AiChuang AI Platform

What Problem It Solves

E-commerce operators frequently need to generate visual assets such as product images, videos, and detail pages. This traditional process often involves opening design software, repeatedly adjusting parameters, and switching between different platforms (like image generators and video editors), which is fragmented and time-consuming. For developers or automation users, there is a lack of a service-oriented interface that allows invoking these creative capabilities directly via conversational instructions.

How The Skill Works

This skill encapsulates the core functionalities of the "AiChuang AI" platform into callable APIs, enabling you to generate content by issuing commands directly within an AI conversation. The core mechanism is completing the entire creative process through HTTP requests.

Core Capabilities Covered

• AI Image Generation: Supports 18 generation modes, including the recommended Agent mode. In Agent mode, users upload a reference image and describe their needs, after which the AI autonomously analyzes and generates optimized images. Other modes like "Remove Watermark" or "AI Try-On" are scene-specific with fixed logic.
• AI Detail Page Generation: After uploading product images, the AI automatically analyzes and plans a complete e-commerce detail page (including modules like selling points, parameters, etc.), then generates each image sequentially.
• Style Replication: By uploading a set of "reference design images" and "product material images", the AI learns the style from the references and applies it to your products to generate new images with a consistent aesthetic.
• Video Generation: Supports the creation of various short videos, such as product explanations and promotional content.
• Asset Management: Queries or batch-manages all your generated assets (images, videos, detail pages, etc.) on the platform.

Key Workflow Steps

1. Authentication & Authorization: All API calls require a user Token. On first use, the skill guides the user to log in to the AiChuang AI platform via a browser and copy their personal Token. This Token is saved in the session context for all subsequent requests.
2. Task Submission: The core operation for generating content is sending a request to the POST /api/v1/Task/Publish endpoint. For image generation, you must specify the sceneMode (e.g., free_creation for free creation or Agent mode), provide a prompt (a mandatory field describing the desired output), and include the reference image's imageUrl. For detail pages and style replication, you first create a batch (CreateBatch), then proceed with confirming the plan and generating images step-by-step.
3. Resource Handling: If the user provides a local image, the skill first encodes it to Base64 and uploads it to the platform's OSS to obtain a publicly accessible URL for the generation request.
4. Status Polling & Result Retrieval: After submitting a task, a taskJobId or conversationId is returned. You need to poll interfaces like GetProgress to check the task status until it completes or fails, then retrieve the URL of the generated result.
5. Session Management & Cost: To maintain historical record continuity, the skill queries and reuses an existing conversationId before publishing a task, rather than creating a new session for each generation. All generation operations consume credits, and the skill informs the user of the estimated cost before submission.

Applicable Boundaries and Important Notes

• Token is a Mandatory Credential: Without a valid Token, all API calls will fail (returning a 401 error). The responsibility for obtaining and securing the Token lies with the user, and it should be used in a private environment.
• Credit-Based Model: Generation is metered, with different costs for different functions (e.g., image generation typically costs 5 credits/image, detail pages 15 credits/image). The skill will reference a pricing table for prompts, but querying an exact balance requires an additional two-step authentication process.
• Strict Prompt Rules: For fixed modes other than Agent/free creation (e.g., "Remove Watermark"), you must use the default preset prompt for that mode. Using a custom prompt will cause the task to fail with a parameter error (error code 999). In Agent/free creation mode, the prompt is freely described by the user.
• Asynchronous Tasks and Queues: Generation tasks are processed asynchronously. After submission, there is a processing delay, and you need to actively poll for results. Under high load, tasks may be queued, requiring patience during polling.
• API Operation Robustness: While the skill provides detailed API call examples (covering macOS/Linux and Windows), practical execution can be affected by network issues, timeouts, and response formats, necessitating appropriate error handling.