len(c.added)} -{len(c.removed)}"))
change = graph.update_file("src/middleware.py")

# 6. 热力图
for entry in graph.heatmap()[:5]:
    print(f"  {entry['file']}: heat={entry['heat']}")

# 7. 调用链追踪
chain = graph.trace_call_chain("src/main.py::App.run", direction="both")

# 8. 导出 Mermaid 图
print(graph.export_mermaid(rel_path="src/main.py"))

零依赖（零依赖）

测试（测试）

python -m pytest test_storm.py -v   # 45/45 tests, < 1s

设计哲学（Philosophy）