Motivation

• Make the memory health check container-aware by reading cgroup limits/usage so reported totals reflect container limits when running in Docker/K8s.
• Prefer cgroup v2 (/sys/fs/cgroup/memory.max + /sys/fs/cgroup/memory.current) and fall back to cgroup v1 or host metrics when limits are unlimited or unavailable.

Description

• Added cgroup-aware logic in packages/healthcheck/src/checks/memory.ts that reads v2 files and two common v1 path variants and returns a MemorySnapshot when a finite container limit is detected.
• Treat memory.max === "max" (v2) and very large v1 limits (heuristic 2 ** 60) as unlimited and fall back to host metrics from systeminformation.mem() while keeping the existing free-ratio threshold checks unchanged.
• Added metrics.source ("cgroup" | "host") and preserved the existing API surface and free/degraded/unhealthy logic; all cgroup reads are guarded to avoid crashes when files are missing.
• Added a lightweight verification script packages/healthcheck/scripts/print-memory-source.mjs and a memory:inspect script entry in packages/healthcheck/package.json to print computed source and values.

Testing

• Ran pnpm --filter @nexload-sdk/healthcheck build which completed successfully.
• Ran pnpm --filter @nexload-sdk/healthcheck memory:inspect which executed the new script and printed the computed metrics.source and memory values successfully in this environment.
• Ran pnpm --filter @nexload-sdk/healthcheck lint which failed in this environment due to a workspace @nexload-sdk/eslint-config export issue (ERR_PACKAGE_PATH_NOT_EXPORTED), not related to the changes in the healthcheck logic.

Codex Task