resource forecasting

A Bash-first playbook for AI/vector DB capacity on Linux: define SLOs, dims, datatypes, RF; size storage and ingest with shell functions; baseline with sysstat/iostat/iotop; choose NVMe/RAID/XFS and separate WAL/data; validate with fio (sequential WAL writes, random ANN reads); budget for replication, snapshots, and zstd backups; watch p95/p99 and write amp; revisit after model/schema changes.