memory / scratchpad.py — Scratchpad

In-memory LRU + optional SQLite persistence with per-entry TTL (default 24 h).

Storage

OrderedDict in process memory (LRU).
Optional SQLite at persist_path/scratchpad.db with table scratchpad(key PK, value JSON, created_at REAL, accessed_at REAL).

API

Method	Purpose
`__init__(max_entries=50, persist_path=None, ttl=86400)`	LRU + TTL config.
`set(key, value) -> str`	Store; LRU evict if over capacity.
`get(key) -> Optional[Any]`	Retrieve; TTL check; refresh `accessed_at` (sliding-window TTL).
`list_all() / count() / delete(key) / clear()`	Standard helpers.

Eviction model

Age-based: any read computes cutoff = now − ttl; expired entries are deleted lazily.
Capacity-based: oldest LRU entry evicted when capacity exceeded.
Initial load on startup re-reads only non-expired entries.

threading.RLock guards in-memory dict + SQLite.