Co Occurrence Engine

独立的 Hebbian 共现图引擎，记录记忆共现，计算关联权重，支持关联查询和数据库维护，服务星型记忆架构。

Security Scan

The skill's code, instructions, and resource use match its stated purpose (a local co‑occurrence tracker using a SQLite DB) and do not request unrelated credentials or network access.

Purpose & Capability

Name/description (Hebbian co-occurrence engine) matches the included implementation: a Python tracker that records co-occurrence edges in a local SQLite DB under ~/.config/cortexgraph. No unrelated services, credentials, or binaries are requested.

Instruction Scope

SKILL.md and the Python file confine operations to local DB read/write, queries, decay, and maintenance. Instructions reference only the DB path and optional half-life setting; they do not instruct reading unrelated system files or sending data externally.

Install Mechanism

No install spec or remote downloads are present. The skill is instruction-only with a bundled Python script, so nothing is fetched from external URLs during install.

Credentials

No required environment variables or credentials are declared. SKILL.md documents optional env vars (CO_OCCURRENCE_DB_PATH, CO_OCCURRENCE_HALF_LIFE_DAYS) that are reasonable for configuration. The code writes only to a per-user config path.

Persistence & Privilege

always is false and the skill does not request elevated privileges or modify other skills. It creates/uses a local DB file under the user's config directory (normal for this kind of tool).

Assessment

This skill appears coherent and local-only: it creates and updates a SQLite DB at ~/.config/cortexgraph/co_occurrence.db and does not contact external endpoints or request secrets. Before installing, consider: (1) any memory IDs you pass to the tracker will be stored on disk — avoid feeding sensitive secrets if you don't want them persisted; (2) the code optionally imports a third-party 'forgetting_curve' module if installed — only install/allow that package from a trusted source; (3) confirm the DB path is acceptable or override CO_OCCURRENCE_DB_PATH if you prefer a different location; (4) if you integrate this with other components (adapter_factory or external adapters), review those adapter implementations to ensure they don't leak data. Overall the skill is internally consistent and low-risk, but treat stored memory contents as persistent data you control.

独立的 Hebbian 共现图引擎，为星型记忆架构提供记忆关联发现与权重计算服务。

功能

• 共现记录：自动记录同时被检索的记忆对，建立关联边
• 关联查询：查询与指定记忆最相关的其他记忆
• 权重计算：基于使用频率与时间衰减计算关联强度
• 统计信息：提供边数、唯一记忆数、平均权重等统计
• 维护工具：清理过旧的边，保持数据库健康

架构

┌─────────────────────────────────────────────────────┐
│               共现图引擎 (独立插件)                  │
│                                                     │
│  • CoOccurrenceEngine 类                            │
│  • SQLite 数据库 (共现图)                            │
│  • 遗忘曲线集成 (可选)                               │
└─────────────────────────────────────────────────────┘
                               │
                               ▼
┌─────────────────────────────────────────────────────┐
│               适配器 (co_occurrence_adapter)         │
│                                                     │
│  • 实现统一记忆接口                                   │
│  • 注册到星型架构适配器框架                           │
└─────────────────────────────────────────────────────┘

核心类

CoOccurrenceEngine

主引擎类，提供共现图的所有操作。

engine = CoOccurrenceEngine(db_path="~/.config/cortexgraph/co_occurrence.db")

主要方法：

• record_co_occurrence(memory_ids: List[str], context: str = "")
• get_co_occurrence_score(memory_id: str, related_ids: List[str] = None) -> float
• get_related_memories(memory_id: str, top_k: int = 10) -> List[Tuple[str, float]]
• get_stats() -> Dict
• decay_old_edges(days: int = 90) -> int

## 数据库

默认数据库位置：`~/.config/cortexgraph/co_occurrence.db`

表结构：
```sql
CREATE TABLE co_occurrence (
    memory_a TEXT,
    memory_b TEXT,
    weight REAL,
    last_updated TEXT,
    created_at TEXT,
    PRIMARY KEY (memory_a, memory_b)
)

集成

1. 独立使用

from scripts.co_occurrence_tracker import CoOccurrenceEngine

engine = CoOccurrenceEngine()
engine.record_co_occurrence(["mem_001", "mem_002"])
related = engine.get_related_memories("mem_001")

2. 通过适配器集成

from adapter_factory import AdapterFactory

factory = AdapterFactory()
adapter = factory.get_adapter("co_occurrence")
results = adapter.search("mem_001", max_results=10)

配置

通过环境变量或配置文件：

• CO_OCCURRENCE_DB_PATH：数据库路径（默认：~/.config/cortexgraph/co_occurrence.db）
• CO_OCCURRENCE_HALF_LIFE_DAYS：衰减半衰期（默认：30天）

依赖

• Python 3.8+
• SQLite3（内置）
• 可选：forgetting-curve 插件（提供更精确的衰减计算）

版本历史

• v0.1.0（初始版本）：从 memory-sync-enhanced 中提取的共现图引擎