Doubao Seedream 5.0 Lite Fast Image Generation

The Problem

In Agent workflows, calling image generation APIs often requires handling cumbersome HTTP requests, file uploads, and asynchronous polling. The seedream 5.0 lite skill encapsulates the image generation capabilities of Doubao Seedream 5.0 Lite into a lightweight CLI client. It directly supports text-to-image and image-to-image, allowing Agents to invoke cloud inference services with minimal friction.

How It Works

This skill is essentially a thin client over the dLazy hosted API, with the following core steps:
- API Authentication: Requests rely on a dLazy API key, which can be saved via the CLI in ~/.dlazy/config.json or injected via the DLAZY_API_KEY environment variable.
- Inference & File Routing: When executing dlazy seedream-5.0-lite, parameters are sent to api.dlazy.com. If local files are passed, the CLI automatically uploads them to files.dlazy.com for the model to read.
- Piping & Async Support: Parameters support pipe references (e.g., @N to extract the Nth output), facilitating data chaining in pipelines. When the --no-wait async mode is enabled, the API returns task: { generateId, status }, which can then be polled using dlazy status --wait.

Boundaries & Considerations

• Cloud Dependency: Entirely relies on dLazy's SaaS service and cannot be used in purely offline environments.
• Error Handling: When receiving 401 (Unauthorized), 503 (Insufficient balance), or 504 (safety policy violation), users must be guided to check their API key, recharge their balance, or adjust their prompts.
• Resource Limits: As a Lite version, generation speed is constrained by dLazy's backend scheduling strategies.