| --- |
| license: apache-2.0 |
| pretty_name: SaaS-Bench Docker Images |
| language: |
| - en |
| tags: |
| - benchmark |
| - agents |
| - computer-use |
| - llm-agents |
| - browser-use |
| - saas |
| - docker |
| task_categories: |
| - other |
| size_categories: |
| - n<1K |
| --- |
| |
| # SaaS-Bench Docker Images |
|
|
| Docker image archives for the **SaaS-Bench** benchmark — a suite of 23 |
| self-hosted SaaS applications used to evaluate computer-use LLM agents on |
| real, multi-step business workflows. |
|
|
| This repository hosts the **prebuilt `.tar` images** (≈ 63 GB total) so you |
| can reproduce the benchmark environment without rebuilding each app from |
| source. The eval harness, task definitions, and verifiers live in the main |
| [SaaS-Bench](https://github.com/keanshi-nlp/SaaS-Bench) repository. |
|
|
| > **Paper:** *SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS |
| > to Solve Professional Workflows?* |
|
|
| ## Overview |
|
|
| SaaS-Bench evaluates browser-driving LLM agents on **106 task instances |
| across 6 domains**, running on **23 self-hosted SaaS applications**. |
| Each task asks the agent to complete a multi-step workflow (e.g. create a |
| purchase order, configure a project board, schedule a patient visit); a |
| per-task `verify.py` script then inspects the running application's state |
| (DB rows, API responses, filesystem) and returns a pass/fail. |
|
|
| | Track | Domain | Tasks | Representative apps | |
| | ------- | ------ | ----- | ------------------------------------------------ | |
| | uni-m | BOF | 15 | Twenty, Bigcapital, HRMS, Pretix | |
| | uni-m | HA | 16 | OpenEMR, OnlyOffice, OpnForm | |
| | uni-m | SEPM | 31 | Baserow, OpenProject, code-server, Metabase | |
| | uni-m | TCDW | 12 | OnlyOffice, Mattermost, RoundcubeMail, ownCloud | |
| | multi-m | AASC | 12 | Grocy, farmOS, Recipya, e-label | |
| | multi-m | IMC | 20 | SiYuan, Watcharr, BookLore, PhotoPrism, MediaCMS | |
|
|
| Domains: **BOF** = Business Operations & Finance · **HA** = Healthcare & |
| Administration · **SEPM** = Software Eng. & Project Mgmt. · **TCDW** = |
| Team Comms & Document Workflows · **AASC** = Agriculture, Authoring & |
| Supply Chain · **IMC** = Information Mgmt. & Creative. |
|
|
| ## Contents |
|
|
| 23 Docker image archives (`mw-*.tar`) covering every app used by the |
| benchmark. |
|
|
| | File | App / Stack | Size | |
| | ------------------------- | ---------------------- | ------- | |
| | `mw-baserow.tar` | Baserow | 3.07 GB | |
| | `mw-bigcapital.tar` | Bigcapital | 2.94 GB | |
| | `mw-booklore.tar` | BookLore | 1.55 GB | |
| | `mw-code-server.tar` | code-server | 8.46 GB | |
| | `mw-elabel.tar` | e-label | 2.00 GB | |
| | `mw-farmos.tar` | farmOS | 1.13 GB | |
| | `mw-grocy.tar` | Grocy | 286 MB | |
| | `mw-hrms.tar` | HRMS | 5.88 GB | |
| | `mw-mattermost.tar` | Mattermost (+Postgres) | 1.53 GB | |
| | `mw-mediacms.tar` | MediaCMS | 1.76 GB | |
| | `mw-metabase.tar` | Metabase | 889 MB | |
| | `mw-onlyoffice.tar` | OnlyOffice (4-image stack) | 10.46 GB | |
| | `mw-openemr.tar` | OpenEMR | 4.48 GB | |
| | `mw-openproject.tar` | OpenProject | 2.27 GB | |
| | `mw-opnform.tar` | OpnForm | 548 MB | |
| | `mw-owncloud.tar` | ownCloud | 2.14 GB | |
| | `mw-photoprism.tar` | PhotoPrism | 3.82 GB | |
| | `mw-pretix.tar` | Pretix | 2.39 GB | |
| | `mw-recipya.tar` | Recipya | 624 MB | |
| | `mw-roundcubemail.tar` | Roundcube Mail | 1.34 GB | |
| | `mw-siyuan.tar` | SiYuan Notes | 3.06 GB | |
| | `mw-twenty.tar` | Twenty CRM | 2.11 GB | |
| | `mw-watcharr.tar` | Watcharr | 250 MB | |
|
|
| Each tar already contains the `:latest` tag; image names follow the |
| `mw-<app>[-<component>]` convention so loaders can resolve them |
| deterministically. |
|
|
| ## Download |
|
|
| ```python |
| from huggingface_hub import snapshot_download |
| |
| snapshot_download( |
| repo_id="Marti844/SaaS-Bench-docker", |
| repo_type="dataset", |
| local_dir="docker/images", |
| allow_patterns=["*.tar"], |
| ) |
| ``` |
|
|
| Or with the CLI: |
|
|
| ```bash |
| hf download Marti844/SaaS-Bench-docker \ |
| --repo-type dataset --local-dir docker/images \ |
| --include "*.tar" |
| ``` |
|
|
| ## System requirements |
|
|
| - **Disk:** ≥ 130 GB free — ~63 GB for the archives plus the loaded images. |
| - **RAM:** ≥ 500 GB recommended if you run the full eval with the default |
| 4-way parallelism — most stacks bundle their own DB / search / |
| document-server, so total memory grows quickly under concurrency. |
| - **Host OS:** Linux. Tested on Ubuntu 22.04 and Alibaba Cloud Linux. |
| - **Docker:** 24+ with the `compose` plugin. |
|
|
| ## Licensing |
|
|
| - **This card:** Apache 2.0. |
| - **Each bundled Docker image** retains the license of its upstream |
| project (e.g. OnlyOffice — AGPLv3, Mattermost — MIT/AGPLv3 dual, |
| OpenEMR — GPLv3, etc.). The images are redistributed for benchmarking |
| convenience only. Verify upstream terms before any non-research use. |
|
|
