Metadata-Version: 2.4
Name: gpu-usage-audit
Version: 1.7.2
Summary: Single-host daemon that surfaces 'idle-held' NVIDIA GPU memory — the embarrassing category conventional dashboards miss.
Project-URL: Homepage, https://github.com/AI-Ocean/gpu-usage-audit
Project-URL: Issues, https://github.com/AI-Ocean/gpu-usage-audit/issues
Project-URL: Releases, https://github.com/AI-Ocean/gpu-usage-audit/releases
Author: AI-Ocean
License:                                  Apache License
                                   Version 2.0, January 2004
                                http://www.apache.org/licenses/
        
           TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
        
           1. Definitions.
        
              "License" shall mean the terms and conditions for use, reproduction,
              and distribution as defined by Sections 1 through 9 of this document.
        
              "Licensor" shall mean the copyright owner or entity authorized by
              the copyright owner that is granting the License.
        
              "Legal Entity" shall mean the union of the acting entity and all
              other entities that control, are controlled by, or are under common
              control with that entity. For the purposes of this definition,
              "control" means (i) the power, direct or indirect, to cause the
              direction or management of such entity, whether by contract or
              otherwise, or (ii) ownership of fifty percent (50%) or more of the
              outstanding shares, or (iii) beneficial ownership of such entity.
        
              "You" (or "Your") shall mean an individual or Legal Entity
              exercising permissions granted by this License.
        
              "Source" form shall mean the preferred form for making modifications,
              including but not limited to software source code, documentation
              source, and configuration files.
        
              "Object" form shall mean any form resulting from mechanical
              transformation or translation of a Source form, including but
              not limited to compiled object code, generated documentation,
              and conversions to other media types.
        
              "Work" shall mean the work of authorship, whether in Source or
              Object form, made available under the License, as indicated by a
              copyright notice that is included in or attached to the work
              (an example is provided in the Appendix below).
        
              "Derivative Works" shall mean any work, whether in Source or Object
              form, that is based on (or derived from) the Work and for which the
              editorial revisions, annotations, elaborations, or other modifications
              represent, as a whole, an original work of authorship. For the purposes
              of this License, Derivative Works shall not include works that remain
              separable from, or merely link (or bind by name) to the interfaces of,
              the Work and Derivative Works thereof.
        
              "Contribution" shall mean any work of authorship, including
              the original version of the Work and any modifications or additions
              to that Work or Derivative Works thereof, that is intentionally
              submitted to Licensor for inclusion in the Work by the copyright owner
              or by an individual or Legal Entity authorized to submit on behalf of
              the copyright owner. For the purposes of this definition, "submitted"
              means any form of electronic, verbal, or written communication sent
              to the Licensor or its representatives, including but not limited to
              communication on electronic mailing lists, source code control systems,
              and issue tracking systems that are managed by, or on behalf of, the
              Licensor for the purpose of discussing and improving the Work, but
              excluding communication that is conspicuously marked or otherwise
              designated in writing by the copyright owner as "Not a Contribution."
        
              "Contributor" shall mean Licensor and any individual or Legal Entity
              on behalf of whom a Contribution has been received by Licensor and
              subsequently incorporated within the Work.
        
           2. Grant of Copyright License. Subject to the terms and conditions of
              this License, each Contributor hereby grants to You a perpetual,
              worldwide, non-exclusive, no-charge, royalty-free, irrevocable
              copyright license to reproduce, prepare Derivative Works of,
              publicly display, publicly perform, sublicense, and distribute the
              Work and such Derivative Works in Source or Object form.
        
           3. Grant of Patent License. Subject to the terms and conditions of
              this License, each Contributor hereby grants to You a perpetual,
              worldwide, non-exclusive, no-charge, royalty-free, irrevocable
              (except as stated in this section) patent license to make, have made,
              use, offer to sell, sell, import, and otherwise transfer the Work,
              where such license applies only to those patent claims licensable
              by such Contributor that are necessarily infringed by their
              Contribution(s) alone or by combination of their Contribution(s)
              with the Work to which such Contribution(s) was submitted. If You
              institute patent litigation against any entity (including a
              cross-claim or counterclaim in a lawsuit) alleging that the Work
              or a Contribution incorporated within the Work constitutes direct
              or contributory patent infringement, then any patent licenses
              granted to You under this License for that Work shall terminate
              as of the date such litigation is filed.
        
