hardwareSummary

A Swiss Army Knife library for system, hardware, and filesystem utilities. Reports OS / CPU / GPU / memory / disk / network, runs a CPU benchmark, organizes a downloads folder, indexes a filesystem to SQLite, generates a Mermaid gitGraph, and ships a streaming upload server plus a custom TCP file-transfer client/server. Python 3.6+, single runtime dependency (tabulate), heavy deps lazy-installed on first use.

pip install nodespecs
python -m specs                      # compact system report
python -m specs bcpu                 # CPU benchmark
python -c "import specs; specs.info_gpu()"

install with UV

curl -LsSf https://astral.sh/uv/install.sh | sh
source $HOME/.local/bin/env
uv python install 3.12

# On modern Linux (Ubuntu 22.04+, Debian 12+) the system Python is
# marked as "externally managed" by PEP 668. If a bare `pip install`
# fails with `externally-managed-environment`, either add
# `--break-system-packages` or use a virtualenv:
sudo apt install python-is-python3
sudo apt install python3-pip
pip install nodespecs --break-system-packages

uv init
uv add nodespecs
# v0.4.2+ auto-installs heavy deps (psutil, etc.) on first use via
# the `@ensure_lib` decorator — no separate `uv pip install psutil`.
uv run -m specs
uv run -m specs bcpu

# or
source .venv/bin/activate
python -m specs bcpu

install and use with pip

sudo apt install python-is-python3
sudo apt install python3-pip

Note on PEP 668. On modern Linux distros (Ubuntu 22.04+, Debian 12+, etc.), the system Python is marked as "externally managed." If pip install nodespecs fails with externally-managed-environment, add --break-system-packages or use a virtual environment (python3 -m venv .venv && source .venv/bin/activate).

v0.4.2+ no longer requires a separate pip install psutil — the @ensure_lib decorator auto-installs missing heavy deps (psutil, etc.) on first use. You only need pip install nodespecs.

pip install nodespecs
python -m specs
python -c "import specs; specs.bench_cpu()"
python -c "import specs; specs.hardware.get_system_info()"

python -c "import specs; specs.info_gpu()"
python -c "from specs import disk;print(disk.resolve_user_path())"
python -c "from specs.disk import resolve_user_path;print(resolve_user_path())"

# clean and orgnized download folder for WINDOWS user
python -m specs fclean

# for the machine not compatible with psutil
python -m specs -v
python -m specs -l=1

python -m specs bcpu
python -m specs upload
python -m specs server
python -m specs cld -i='172.25.1.228'

File Indexer Script

python -m specs.disk.drive walks a directory tree, captures per-file metadata (logical size, on-disk usage in 512-byte blocks, hardlinks, inode, device), and writes the result to a SQLite database. It also prints a side-by-side comparison between the indexer's on_disk_size and shutil.disk_usage(root_dir) so the gap from filesystem metadata, snapshots, and unread subtrees is visible.

The new on-disk size is deduped by inode, so a file with N hardlinks contributes its real allocation once instead of N times. Cluster rounding on NTFS / exFAT and sparse files are automatically accounted for via st_blocks * 512.

Usage

Run from the repository root. --dir defaults to the current working directory on Linux / macOS and to F:\\ on Windows:

python -m specs.disk.drive

To index a custom directory, pass --dir explicitly:

python -m specs.disk.drive --dir "F:\\"            # Windows
python -m specs.disk.drive --dir /home/me/data    # Linux / macOS

To set the SQLite output path explicitly:

python -m specs.disk.drive --dir "F:\\" --db "F:\\file_index.db"

Expected Output

On success:

Indexed <N> files (<U> unique inodes)
  logical size:         <bytes> bytes
  on-disk size:         <bytes> bytes  (deduped by inode; accounts for cluster rounding + sparse files)
  hardlink paths:       <count> (extra paths sharing an inode)
  skipped:              <count> (permission / access errors during walk)
Successfully exported to SQLite database: <path>/file_index.db

OS-reported usage of the volume containing <root_dir> (indexer covers this subtree only):
  total:                <bytes> bytes
  used:                 <bytes> bytes
  free:                 <bytes> bytes
  OS - indexer:         <bytes> bytes  (includes every file outside the indexed subtree, plus metadata, snapshots, etc.)

When the indexed directory is the volume root (e.g. C:\\ on Windows or / on Linux), the OS-reported block reports the volume itself and the gap is attributed to "metadata, system files, unread subtrees, snapshots" instead of "files outside the indexed subtree."

On failure to export:

Export failed: <error message>

The SQLite files table has columns for filepath, filename, extension, size, on_disk, blocks, nlinks, inode, device, last_modified, and indexed_at. The statistics table holds the same counters that the CLI prints.

## server
python -c "import specs; print(specs.__version__)"

python -c "import specs; specs.whoish()"
python -c "import specs; specs.server()"
## client upload wt progress bar
python -c "import specs; specs.client('172.25.1.175', 12345,'./README.md',False)"

## client upload wo progress bar
python -c "import specs; specs.client('172.25.1.175', 12345,'./README.md',False)"
## client upload top level files in the cwd
python -c "import specs; specs.clientd('192.168.0.157')"
python -c "import specs; specs.clientd('192.168.0.157', 12345)"

sudo apt update
sudo apt install python3-pip
python3 -m pip install nodespecs && python3 -m specs

import specs

specs.info_gpu()
specs.bench_cpu()

Deprecated (install and use with git)

!git clone https://github.com/jinsanity07git/hardwareSummary && python hardwareSummary/hardware.py && python hardwareSummary/cpu-benchmark.py

git clone https://github.com/jinsanity07git/hardwareSummary && python hardwareSummary/hardware.py && python hardwareSummary/cpu-benchmark.py

sudo apt upgrade
sudo apt install python3-pip
git clone https://github.com/jinsanity07git/hardwareSummary && python3 hardwareSummary/hardware.py && python3 hardwareSummary/cpu-benchmark.py

CPU collection

• Note
  • Kinds of Arch explanation⁴

GPU collection

Feature in develop

1. work through ws, transfer file using websocket
2. py-ios-device python based Apple instruments protocol，you can get CPU, Memory and

• Streaming upload server in Python extended from uploadserver.other metrics from real iOS devices

Performance source

• https://browser.geekbench.com/ *
• https://technical.city/en/video/GeForce-RTX-4060-vs-Tesla-T4
• https://technical.city/en/video/Tesla-T4-vs-T600