CoView

KO | EN

gitlite — search

#javascript #react #llm #go #machine-learning #rust #linux #docker #java #awesome #awesome-list #pytorch #openai #open-source

(★ 19)

A desktop AI companion that sees, acts, and collaborates, turning your computer into a living workspace.

CoView 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1 align="center">CoView - Desktop AI Companion</h1>

<a href="README.zh-CN.md">中文文档</a> ·
 <a href="docs/en/README.md">Full Documentation</a>

<a href="http://www.coview.top/">Visit Official Website: www.coview.top</a>

A desktop AI companion that sees, acts, and collaborates, turning your computer into a living workspace.

---

## Why CoView

CoView is built for a simple idea: your computer should not just wait for commands. It should see what you see, understand what you are trying to do, and work with you across the apps already on your screen.

At its core, CoView is powered by a local visual control loop:

```text
Observe screen -> reason about the task -> execute one action -> observe again
```

That means CoView is not limited to chat. It can operate browsers, editors, documents, websites, desktop apps, and coding workspaces through mouse, keyboard, voice, and background Code Agent workflows.

## Highlights

| Capability                   | What CoView Does                                                                                                   |
| ---------------------------- | ------------------------------------------------------------------------------------------------------------------ |
| 👀 See                       | Observes your desktop through screenshots and visual reasoning.                                                    |
| 🖥️ Multi-Screen Control    | Understand and manipulate windows and content across multiple monitors, and quickly perform GUI operations.        |
| 🖱️ Act                     | Clicks, drags, scrolls, uses hotkeys, inputs text, and reads pages or documents.                                   |
| 🤝 Collaborate               | Works from a floating assistant UI with task input, stop control, logs, settings, and suggestions.                 |
| 🎙️ Listen & Respond        | Supports ASR, TTS, local wake-word detection, and voice indicators.                                                |
| 🧑‍💻 Background Code Agent | Run codes and automated tasks asynchronously in the background, supporting Codex, Claude Code, Kimi Code and more. |
| 🌐 Chinese & English         | Provides bilingual product flows and documentation.                                                                |
| 🔌 Model Flexible            | Uses OpenAI-compatible endpoints through `base_url`, `api_key`, and `model_name`.                            |

## How CoView Works

## Voice Interaction

Default wake words: `你好小彤` and `hey Lucy`.

## Quick Start

Requirements:

- Python 3.10 or newer. Python 3.11/3.12 are recommended.
- macOS or Windows.
- A visual model endpoint compatible with the OpenAI-style API shape used by this project.

### 1. Clone the Repository

| macOS / Linux                                          | Windows PowerShell                                     |
| ------------------------------------------------------ | ------------------------------------------------------ |
| `git clone https://github.com/mini-yifan/CoView.git` | `git clone https://github.com/mini-yifan/CoView.git` |
| `cd CoView`                                          | `cd CoView`                                          |

### 2. Create and Activate a Virtual Environment

| macOS / Linux                 | Windows PowerShell             |
| ----------------------------- | ------------------------------ |
| `python3 -m venv .venv`     | `py -3 -m venv .venv`        |
| `source .venv/bin/activate` | `.venv\Scripts\Activate.ps1` |

If PowerShell blocks activation, run this once in the same PowerShell window:

```powershell
Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
```

### 3. Install CoView

| macOS                                              | Windows                                 |
| -------------------------------------------------- | --------------------------------------- |
| `python3 -m pip install -U pip`                  | `py -m pip install -U pip`            |
| `python3 -m pip install -e ".[macos,voice,tts]"` | `py -m pip install -e ".[voice,tts]"` |

### 4. Configure Your Model

CoView reads `config.json` from the repository root and merges it with defaults from `src/baodou_ai/core/config.py`.

```json
{
  "api_config": {
    "api_key": "YOUR_API_KEY",
    "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
    "model_name": "qwen3.6-35b-a3b"
  }
}
```

Do not commit real API keys to a public repository.