           4. Redistribution. You may reproduce and distribute copies of the
              Work or Derivative Works thereof in any medium, with or without
              modifications, and in Source or Object form, provided that You
              meet the following conditions:
        
              (a) You must give any other recipients of the Work or
                  Derivative Works a copy of this License; and
        
              (b) You must cause any modified files to carry prominent notices
                  stating that You changed the files; and
        
              (c) You must retain, in the Source form of any Derivative Works
                  that You distribute, all copyright, patent, trademark, and
                  attribution notices from the Source form of the Work,
                  excluding those notices that do not pertain to any part of
                  the Derivative Works; and
        
              (d) If the Work includes a "NOTICE" text file as part of its
                  distribution, then any Derivative Works that You distribute must
                  include a readable copy of the attribution notices contained
                  within such NOTICE file, excluding those notices that do not
                  pertain to any part of the Derivative Works, in at least one
                  of the following places: within a NOTICE text file distributed
                  as part of the Derivative Works; within the Source form or
                  documentation, if provided along with the Derivative Works; or,
                  within a display generated by the Derivative Works, if and
                  wherever such third-party notices normally appear. The contents
                  of the NOTICE file are for informational purposes only and
                  do not modify the License. You may add Your own attribution
                  notices within Derivative Works that You distribute, alongside
                  or as an addendum to the NOTICE text from the Work, provided
                  that such additional attribution notices cannot be construed
                  as modifying the License.
        
              You may add Your own copyright statement to Your modifications and
              may provide additional or different license terms and conditions
              for use, reproduction, or distribution of Your modifications, or
              for any such Derivative Works as a whole, provided Your use,
              reproduction, and distribution of the Work otherwise complies with
              the conditions stated in this License.
        
           5. Submission of Contributions. Unless You explicitly state otherwise,
              any Contribution intentionally submitted for inclusion in the Work
              by You to the Licensor shall be under the terms and conditions of
              this License, without any additional terms or conditions.
              Notwithstanding the above, nothing herein shall supersede or modify
              the terms of any separate license agreement you may have executed
              with Licensor regarding such Contributions.
        
           6. Trademarks. This License does not grant permission to use the trade
              names, trademarks, service marks, or product names of the Licensor,
              except as required for describing the origin of the Work and
              reproducing the content of the NOTICE file.
        
           7. Disclaimer of Warranty. Unless required by applicable law or
              agreed to in writing, Licensor provides the Work (and each
              Contributor provides its Contributions) on an "AS IS" BASIS,
              WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
              implied, including, without limitation, any warranties or conditions
              of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
              PARTICULAR PURPOSE. You are solely responsible for determining the
              appropriateness of using or redistributing the Work and assume any
              risks associated with Your exercise of permissions under this License.
        
           8. Limitation of Liability. In no event and under no legal theory,
              whether in tort (including negligence), contract, or otherwise,
              unless required by applicable law (such as deliberate and grossly
              negligent acts) or agreed to in writing, shall any Contributor be
              liable to You for damages, including any direct, indirect, special,
              incidental, or consequential damages of any character arising as a
              result of this License or out of the use or inability to use the
              Work (including but not limited to damages for loss of goodwill,
              work stoppage, computer failure or malfunction, or any and all
              other commercial damages or losses), even if such Contributor
              has been advised of the possibility of such damages.
        
           9. Accepting Warranty or Support. While redistributing the Work or
              Derivative Works thereof, You may choose to offer, and charge a
              fee for, acceptance of support, warranty, indemnity, or other
              liability obligations and/or rights consistent with this License.
              However, in accepting such obligations, You may act only on Your
              own behalf and on Your sole responsibility, not on behalf of any
              other Contributor, and only if You agree to indemnify, defend,
              and hold each Contributor harmless for any liability incurred by,
              or claims asserted against, such Contributor by reason of your
              accepting any such warranty or support.
        
           END OF TERMS AND CONDITIONS
        
           APPENDIX: How to apply the Apache License to your work.
        
