Modern software development is increasingly assisted by Large Language Model (LLM) agents. To help these tools understand MindFusion API and generate accurate code, we provide a unified LLM-friendly documentation index, following https://llmstxt.org/ proposal. The llms.txt for MindFusion Diagramming Toolkit for Java Swing is located here:

https://mindfusion.dev/docs/java-swing/diagramming/llms.txt

How to use with your AI Assistant

Point your agent to https://mindfusion.dev/docs/java-swing/diagramming/llms.txt, or download a local copy for it to read from the project's repository. If you are working with smaller context-window size, you might have to split the file into sections or copy portions from it into individual prompts.

For initial integration of the toolkit into your project, loading the MinApp, Controls, and/or Tutorial sample projects into the assistant's context should help too.

The llms.txt structure

We have divided llms.txt into two main sections, corresponding to the Overview and Reference parts of this documentation:

• High-Level Guides and Overviews (Inlined) - Contains the full content of all conceptual overviews, tutorials, and architectural guidelines in a single, continuous stream. This allows the AI to gain a coherent mental model of the library without needing to crawl separate files.
• API Class Reference (Linked) - Lists the consolidated class/type reference documents in the Markdown folder. This enables the agent to dynamically lazy-load individual class references on demand only when you ask about specific classes.

Context-Window performance

For a complete export of this library's documentation at the time of writing this, the primary index file is approximately 500 KB (equivalent to 120,000 to 130,000 tokens). Here is how this scales across popular AI architectures:

• Large-Context Models with 2M tokens: The entire index takes up less than 6% of the context window. The agent can ingest and retain the entire high-level documentation suite for long-duration conversations with virtually zero context decay.
• Standard-Context Models with 200k tokens: Ingesting the index could take up approximately 65% of the context window. For these models, we recommend using individual sections from the inlined Guides / Overview topics for architectural orientation, copy-pasting them on-demand to preserve space for your active code.