File size: 6,889 Bytes
913eb00
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
---
license: apache-2.0
pretty_name: Dia computer-use environment from a one-line request (REA)
tags:
- computer-use
- cua
- gui-agent
- rl-environment
- agentenv
- desktop
- linux
- rea
- diagram
- flowchart
- from-request
---

# Dia: a computer-use environment built from a one-line request

![An agent solving the task in Dia](media/hero.gif)

**The real Dia 0.98 diagram editor 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.**

Claude Code built this environment from a one-line request in about 22 minutes, with only REA's build skill and MCP tools; [How it was made](#how-it-was-made) has the request and the steps. Dia opens a small order-processing flowchart, `Documents/order-flow.dia`, at launch: four boxes, Receive order, Check stock, Pack items and Ship order, joined by lines. The agent has to rename the Check stock box to Verify inventory and save. 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)
- 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 my order flowchart, Documents/order-flow.dia, rename the "Check stock" box to "Verify inventory" and save it.

`bundle/artifacts/rename-stock-box-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 |
| --- | --- | --- |
| `renamed` | `text_equals` | `Documents/order-flow.dia` |
| `one-box-named` | `member_count` | `Documents/order-flow.dia` |
| `four-boxes` | `member_count` | `Documents/order-flow.dia` |
| `guard-receive` | `text_equals` | `Documents/order-flow.dia` |
| `guard-pack` | `text_equals` | `Documents/order-flow.dia` |
| `guard-ship` | `text_equals` | `Documents/order-flow.dia` |

A `.dia` file is XML. The seed is gzip-compressed, as Dia saves a diagram when its "Compress diagram files" preference is ticked, and REA reads the XML inside. Dia writes a box's text between `#` marks. The checks select the diagram's layer by what it holds, the layer with the Receive order box (`[>dia:object=#Receive order#]`, a selector on a child element's text), then find boxes by their object `id`: box O1 must read "Verify inventory" exactly, and Receive order, Pack items and Ship order must keep their text. Two member counts complete them: exactly one flowchart box reads Verify inventory, and the layer holds exactly four boxes. REA's draft suggested the text check plus four checks on the box's position and size, which Dia recomputes when the text changes; Claude Code kept the text check, discarded the geometry checks, and added the counts.

## Results

2 of 2 runs passed, about $1.75 per run.

| Run | Outcome | Turns | Cost | Video |
| --- | --- | --- | --- | --- |
| 1 | passed | 39 | $1.47 | [video](media/solve-1.mp4) |
| 2 | passed | 56 | $2.03 | [video](media/solve-2.mp4) |

## Run it

You need Docker. `agentenv-framework` builds the computer: Ubuntu 24.04 with the app installed from Ubuntu's archive (`dia` 0.98+git20240130).

```bash
pip install agentenv-framework==0.9.1431
hf download earakely-scale/DiaFromRequestComputerEnv-REA --repo-type dataset --local-dir dia-request
agent-env run dia-request/bundle --task rename-stock-box --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

Claude Code built this environment from a one-line request, given only [REA](https://github.com/morluto/rea)'s build skill and REA's MCP tools:

> Build me a Dia environment where the agent renames a box in a small flowchart and saves it.

It asked only the skill's upfront questions, wrote no app-specific scripts, and finished in about 22 minutes.

The first witness failed at `after_edit`: Dia adds an empty layer when it reopens a diagram, so after a save the file held two layers and the checks' plain layer step matched both. Claude Code rewrote the checks to select the layer by what it holds, exported again, and the second witness held. The steps were REA's usual ones:

1. **Capture.** REA launched Dia on a Linux desktop and recorded its package, windows and opened files while the edit was made. 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.** In a fresh computer, the edit was made through computer actions alone, and the checks were read at four stages. With the rewritten checks they held:

| Stage | Checks |
| --- | --- |
| `seed` | fail, as expected: the box reads Check stock; the count and kept-box checks pass |
| `after_edit` | pass |
| `near_miss` | fail, as expected: the box was retyped as "Verify inventry" |
| `after_reset` | fail, as expected: `data/reset` restored the seed |

The agent runs came after the build: Claude Code with Claude Sonnet 5.5 (`--model anthropic/claude-sonnet-5-5 --no-emulate`, on an amd64 host) solved the task twice, 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

- Only two runs: the build stopped at its $4 solve budget before a third. Two runs show the task is solvable, not how hard it is.
- Dia's window opens small, with most of the flowchart out of view and its title bar under the desktop's top panel, so it can't be maximized from the title bar (Alt+F10 does it). The runs took 39 and 56 turns.
- The renamed box is found by its object id, O1. A new box drawn in place of Check stock does not get that id, so it fails even when the chart reads right. The lines between the boxes are not checked.
- The computer installs whatever `dia` Ubuntu 24.04's archive holds at build time (0.98+git20240130 for these runs); it is not pinned.

## Licences

The bundle code, the task and the synthetic seed files are Apache-2.0. Dia itself is not included: the computer installs it from Ubuntu's archive (`dia` 0.98+git20240130) at build time, under its own licence (GPL-2.0-or-later).