              To apply the Apache License to your work, attach the following
              boilerplate notice, with the fields enclosed by brackets "[]"
              replaced with your own identifying information. (Don't include
              the brackets!)  The text should be enclosed in the appropriate
              comment syntax for the file format. We also recommend that a
              file or class name and description of purpose be included on the
              same "printed page" as the copyright notice for easier
              identification within third-party archives.
        
           Copyright 2026 AIOcean
        
           Licensed under the Apache License, Version 2.0 (the "License");
           you may not use this file except in compliance with the License.
           You may obtain a copy of the License at
        
               http://www.apache.org/licenses/LICENSE-2.0
        
           Unless required by applicable law or agreed to in writing, software
           distributed under the License is distributed on an "AS IS" BASIS,
           WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
           implied, including, without limitation, any warranties or conditions
           of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
           PARTICULAR PURPOSE. See the License for the specific language governing
           permissions and limitations under the License.
License-File: LICENSE
Keywords: gpu,idle-detection,monitoring,nvidia,sqlite
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Operating System :: POSIX :: Linux
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: System :: Monitoring
Requires-Python: >=3.12
Requires-Dist: nvidia-ml-py>=12.535
Requires-Dist: websockets>=13
Provides-Extra: nvml
Description-Content-Type: text/markdown

# gpu-usage-audit

Single-host NVIDIA GPU usage audit for finding **idle-held** GPUs: cards that look idle by utilization, but are still held by a process through GPU memory.

[![PyPI](https://img.shields.io/pypi/v/gpu-usage-audit.svg)](https://pypi.org/project/gpu-usage-audit/)
[![Python 3.12+](https://img.shields.io/badge/python-3.12%2B-blue.svg)](https://pypi.org/project/gpu-usage-audit/)
[![License](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE)
[![GitHub Release](https://img.shields.io/github/v/release/AI-Ocean/gpu-usage-audit)](https://github.com/AI-Ocean/gpu-usage-audit/releases)

[English](README.md) · [한국어](README.ko.md) · [Releases](https://github.com/AI-Ocean/gpu-usage-audit/releases) · [Issues](https://github.com/AI-Ocean/gpu-usage-audit/issues)

---

![gua top — live local GPU view](https://raw.githubusercontent.com/AI-Ocean/gpu-usage-audit/main/docs/img/gua-top.png)

*`gua top` — a live local view, no board or web UI required. GPU0 is genuinely busy; GPU1 sits at 3% utilization while still holding 8.2 GB (**idle-held**); GPU2 is truly free.*

## About

gpu-usage-audit watches local NVIDIA/NVML telemetry and answers one question other dashboards skip: **is this GPU actually free, or just sitting idle while holding memory?** It separates every GPU card-tick into:

- `active`: utilization is doing real work
- `idle-held`: utilization is low, but a process still holds GPU memory
- `truly-idle`: no meaningful GPU process memory is present

The second category is the point. A notebook can sit at 1% SM utilization while keeping an 8 GB tensor allocated. Conventional dashboards usually flatten that into “idle”; this tool shows that the card is effectively unavailable.

Two ways to see it: **live** with `gua top` (1s utilization graph + per-GPU process table, right in your terminal), and **retrospective** with `gua report` (records into SQLite, then splits the window into the three states above with idle-capacity in GPU-hours). For a shared lab across several hosts, the optional [GUA Board](#cloud-sync-gua-board-optional) folds the same telemetry into one web view.

## Features

- Single-host, bare-metal NVIDIA GPU audit
- `gua doctor` readiness check for `/dev/nvidia*`, `nvidia-smi`, NVML, and DB path
- Background collector with `gua daemon`, `gua status`, and `gua stop`
- Live local view with `gua top` — 1s util graph + per-GPU process table, no board needed
- SQLite history database at `~/.gua/gua.db` by default
- Report sections for headline split, idle capacity, per-GPU state, top identities, and time-of-day heatmap
- Daemon interval metadata stored per run, so reports compute GPU-hours correctly across mixed 30s / 10s runs
- GPU-less `gua demo` command with deterministic fake telemetry
- Optional cloud sync to [GUA Board](#cloud-sync-gua-board-optional): the same 1s util stream, surfaced across many hosts in one web view
- No cluster runtime dependency; no Kubernetes, Slurm, Docker, or remote-node scan

