Inspiration

Specklethron is a highly advanced AI entity originating from Speckleverse that is designed for efficient data search and retrieval. Thanks to its connection to the mother starship "Speckle," it can access your Building Information Modeling (BIM) data seamlessly.

What it does

Specklethron can connect to your Speckle account and browse streams through a simple CLI interface, familiar to every hackathon enthusiast. Occasionally, it motivates developers to add some missing planned functionalities, enhancing the overall experience and capabilities of the system.

How we built it

Semantic Kernel Integration: The Semantic Kernel acts as the brain of Specklethron, handling user inputs and forming prompts based on both agent prompts and Speckle functions. This allows Specklethron to interpret user requests accurately and execute the appropriate actions within the Speckle ecosystem. Loose Coupling of Functions: Functions are designed to be loosely coupled, meaning they can operate independently yet work together cohesively. During the execution of prompts, chat-gpt is calling c# functions - MAGIC!!

Challenges we ran into

The size of prompts and frequency can be challenging to optimize due to open API limits Designing clear descriptions can be sometimes challenging

Accomplishments that we're proud of

Ability to connect chat-gpt with Speckle :D Query models and create some streams.

What we learned

The kernel is able to figure out by itself function chains like - login, get streams, get elements. It is an interesting way to make the back end of the app. Still challenging due to the generative nature of llm's. But allows more flexibility in terms of app structure. Relation between data objects doesn't have to be strictly defined.

What's next for Specklethron

  • Commit comparison
  • 3D view
  • Fancy chat UI
  • RAC

Built With

  • c#
  • chatgpt
  • semantic-kernel
  • speckel
Share this project:

Updates