Default model services:

| Capability            | Default Provider / Model                                 | Notes                                                              |
| --------------------- | -------------------------------------------------------- | ------------------------------------------------------------------ |
| Visual language model | Alibaba Cloud Model Studio /`qwen3.6-35b-a3b`          | Can be changed to another OpenAI-compatible visual model provider. |
| Realtime ASR          | Alibaba Cloud Model Studio /`qwen3-asr-flash-realtime` | Provider switching is not supported yet.                           |
| TTS                   | Alibaba Cloud Model Studio /`cosyvoice-v3-flash`       | Provider switching is not supported yet.                           |

Apply for the API key from [Alibaba Cloud Model Studio](https://www.alibabacloud.com/help/zh/model-studio/get-api-key). The same DashScope API key is used for the default visual model, [ASR](https://help.aliyun.com/zh/model-studio/asr-model/), and [TTS](https://www.alibabacloud.com/help/zh/model-studio/text-to-speech). Support for additional ASR and TTS providers will be added gradually.

### 5. Run the App

| Goal                   | macOS                                                           | Windows                                              |
| ---------------------- | --------------------------------------------------------------- | ---------------------------------------------------- |
| Start the floating GUI | `coview`                                                      | `coview`                                           |
| Run one CLI task       | `coview-cli "Open the browser and search today's weather"`    | `coview-cli "Open Notepad and type Hello"`         |
| Limit task steps       | `coview-cli "Summarize the current page" --max-iterations 20` | `coview-cli "Open Calculator" --max-iterations 20` |
| Stop a CLI task        | `Ctrl+C`                                                      | `Ctrl+C`                                           |

### Voice Wakeup

To use hands-free wakeup, download the local wake-word model first:

| macOS / Linux                                   | Windows                                    |
| ----------------------------------------------- | ------------------------------------------ |
| `python3 scripts/download_wake_word_model.py` | `py scripts\download_wake_word_model.py` |

Default wake words are `你好小彤` and `hey Lucy`. Say `close program`, `quit app`, or similar `close/exit/quit program/app` commands to quit CoView by voice.

## First Interaction in 60 Seconds

1. Start the GUI with `coview`.
2. Open the floating assistant settings and enter your model API key, base URL, and model name. You can type `/setting`, `/settings`, or `/设置` in the input box to open Settings directly.
3. Click the floating assistant input, type a task, and press Enter.
4. Watch CoView observe the screen, choose an action, execute it, and observe again.
5. Use the stop button if the task is wrong, too broad, or interacting with the wrong window.

Good first tasks:

```text
Open the browser and search for the weather in Shanghai.
Summarize the article in the current browser tab.
Open Calculator and compute 128 * 46.
Read the visible document and list the action items.
Create a background code-agent task to inspect this repository's test structure.
```

## Documentation

| Guide                    | Link                                                          |
| ------------------------ | ------------------------------------------------------------- |
| Full documentation index | [docs/en/README.md](docs/en/README.md)                           |
| Product overview         | [docs/en/product-overview.md](docs/en/product-overview.md)       |
| Setup and configuration  | [docs/en/setup-configuration.md](docs/en/setup-configuration.md) |
| Voice interaction        | [docs/en/voice-interaction.md](docs/en/voice-interaction.md)     |
| CLI and Python API       | [docs/en/usage-cli-api.md](docs/en/usage-cli-api.md)             |
| Architecture             | [docs/en/architecture.md](docs/en/architecture.md)               |
| Agent protocol           | [docs/en/agent-protocol.md](docs/en/agent-protocol.md)           |
| Development              | [docs/en/development.md](docs/en/development.md)                 |
| Safety and contributing  | [docs/en/safety-contributing.md](docs/en/safety-contributing.md) |

## Open Source Acknowledgements

CoView is built on top of excellent open source projects, including [PyQt5](https://pypi.org/project/PyQt5/), [OpenCV](https://opencv.org/), [NumPy](https://numpy.org/), [MSS](https://github.com/BoboTiG/python-mss), [PyAutoGUI](https://github.com/asweigart/pyautogui), [OpenAI Python SDK](https://github.com/openai/openai-python), [tiktoken](https://github.com/openai/tiktoken), [Pydantic](https://github.com/pydantic/pydantic), [Pillow](https://python-pillow.org/), [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx), [SentencePiece](https://github.com/google/sentencepiece), and [sounddevice](https://github.com/spatialaudio/python-sounddevice).

Please refer to each upstream project for its license and attribution requirements.

## License

CoView is released under the [MIT License](LICENSE).

CoView

CoView 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

CoView

CoView 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README