> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gravitex.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# WorkBuddy-Image Generation

> Connect GravitexAI text-to-image and image-editing models to WorkBuddy with custom connectors

This guide applies to macOS and Windows. Except for fully quitting the app, the steps are the same. Custom connectors let WorkBuddy use GravitexAI image models for text-to-image, image editing, and same-prompt comparisons.

## 1. Capabilities

One JSON configuration exposes multiple image models without creating a separate account for each.

| Capability | Description |
| - | - |
| One key, multiple models | Use GPT, Gemini, and other image models with one GravitexAI key |
| Local key storage | Connector configuration, including the key, is stored locally rather than uploaded to the cloud |
| Short names | Give each model a name such as gemini-pro and request it in chat |
| Model comparisons | Ask several models to draw the same subject |
| Usage-based billing | Image API usage is billed to GravitexAI, without consuming WorkBuddy credits or plan allowance |

## 2. Requirements

* Windows 10 or later; Windows 7, 8, and 8.1 are unsupported.
* macOS 12 Monterey or later.

> [!TIP]
>
> WorkBuddy cannot start on systems that do not meet these requirements.

## 3. Prerequisites

* WorkBuddy installed.
* A key (`sk-xxxxxxxxxx`) from the [GravitexAI Console](https://maas.gravitex.ai/#/api-keys).
* **On Windows:** install [Node.js LTS](https://nodejs.org/) with the default options so `npx` is available.

## 4. Quick start

### Step 1: Create an API key

Open the [GravitexAI Console](https://maas.gravitex.ai/#/api-keys), create and name a token, and copy the key beginning with `sk-`.

### Step 2: Open Custom connectors

In the left sidebar, open **Experts · Skills · Connectors**, switch to **Connectors**, and click **Custom connector** at the top right. The screenshot shows the Chinese UI.

![WorkBuddy custom connector entry](https://mintcdn.com/gravitexai/QXITE-rsvJBzaoaP/images/workbuddy-image-generation-1.png?fit=max\&auto=format\&n=QXITE-rsvJBzaoaP\&q=85\&s=51602bc7ae5132f1a9b2c39f57e38e69)

### Step 3: Paste the JSON configuration

Paste the complete configuration below and **replace every `sk-your-api-key` with your key**.

```json theme={null}
{
  "mcpServers": {
    "gpt-image-2": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gpt-image-2"
      }
    },
    "gpt-flare": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gpt-image-2.5-flare"
      }
    },
    "gpt-sunburst": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gpt-image-2.5-sunburst"
      }
    },
    "gpt-image-1": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gpt-image-1"
      }
    },
    "gemini-pro": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gemini-3-pro-image"
      }
    },
    "gemini-flash": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gemini-3.1-flash-image"
      }
    },
    "gemini-2.5-flash": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gemini-2.5-flash-image"
      }
    },
    "gemini-flash-lite": {
      "command": "npx",
      "args": ["-y", "@junjun-org/proxy-gpt-image@latest"],
      "env": {
        "OPENAI_BASE_URL": "https://api.gravitex.ai/v1",
        "OPENAI_API_KEY": "sk-your-api-key",
        "IMAGE_MODEL": "gemini-3.1-flash-lite-image"
      }
    }
  }
}
```

> [!NOTE]
>
> Connector names such as gpt-flare are customizable and appear in WorkBuddy. On Windows, install Node.js LTS first and use `npx` as command, not `npx.cmd`.

### Step 4: Restart and trust

Fully quit and reopen WorkBuddy:

* macOS: `Cmd + Q`; closing a window alone is insufficient.
* Windows: exit from the taskbar/tray, or use `Ctrl + Shift + Esc` to confirm the process has stopped before reopening.

Return to **Connectors → Custom connectors** and click **Trust** for the connectors you want. Start with two or three frequently used models; each connector runs an `npx` process, so trust others when needed.

> [!WARNING]
>
> New connector settings may not take effect until WorkBuddy is fully restarted. Confirm the process has exited, rather than just closing the window.

