---
license: cc-by-4.0
tags:
- vision-action
- desktop-agent
- os-task
- imitation-learning
- OWA
task_categories:
- robotics
---

# OSTask-demo

[OWA Toolkit Documentation](https://open-world-agents.github.io/open-world-agents/)

A demo dataset of general OS task trajectories with synchronized video and input events. Each trajectory consists of a natural language instruction paired with a human demonstration recording.

## Visualize

Explore recordings directly in your browser with synchronized keyboard/mouse overlay: **👉 [Open in Dataset Visualizer](https://huggingface.co/spaces/open-world-agents/visualize_dataset?repo_id=open-world-agents/OSTask-demo)**

<img src=".github/assets/viewer.png" alt="Dataset Visualizer Preview" width="600">

## Task Domains

- **Browser** (Chromium): Cache cleanup, web navigation
- **Office** (Excel): Spreadsheet manipulation
- **Code Editor** (VS Code): Find and replace, text editing
- **General OS**: Application management, system operations
- **Workflow**: End-to-end tasks spanning multiple applications and tools

## File Structure

Each task has three files:

```
chromium_01.json   # Task instruction
chromium_01.mcap   # Timestamped events + frame references
chromium_01.mkv    # Video recording (HEVC)
```

**JSON**: Contains the natural language instruction for the task.

**MCAP + MKV**: Video and input events in [OWAMcap](https://open-world-agents.github.io/open-world-agents/data/getting-started/why-owamcap/) format—keyboard, mouse, and screen capture with nanosecond timestamps.

## Load the Data

```bash
pip install mcap-owa-support owa-msgs huggingface_hub
```

```python
import json
from huggingface_hub import hf_hub_download
from mcap_owa.highlevel import OWAMcapReader

# Load instruction
json_file = hf_hub_download(
    repo_id="open-world-agents/OSTask-demo",
    filename="chromium_01.json",
    repo_type="dataset"
)
instruction = json.load(open(json_file))["instruction"]

# Load trajectory
mcap_file = hf_hub_download(
    repo_id="open-world-agents/OSTask-demo",
    filename="chromium_01.mcap",
    repo_type="dataset"
)

with OWAMcapReader(mcap_file) as reader:
    for msg in reader.iter_messages(topics=["screen"]):
        screen = msg.decoded
        screen.resolve_relative_path(mcap_file)
        frame = screen.load_frame_array()  # numpy array (H, W, 4) RGBA
        break
```

## Usage

This dataset can be used for:
- Training desktop/OS automation agents
- Evaluating computer use models
- Research on GUI agents and multimodal agents