## Installation

The recommended install path is PyPI via [uv](https://docs.astral.sh/uv/):

```sh
uv tool install gpu-usage-audit
```

Update or remove it with:

```sh
uv tool install gpu-usage-audit@latest   # picks up a just-published release (upgrade may miss it due to index cache)
uv tool uninstall gpu-usage-audit
```

Manual wheel downloads are available from GitHub Releases (swap in the [latest tag](https://github.com/AI-Ocean/gpu-usage-audit/releases)):

```sh
BASE="https://github.com/AI-Ocean/gpu-usage-audit/releases/download/v1.6.1"
WHEEL="gpu_usage_audit-1.6.1-py3-none-any.whl"

curl -fsSLO "$BASE/$WHEEL"
curl -fsSLO "$BASE/SHA256SUMS"
sha256sum -c SHA256SUMS --ignore-missing

uvx --from "./$WHEEL" gua doctor
```

## Quick Start

On an NVIDIA GPU host:

```sh
gua doctor
gua daemon --interval 30s
gua status
gua report --since 1h
gua stop
```

`gua doctor` is read-only. It does not need `sudo`; run it as the same user that will run the daemon.

For a live look without the daemon or a report, just run `gua top` (press `q` or Ctrl-C to quit):

```sh
gua top            # 1s util graph + per-GPU process table
gua top --fake     # try it on a machine with no GPU
```

Default local state lives under `~/.gua/`:

| Path | Purpose |
| --- | --- |
| `~/.gua/gua.db` | SQLite history database |
| `~/.gua/gua.pid` | background daemon PID file |
| `~/.gua/gua.log` | daemon stdout/stderr log |

The default DB is an appendable local history database. Later daemon runs append to it. If you pass a custom `--db PATH`, daemon still refuses an existing file to avoid mixing ad hoc runs by accident.

## Report Preview

```text
$ gua report --since 1h
gua — lab-a100 (bare, driver 560.35.05)  Window: 1:00:00

§1 Headline
  basis: one sample = one GPU card at one daemon tick
  rules: active >=10% util; idle-held <10% util with >100 MB process memory
  active       █   15.7%
  idle-held    ▒   45.1%
  truly-idle   ░   39.2%
  (51 samples)

§2 Idle capacity
  converted from card-ticks to GPU-hours using recorded daemon interval
  idle-held: ~0.31 GPU-hours, ~1.53 GPUs equivalently unavailable
  truly-idle: ~0.12 GPU-hours, ~1.00 GPUs equivalently free

§3 Per-GPU
§4 Top identities
§5 Time-of-day heatmap (UTC)
```

Reports can run while the daemon is writing; SQLite WAL mode handles concurrent reads. Reports also work after the daemon has stopped, as long as the DB file exists.

## Commands

| Command | Description |
| --- | --- |
| `gua doctor` | Check local NVIDIA/NVML readiness and DB path status |
| `gua daemon` | Start background collection on the local NVIDIA host (`--cloud` also streams to GUA Board) |
| `gua start` | Alias for `gua daemon` |
| `gua status` | Show whether the managed background collector is running |
| `gua stop` | Stop the managed background collector |
| `gua top` | Live local GPU view (1s util graph + processes), no board required |
| `gua report` | Render the retrospective report from SQLite |
| `gua demo` | Generate a fake local report without a GPU |
| `gua enroll` | Connect this host to a GUA Board workspace (optional cloud sync) |
| `gua sync-once` | Collect one snapshot and push the latest state to GUA Board |
| `gua version` | Print version |

## Important Options

```sh
gua daemon [--db PATH] [--interval D] [--pid-file PATH] [--log-file PATH]
gua daemon --cloud [--config PATH]        # also stream to GUA Board (after `gua enroll`)
gua daemon --foreground [--db PATH] [--interval D]
gua top [--interval D] [--fake]
gua report [--db PATH] [--since D] [--interval D] [--width N]
gua demo [--db PATH] [--ticks N] [--interval D]
```

- `--interval` on `daemon` controls sampling cadence. Default: `30s`.
- `--interval` on `report` is optional. New DB rows use the interval recorded by each daemon run. Use report `--interval D` only as an override or for legacy rows without interval metadata.
- `--since` accepts `ms`, `s`, `m`, `h`, and `d`, with no upper bound.
- `--foreground` is intended for systemd and debugging.

