Testing

While our AI systems aim to follow the given prompts closely, their output may not always align perfectly with the exact requirements. To ensure seamless operations, we have implemented multiple safeguards to prevent runtime errors or crashes. Let's look at a simple test we conducted:

We conducted a basic test:

1. Prompt: Build a simple bridge.

2. Agent Allocator: All three agents were allocated for the task.

   1. Agent 1 - Image Generation Prompt: "Please provide the dimensions and location of the bridge you need to build."

   2. Agent 2 - TTC Prompt: "Build a bridge with the provided dimensions and location."

   3. Agent 3 - Web Agent: The agent gathered relevant links to free 3D models of bridges in glb and glTF formats, and provided the following summary: "The results include free downloadable 3D models of bridge building in glb and glTF formats from various sources. Specifically, the results include links to DownloadFree3D.com, TurboSquid, and Sketchfab. Other possible queries include searching for bridge building software or bridge building tutorials."

      1. https://downloadfree3d.com/file-format/glb/

      2. https://www.turbosquid.com/Search/3D-Models/free/gltf

      3. https://sketchfab.com/tags/building

Resutls

The image generation process was straightforward, despite a slightly odd prompt. The generated image from DALL-E was as expected. The interesting part was the interaction between the DALL-E and TTC agents. As shown in the accompanying image, TTC attempted to replicate the design present in DALL-E's output, achieving some level of similarity.

The complex feature of connecting TTI and TTC is essentially a text-to-image generation process. A simplistic approach, details of which are not disclosed currently, has been used for this purpose. While it may not always provide accurate results, it functions satisfactorily for the moment. Upgrades to this process are forthcoming.

Generated Image

Generated Structure