File size: 1,280 Bytes
ec37783
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ed006ec
 
 
ec37783
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: mit
task_categories:
  - robotics
  - reinforcement-learning
tags:
  - computer-use
  - trajectory
  - human-demonstration
  - cua-format
---

# trajectory

Computer Use Agent (CUA) format trajectories for training computer use models.

## Dataset Description

This dataset contains human-recorded computer use trajectories in CUA format.
Each trajectory represents a sequence of actions (clicks, typing, scrolling, etc.)
performed on a computer to accomplish a task.

## Latest Trajectory

- **Name**: Pole_1
- **Description**: No description
- **Total Steps**: 9
- **Tags**: None

## Format

Each trajectory file is a JSON object with the following structure:

```json
{
  "timestamp": "ISO-8601 timestamp",
  "session_id": "unique trajectory ID",
  "name": "recording name",
  "tool_calls": [
    {
      "type": "computer_call",
      "call_id": "unique call ID",
      "status": "completed|failed",
      "action": {
        "type": "click|type|scroll|keypress|...",
        "x": 123,
        "y": 456,
        ...
      }
    }
  ],
  "messages": [...],
  "tags": [...],
  "images": ["base64 screenshots..."]
}
```

## Source

Generated by [Project OWL](https://github.com/growthpods/project-owl) - Computer Use Agent Training Platform

## License

MIT License