Kujiale AI Design Assistant

The Problem It Solves

Manually translating a floorplan or a set of design preferences into a complete interior design—complete with furniture layout, renderings, and panoramic views—is a fragmented process requiring multiple tools and significant time. There's a clear gap: how to automate the journey from a text description or a single image into a professional, multi-output design package.

How the Skill Works

This skill automates that entire pipeline by orchestrating calls to the Kujiale platform's API suite, following a strict four-phase sequence:

Core Capabilities and Key Steps

1. Floorplan Ingestion & Project Setup: It accepts two input modes.
   • Text Search: Queries the floorplan database (via planSearch) using city and community details. After user selection, it creates a design project (createDesign).
   • Image Recognition: Monitors an inbound media directory for new images. Upon detection, it handles the upload (upload) and trace import (planUrlCreate) process, polling (planUrlCreateResult) until a floorplan is identified and a design basis is created.
2. Style Selection & Intelligent Layout: Based on user-selected preference tags (customTag) and a matched hard finish style (hardStyle), it triggers the backend "Smart Layout" (layout) function. This step consumes account quotas ("beans"), and the skill will prompt for user confirmation. It then polls for the layout result (layoutResult) to retrieve furniture placements per room.
3. Automated Rendering: Once layout is confirmed, it immediately triggers the rendering job. The system generates images asynchronously. The skill continuously polls (renderResult) until completion or a timeout (approx. 5 minutes) is reached.
4. Structured Result Delivery: The final output strictly follows the result.md template, delivering: design highlights summarized from the primary rendering, renderings ordered by room priority (living/dining room -> master bedroom -> secondary bedrooms -> others), a panoramic view link, and a project details link.

Throughout, progress updates are sent via message(action=send), and script versions are validated (versionCheck) to ensure a consistent execution environment.

Applicable Boundaries & Caveats

• Rigid Workflow: The skill executes the documented flow exactly; it cannot handle custom, complex commands that bypass steps (e.g., skipping style selection to go directly to layout).
• Account Dependency: Core functions (smart layout, rendering) depend on the linked Kujiale account's permissions and quota. Insufficient quota or API errors will halt the process.
• Input Requirements: For the image path, files must be placed in a specific directory (~\.openclaw\media\inbound) and are detected after a polling interval (every 5 seconds). Text search accuracy relies on the coverage of the Kujiale floorplan library.
• Result Quality: Final image quality is determined by Kujiale's rendering engine and template library; the skill does not modify the underlying algorithms. Output is filtered by pictype, displaying only standard renderings (pictype=0) and panoramic views (pictype=1).