AI Playground history and background

The AI Playground helps developers share their ideas and get them into production quickly.

  • 🧭 V0.1 Dec 2024: Implementation of long range planning methodology (SOFT)

    The first idea is the implementation of a long range planning methodology. Many companies use SWOT. SWOT probably originated from the Stanford Research Institute's Long Range Planning Service, known as SOFT. SWOT is nothing more than a re-labelling of the SOFT acronyms. The team of open source developers is dedicated to modernising this approach by integrating the latest Agentic RAG technology. The SOFT management framework is relatively unknown to the general public. This gives us the advantage of experimenting with untrained AI models. Most models only know the SWOT methodology.

  • ⚙️ V0.1 Dec 2024: Achievement

    The AI correctly picks up the sentiment of the problem from the user by selecting the "T" for threats. It suggests a good description, a reference and actions to take. The user can use this as a source of inspiration. Accuracy to the standard jumps from 40 to 90%.

  • ⚙️ V0.2 Jan 2025: UI-Concept

    Inspired by mobile mood apps, an innovative user interface concept has been adopted that encourages user interaction through chat. Based on user input, the application can suggest context, configurations and workflows. The AI in the background is instructed to distinguish between assistance and task input. The user can agree or disagree to keep the behaviour transparent. The app must be able to undo its actions upon the users request.

  • 💡 V0.3 March 2025: Use business process workflow tool called n8n

    The open source team plans to simplify the chat and make it fit into any workflow. The developers are currently investigating a business process management tool called n8n, which is known for its agentic AI capabilities. The idea is that the AI figures out what the user wants to do and chooses the right workflow to do it.

  • ✨ V0.4 March 2025: User in the loop

    The “human in the loop” provision in the EU AI Act is a cornerstone of the EU’s approach to regulating high-risk AI systems. By mandating meaningful human oversight, the Act aims to mitigate risks associated with automation while ensuring that AI remains a tool that serves, rather than replaces, human judgment. This balanced approach was embraced by the development team and rephrased as "user in the loop," meaning that the AI-reasoning is shared between the frontend and the backend, giving control optionally back to the user.

  • ⚙️ V0.5 Mai 2025: 100% n8n-background

    The goal of having an all n8n background was challenging but made fun in the end. Workflows in n8n are sequential in nature. We felt that the app needed asynchronous processing to keep it alive during user interactions. This was achieved by adding the simple MQTT message broker mosquitto to the pod. Backend security was also a challenge. This was solved by using auth_request from NGINX. We now use JWT by default to secure the n8n background and we have a good feeling about it.
    auth_request:

    The auth_request module in NGINX enables external authorization for HTTP requests by delegating authentication decisions to the n8n service (in this case). This approach allows NGINX to act as a gatekeeper, permitting or denying access based on the response from an external authentication server.

    Source: Perplexity

  • 🧭 Summer 2025: Probing the direction of AI – exciting times ahead!

    I'm thrilled to tell you that, at the time of writing, the big drive of AI is making great strides on standards like Agent Communication Protocol (ACP), Agent2Agent Protocol (A2A) and, most of all, the Model Context Protocol (MCP), which finds most implementations. The development team is super excited to figure out where we will land and which of the competing standards will be embraced.

Video: Older version V0.2