## Demo Without a GPU

```sh
gua demo
```

The demo records deterministic fake telemetry and immediately prints the report shape.

## Systemd Example

```ini
[Unit]
Description=gua daemon
After=network.target

[Service]
Type=simple
ExecStart=/usr/local/bin/gua daemon --foreground --db /var/lib/gua/gua.db --interval 30s
Restart=on-failure
User=gua

[Install]
WantedBy=multi-user.target
```

Then run:

```sh
systemctl enable --now gpu-usage-audit
```

## Cloud Sync (GUA Board, optional)

`gpu-usage-audit` runs fully local by default. **GUA Board** is a separate service that folds the same telemetry from many hosts into one web view — live utilization graphs next to a reservation timeline, so a shared lab can see at a glance which GPUs are genuinely free, which are reserved, and which are *reserved but sitting idle*.

![GUA Board — live availability across hosts](https://raw.githubusercontent.com/AI-Ocean/gpu-usage-audit/main/docs/img/board.png)

Connect a host in three steps:

```sh
# 1. In the GUA Board web UI, register a server and copy the one-time enrollment token.
# 2. On the GPU host:
gua enroll --server-url https://board.example.com --enrollment-token <TOKEN>
# 3a. Live: run the daemon in cloud mode — pushes snapshots and streams 1s util over WebSocket:
gua daemon --cloud
# 3b. Or one-shot: collect a single snapshot and push the latest state (run on a timer):
gua sync-once
```

How it works and what it does not do:

- `enroll` exchanges the one-time token for a host-scoped, write-only agent token, stored in `~/.gua/cloud.json` with mode `0600`. The token can only write this host's observations — it cannot read reservations, users, or other hosts.
- `daemon --cloud` keeps writing local history as usual, and additionally streams the 1s util samples to the board (so the board's graphs scroll live) and pushes periodic snapshots. The board buffers util in memory only; it stores no per-second history.
- `sync-once` collects one snapshot, **writes it to the local database first**, then pushes only the latest state. A failed push never blocks or rolls back the local write.
- Only the latest state is sent. Historical ticks are kept locally and are never replayed to the server.
- Process telemetry is limited to PID, Linux user, process name (`/proc/<pid>/comm`), and GPU memory — never full command lines.
- The agent only pushes outward. There is no tunnel, no pull, and no remote command execution — the board cannot reach into a host.

Override the config or database path with `--config PATH` / `--db PATH`, and use `gua sync-once --fake` to exercise the flow without a GPU.

## Classification Rules

Each daemon tick records per-card utilization and per-process GPU memory. The report classifies each GPU card at each tick with these rules:

```text
util >= 10                  -> active
util <  10 AND mem >  100   -> idle-held
util <  10 AND mem <= 100   -> truly-idle
```

The 100 MB threshold absorbs runtime baselines such as importing PyTorch or TensorFlow.

## Development

```sh
git clone https://github.com/AI-Ocean/gpu-usage-audit
cd gpu-usage-audit
uv sync
uv run python -m pytest
uv run ruff check
uv run ruff format --check
uv run python -m mypy
uv run gua demo
```

CI runs ruff, format check, mypy, pytest, build, and wheel smoke tests. Tag pushes (`v*`) build release assets and publish to PyPI through Trusted Publishing.

## Non-goals

This is a single-host tool — live (`gua top`) and retrospective (`gua report`) views of the GPUs on the machine it runs on. It does not integrate with cluster schedulers: no Kubernetes cluster scans, Slurm joins, quotas, Docker/Podman runtime fallback, or pod-name resolution. The agent never scans or reaches other hosts. Aggregating many hosts into one live view is the job of the optional [GUA Board](#cloud-sync-gua-board-optional), which the agent only ever pushes to.

The Go v0.1.0 implementation remains available at tag `v0.1.0` and branch [`go-archive`](https://github.com/AI-Ocean/gpu-usage-audit/tree/go-archive).

## License

Apache License 2.0. See [LICENSE](LICENSE).