### Step 5: Start generating

In a task, open **+ → Connectors** and select the image models you need. See the examples below.

## 5. Model reference

| Connector name | Model ID | Role |
| - | - | - |
| `gpt-image-2` | `gpt-image-2` | General-purpose GPT model |
| `gpt-flare` | `gpt-image-2.5-flare` | Newer GPT variant |
| `gpt-sunburst` | `gpt-image-2.5-sunburst` | Another newer GPT style |
| `gpt-image-1` | `gpt-image-1` | Previous GPT generation |
| `gemini-pro` | `gemini-3-pro-image` | High-capability Gemini option |
| `gemini-flash` | `gemini-3.1-flash-image` | Fast tier |
| `gemini-2.5-flash` | `gemini-2.5-flash-image` | Previous fast tier |
| `gemini-flash-lite` | `gemini-3.1-flash-lite-image` | Lowest-cost option in this list |

> [!NOTE]
>
> Check exact IDs in the [GravitexAI Model Catalog](https://maas.gravitex.ai/#/api-models). To request a model not listed here, such as `seedream-5-0-260128`, specify its full ID in chat to override the default for that call.

## 6. Daily use

* **Set a task default:** choose a connector in **+ → Connectors**.
* **Switch by short name:** say “Use flash” or “Try pro for another version.”
* **Use another model temporarily:** give its full ID, such as `seedream-5-0-260128`, to override the default without changing the configuration.

To change the saved default model, remove and re-add the custom connector with a new `IMAGE_MODEL` value, then fully restart WorkBuddy; custom connectors do not support editing in place.

For text-to-image, describe the desired picture. For image-to-image, supply a reference and describe the changes.

## 7. Prompt examples

### Text-to-image with the selected default

> Draw a flat red apple icon at 1024x1024.

### Choose a model by name

> Use gemini-pro to draw a sunrise over snowy mountains.

> Use flash to draw the same scene again.

> Use gpt-flare to turn this picture into watercolor.

### Edit an existing image

> Use the lamb-grazing.png file on my desktop and change the background to red.

> Change daytime to nighttime in this picture and keep everything else unchanged.

### Size and quality

> Draw a landscape at 1536x1024 using low quality for a quick result.

> Generate a vertical phone wallpaper at 1024x1536 with high quality.

Sizes: `1024x1024` (square), `1536x1024` (landscape), `1024x1536` (portrait). Quality: `low` (fast/economical), `auto` (default), or `high` (detailed; around 1–2 minutes).

### Compare models

> Have gpt-image-2 and gemini-pro each draw the same subject so I can compare them.

## 8. Troubleshooting

> [accordion] The configured connector is missing
>
> Replace all placeholder keys in the JSON and fully restart WorkBuddy. On macOS use Cmd+Q; on Windows confirm the process has ended in Task Manager before reopening.

> [accordion] The request returns 401
>
> Check that the key begins with sk- and has no extra spaces. Verify token status and balance in the GravitexAI Console; insufficient balance may also produce 401.

> [accordion] Image editing returns 404
>
> Some gateways use a different edits route and require an editPath setting. Check the model's image-editing endpoint and the connector's corresponding env configuration.

> [accordion] Windows cannot find npx
>
> Install Node.js LTS and ensure it is on PATH, then restart WorkBuddy.

> [accordion] Does image generation consume WorkBuddy credits?
>
> Image API costs are billed to GravitexAI rather than WorkBuddy credits or plan allowance. Monitor the GravitexAI balance and usage.

> [accordion] Can I edit the default model in place?
>
> Remove the connector, re-add the JSON with the desired IMAGE\_MODEL value, and fully restart WorkBuddy.

## 9. Resources

* [WorkBuddy website](https://www.workbuddy.cn)
* [GravitexAI API keys](https://maas.gravitex.ai/#/api-keys)
* [GravitexAI models](https://maas.gravitex.ai/#/api-models)
* [WorkBuddy chat-model configuration](https://docs.gravitex.ai/en/integrations/workbuddy.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.