dLazy GPT Image 2 Image Generation and Editing

What problem it solves

This skill is for local workflows that need to call an image model: turning a text prompt into an image, or editing and synthesizing images based on existing reference inputs. It is not an offline local model, and it does not perform complex post-processing. Its core job is to connect an Agent’s inputs to dLazy’s hosted GPT Image 2 inference endpoint while handling authentication, local file reads, async tasks, and common error reporting.

How the skill works

• Text-to-image generation: after providing a prompt and generation parameters, the skill sends the request through the dLazy CLI to the dLazy API and receives either the generated result or a task status.
• Image editing and synthesis: when local image paths are passed in, the CLI uploads those files to dLazy media storage so the model can use them as reference inputs; outputs are typically hosted file URLs.
• Authentication and configuration: requests depend on a dLazy API key, which can come from the user configuration or an environment variable; the key is scoped to the organization and can be rotated or revoked from the dashboard.
• Async handling: when the request is configured not to wait, the API returns task metadata, and the caller must later poll task status to retrieve the final image.
• Error handling: the skill surfaces errors such as insufficient_balance, unauthorized, and missing local files, making it easier for the Agent to tell the user to top up, configure the key, or fix input paths.

Boundaries and cautions

This is a cloud inference client, not a local model service. When local files are used, they are uploaded to dLazy storage; when content is generated, results are subject to model policies and dLazy’s service terms. It fits scripted image generation, reference-image editing, and batch task integration. It is less suitable for fully offline use, environments where sensitive data cannot leave the machine, or workflows that require a private local model.