How does VelocityMind warm application binaries into the page cache safely?

[amandeavor]

What system calls and memory pressure safeguards does velocitymind use to preload desktop binaries without causing OOM thrashing on low-RAM systems?

[amandeavor]

VelocityMind Memory Safety & Page-Cache Warming

velocitymind acts as a lightweight heuristic preloader daemon. Its architecture prioritizes memory conservation:

1. Asynchronous Non-Blocking Preloading: It uses posix_fadvise(fd, 0, 0, POSIX_FADV_WILLNEED) rather than mmap locking, allowing the Linux kernel VFS to read pages into clean page cache without pinning them.
2. Transition Probability Graph: Records application focus transitions in a local SQLite database (~/.local/share/velocitymind/transitions.db). Only binaries with transition confidence >0.75 are queued.
3. PSI (Pressure Stall Information) Safeguard: Before issuing posix_fadvise, the daemon inspects /proc/pressure/memory. If memory pressure exceeds 15% in the last 10-second window, preloading is throttled or aborted to prevent OOM stalls.
4. ZRAM / Swappiness Harmony: Preloaded pages remain clean and reclaimable by the kernel immediately if memory demand spikes.