Project:

I was approached by a Platinum FileMaker Partner who was having difficulty with an AI enabled reporting project. The system was not scaling to handle millions of records as needed. Additionally they had trouble with the LLM interactions with FileMaker.

Project Overview:

Built in Python and using DuckDB as a reporting database, this system allows users to enter natural language queries about key aspects of company performance. The system uses a local LLM to process the queries from natural language to SQL and then runs the queries, returning the results to the user. There is an option for either a FileMaker or web interface, allowing for maximum flexibility. The quality of query results is tracked, thus enabling this system to be useful in a discovery style approach where new questions about company data can be re-used easily.

With open-source components the system is highly cost effective as no additional user license costs are incurred no matter how many people access the data.

Project Benefits:

• Scale: System scales to millions of records with no visible performance degradation.
• Human language querying: The system accepts natural language queries. Ranking the quality of query results enables high quality queries to be saved and performed more quickly.
• Improve Client satisfaction: Ensures the client is happy with their FileMaker solution and doesn’t need to turn elsewhere for advanced reporting
• Cost-effective: No third party licensing costs