Easy up-skill: feed your LLM of choice your FileMaker DDL for a quick context upgrade.
You’re probably already using a coding assistant of some sort. Tools and preferences are shifting quickly, but I want to share a quick boost I’ve been using lately to great effect. I think it could be helpful regardless of your preferred AI tooling. From playing with FileMaker’s AI functions, I know it uses the GetTableDDL function to tell LLMs about the FileMaker architecture, so Perform SQL Query By Natural Language can work.
What is DDL?
DDL stands for Data Definition Language. It’s a subset of SQL used to create, modify, or delete database objects like tables, indexes, and users.
I fiddled with that function in the data viewer until I could extract a DDL of my solution. Here’s what worked (because it wants a JSON array of table occurrences):
GetTableDDL ( JSONFormatElements (JSONMakeArray ( TableNames ( “My Solution Name” ) ; "" ; "string" )) ; 1 )
I copied the result into a new text file, saved it as My Solution Name.ddl, and placed that in the project folder I’m using with my IDE (VS Code, connected to Claude Sonnet via Amazon Bedrock for security).
I asked Claude to create a project brief for this project, and holy patoots it was pretty spot-on. (I of course reviewed it and made necessary adjustments and additions!)
Now I’m able to get to brass tacks right away planning and building an HTML & JavaScript-based dashboard that’s going to live in a web viewer. For any future conversation or code work for this client, the DDL will give continue to give my LLM the context to Guess Less, Determine More! (Who’s making that t-shirt?)
Other Thoughts on DDL:
A DDL just describes your schema. A super-thorough description of your application would come from using Save As XML. However, that file is gigantic, and processing it would require a ton of tokens, time, and environmental impact. With a hope of staying as light as possible, I prefer the DDL. Plus, LLMs already know how to use a DDL. They aren’t as knowledgeable about FileMaker XML files.
Pro tip: You can copy and paste the DDL text into a chat, but you might run out of characters. That’s why I created a text file and placed it within the project folder I’m working in. If you’re a chat-only person, just keep in mind it might take a few pastes to get the full DDL in, and your LLM might not know or tell you that what it could actually process is incomplete.
Also, if you’re using a public LLM, consider whether you’re comfortable sending your architecture to it.
Your FileMaker and AI Development Team
The Soliant team builds custom FileMaker solutions and AI-integrated applications for clients every day. Our approach is part of how we stay sharp and deliver work that fits the systems we’re actually building in. The context-aware workflows, the web viewer dashboards, and the tooling decisions are problems we think through constantly.
Contact us to connect with our FileMaker development team. We’ll take an honest look at your solution, talk through your goals, and tell you exactly where AI tooling can make a real difference for your project.
