Lightweight end-to-end process monitoring for Linux. Unlike top, time, or single-metric profilers, viy captures CPU, memory, I/O, network, threads, and file descriptors together with time-series visualization and performance insights.
It correlates resource spikes with Python function calls, helping to see root causes and understand program behavior.
Named after Viy (Вій), a creature inspired by Ukrainian folklore known for its powerful gaze that sees through everything.
# Download binary
curl -L https://github.com/enkidulan/viy/releases/latest/download/viy -o ~/.local/bin/viy
chmod +x ~/.local/bin/viy$ viy python tests/test_workload.py╭─ viy ────────────────────────────────────────────────────────────────────────╮
│ Command: python tests/test_workload.py │
│ Exit code: 0 │
│ Wall time: 4.72s │
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ Summary ────────────────────────────────────────────────────────────────────╮
│ CPU: 71.6% (user 3.12s, sys 260.0ms) Memory: 9.0MB -> 11.8MB (peak 111.6MB)│
│ I/O: 477.1MB read, 477.3MB written Network: 148B in, 0B out (host)│
│ Ctx switch: 52 vol, 118 invol Page faults: 26.7K minor, 0 major │
│ Threads: 1 -> 1 (peak 6) File descs: 23 (peak 24) │
│ Processes: 5 spawned │
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ CPU Utilization (%) avg 71.6% ─────────────────────────────────────────────╮
│ 116% ▄ █ ▆ ▆ ▇ ▇▅▆ ▆ ▆ ▇ ▇ █ ▇ ▇ █ ▇ ▇▇█▅▇ ▇ ▇▆█ ▆ ▅ │
│ █ █ █ █ █ ███ █ █ █ █ █ █ █ █ █ █████ █ ███ █ █ │
│ 58% █▃█▄█▃█▃█▄███▃█▃█▂█▄█▄█▄█▃█▄█▄█▂█████▃█▄███▃ ▂ ▁ ▂ ▂▂█▄█ │
│ ████████████████████████████████████████████ █ █ █ █████ │
│ 0% ████████████████████████████████████████████ █ █ █ █████ │
│ 0 1.2s 2.4s 3.5s 4.7 │
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ Memory RSS peak 111.6MB ───────────────────────────────────────────────────╮
│ 111.6MB ▁▆▇███ │
│ ▄▅███████ │
│ 55.8MB ▁▂▆▇█████████ │
│ ▃▄█████████████ │
│ 0B ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▆▇███████████████▄▄▄▄▄ │
│ 0 1.2s 2.4s 3.5s 4.7 │
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ I/O Rate 477.1MB read, 477.3MB written ────────────────────────────────────╮
│ 6.6GB/s █ │
│ █ │
│ 3.3GB/s █ │
│ ▄▆▅█ │
│ 0B/s ▅████ │
│ 0 1.2s 2.4s 3.5s 4.7 │
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ Python Timeline → 4.72s ──────────────────────────────────────────────────╮
│ cpu_burn (__main__) [CPU] 163% │
│ ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ │
│ memory_grow_and_shrink (__main__) [MEM] 111.6MB │
│ ░░░░▒▒▒▒▓▓▓▓████▒ │
│ io_write_read (__main__) [I/O] 6.6GB/s │
│ ░░░▒▓ │
│ spawn_children (__main__) [CPU] 682% │
│ ▒ │
│ 0 1.2s 2.4s 3.5s 4.7│
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ Insights ───────────────────────────────────────────────────────────────────╮
│ [THR] Dynamic thread creation detected (peak 6 threads from initial 1) │
│ [MEM] Peak RSS 111.6MB at 91% through execution, then declined │
│ [PY] CPU peak (682%) → spawn_children (__main__) │
│ [PY] Memory peak (111.6MB) → memory_grow_and_shrink (__main__) │
│ [PY] I/O peak (6.6GB/s) → io_write_read (__main__) │
╰──────────────────────────────────────────────────────────────────────────────╯
viy [OPTIONS] <command> [args...]--json— Output JSON to stderr--quiet— Hide insights--silent— Hide entire report (use with--json)--interval <ms>— Sampling interval (default: 10)--no-color— Disable colors--py-filter <glob>— Trace Python frames matching pattern--py-top <n>— Show top N Python functions (default: 10)
# Monitor a build
viy make -j8
# JSON output for CI
viy --json --silent ./benchmark 2> report.json
# Python profiling
viy --py-filter "*train.py" python train.py
# Custom sampling
viy --interval 100 ./long-jobReport goes to stderr (stdout is unmodified):
- Header — command, exit code, wall time
- Summary — CPU, memory, I/O, network, threads, file descriptors
- Charts — time-series for CPU, memory, I/O, network, threads
- Python Timeline — function-level profiling (when tracing Python)
- Insights — automated analysis (e.g., "CPU-bound", "Memory leak")
Use --json for structured output to stderr.
- CPU — user/system time, utilization %
- Memory — RSS, VSZ, page faults
- I/O — bytes read/written, syscalls, cache hit ratio
- Network — bytes sent/received (host namespace)
- Threads — count over time
- Context switches — voluntary/involuntary
- File descriptors — count over time
- Process tree — aggregates all child processes
Auto-generated performance analysis:
[CPU]— CPU-bound, I/O-bound, kernel-heavy, high contention[MEM]— Memory leaks, peak timing, page faults[I/O]— Cache hit ratio, small ops, throughput patterns[THR]— Dynamic thread creation[FD]— File descriptor leaks
Spawns your command and polls /proc/[pid]/{stat,io,status,fd} at 10ms intervals. Aggregates metrics across the full process tree. Overhead < 0.1% CPU.
- Linux (uses
/procfilesystem) - Rust 1.85+ (for building)
MIT