Datasets:
File size: 4,983 Bytes
fe75e07 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 | ---
license: apache-2.0
pretty_name: LibreOffice Writer computer-use environment (REA)
tags:
- computer-use
- cua
- gui-agent
- rl-environment
- agentenv
- desktop
- linux
- rea
- word-processor
- office
- odf
---
# LibreOffice Writer: a computer-use environment
**The real LibreOffice Writer 24.2 word processor running on an Ubuntu 24.04 desktop, as an [agentenv](https://pypi.org/project/agentenv-framework/) computer environment. An agent completes the task only through screenshots, mouse and keyboard, and a verifier grades the saved files with no model.**
Writer opens a delivery notice from a corner shop. The agent has to move the delivery to a new date by editing the line under the title, keep the time, and save the document. It sees only the screen and acts only with the mouse and keyboard.
## Watch
- Agent runs (Claude Code with Claude Sonnet 5.5): [run 1](media/solve-1.mp4), [run 2](media/solve-2.mp4), [run 3](media/solve-3.mp4)
- Scripted witness, which proves the task can be done through the computer: [witness.mp4](media/witness.mp4)
Each video is the computer's own screen recording at 10 frames per second.
## The task
> In the delivery notice open in LibreOffice Writer, move the delivery to Monday 24 November (same time, 8am to 10am) by changing the date under the title, then save the notice.
`bundle/artifacts/move-delivery-check/verify.py` reads the app's files through the computer's `data/get` view and passes only if every check holds:
| Check | Kind | File |
| --- | --- | --- |
| `new-date` | `value_equals` | `Documents/notice.odt` |
| `title-kept` | `value_equals` | `Documents/notice.odt` |
| `signature-kept` | `value_equals` | `Documents/notice.odt` |
The `.odt` is a zip. The checks read `content.xml` inside it and find paragraphs by their paragraph style, never by position. The subtitle must read the new date exactly, and the title and signature must be unchanged.
## Results
3 of 3 runs passed, about $0.58 per run.
| Run | Outcome | Turns | Cost | Video |
| --- | --- | --- | --- | --- |
| 1 | passed | 11 | $0.72 | [video](media/solve-1.mp4) |
| 2 | passed | 14 | $0.57 | [video](media/solve-2.mp4) |
| 3 | passed | 10 | $0.44 | [video](media/solve-3.mp4) |
## Run it
You need Docker. `agentenv-framework` builds the computer: Ubuntu 24.04 with the app installed from Ubuntu's archive (`libreoffice-writer` 24.2.7).
```bash
pip install agentenv-framework==0.9.1431
hf download earakely-scale/LibreOfficeWriterComputerEnv-REA --repo-type dataset --local-dir writer
agent-env run writer/bundle --task move-delivery --agent claude-code --model <model>
```
The agent needs a model provider configured for agentenv. `--task smoke` only boots the computer, with no agent. `bundle/README.md` describes the bundle in detail.
## How it was made
[REA](https://github.com/morluto/rea) built this environment from a real session with the app:
1. **Capture.** REA launched LibreOffice Writer on a Linux desktop and recorded its package, windows and opened files while a scripted user made the edit. It also recorded the app's state folders before and after.
2. **Export.** REA turned the capture into an agentenv `computer_app`. It installs the app, seeds the starting files, starts the app, and waits for its main window. It also serves typed views of the app's files for verifiers.
3. **Task.** The task's checks were validated against the capture: they fail on the starting state and pass on the captured end state.
4. **Witness and runs.** Scripted computer actions solved the task in a fresh computer, and the verifier failed before, passed after and failed again after `data/reset`. Then an agent solved it 3 times, each run recorded.
## What's inside
```
bundle/ the agentenv bundle: envs/ (app module, seed files), tasks/, artifacts/ (verifier), evals/
media/ witness and agent-run recordings (MP4) with short GIF previews
results.json the runs above, with each check's result
```
## Limits
- The seeded Writer profile turns off Writer's revision IDs (Tools > Options > Writer > Comparison, `StoreRSID`). With them on, which is Writer's default, retyping text splits the paragraph into styled runs, and the checks, which read the paragraph's first run, would fail a correct edit. A check on a paragraph's whole text would remove this need.
- The subtitle paragraph is found by its `Subtitle` style. An edit that changes the paragraph's style fails even if the text is right.
- The computer uses the `libreoffice-writer` version in Ubuntu 24.04's archive at build time; it is not pinned.
- Three runs is a small sample: the solve rate shows the task is solvable, not how hard it is.
## Licences
The bundle code, the task and the synthetic seed files are Apache-2.0. LibreOffice Writer itself is not included: the computer installs it from Ubuntu's archive (`libreoffice-writer` 24.2.7) at build time, under its own licence (MPL-2.0).
|