MCPTap v2.5.3

[2.5.3]

Added

• SQLite log retention via background asyncio task — a new
  LogRetentionTask runs every hour inside the aiohttp event loop and
  deletes log entries older than MCP_TAP_LOG_RETENTION_DAYS (default 30).
  The task is wired into the application lifecycle (on_startup /
  on_cleanup) alongside the existing ConfigReloader, and gracefully
  survives transient purge errors (logs and continues).

  • settings.py — new field log_retention_days (env
    MCP_TAP_LOG_RETENTION_DAYS, default 30).
  • log_store.py — new method LogStore.purge_old(retention_days)
    that executes DELETE FROM request_logs WHERE timestamp < ? and
    returns the number of deleted rows.
  • log_retention.py — new module with LogRetentionTask class.
  • app.py — _start_log_retention / _stop_log_retention lifecycle
    hooks; skipped when log store is disabled.

  Tests: tests/test_log_store.py adds TestPurgeOld (4 tests — deletes
  old entries only, returns 0 when nothing to delete / when disabled,
  deletes all when retention is 0). tests/test_log_retention.py (5
  tests — purge_once, disabled store, start/stop lifecycle, purge in loop,
  error resilience).

Full Changelog

v2.5.2...v2.5.3