FileMaker Pro AI Introduction

The Possibilities Offered By Using AI in FileMaker

• Simulating an AI Query and Return chatbot in the style of ChatGPT, Claude etc
• Perform Find By Natural Language using AI
• Perform Semantic Search using AI
• Perform SQL Queries using AI

Getting Started

• What is an LLM?
• Providers and models supported in FileMaker
• Setting up your development environment
• AI models supported
• Setting up a free account with Cohere
• Configuring the AI account connection

Using Generate Response From Model

• The Generate Response From Model script step
• Defining the Prompt and the Response
• Understanding Temperature
• Asking a question and returning a reponse

Using AI to Analyse FileMaker Data

• Performing AI analysis of internally stored data eg. summarising PDFs
• Using the GetTextFromPDF() function
• Combining the prompt with the PDF
• Working with the Markdown included in the response

Interrogating your own data using Instructions

• Enabling your team to perform FileMaker searches using free-form text
• Gathering data, finding records and looping through to build your summarised raw data
• Using the Instructions option to build the equivalent of a System prompt
• Using free-form search terms to find matching records

Simulating an AI chat interface

• Saving context using messages
• Saving message history
• Controlling memory using the Count option
• Noting the memory history is stored as a JSON array
• Likely use cases for memory-enhanced messaging

Using Perform Find by Natural Language

• The options within the script step 'Perform Find by Natural Language'
• Understanding how it works - sending Schema not actual data to the AI
• Found Set vs. 'Found Set by JSON' vs. 'Find Request as JSON'
• Inspecting the Find Requests generated
• Making requests more intelligent using AI reasoning
• Enhancing requests using Parameters and a Prompt template
• Generating the search results as JSON rather than a found set
• Generating just the Find Requests as JSON rather than performing the search , Layouts, Query, Sort order

Understanding the options within 'Perform SQL Query by Natural Language'

• Using Query
• Using Query for data only
• Using Get SQL
• Using Get First SQL
• Using Get DDL

Perform SQL Query by Natural Language

• Choosing tables: From List, By Name or By DDL
• How it works , prompt sent to AI along with DDL, SQL query returned, Data collected then processed by AI then returned
• Further options with JSON
• Using 'Query for data only'
• Using 'Get SQL' to returns only the SQL needed to retrieve the response
• Using 'Get First SQL'

Using DDL in Perform SQL Query by Natural Language

• What is Data Definition Language?
• Understanding DDL in FileMaker , using the GetTableDDL() function to reveal tables, fields and values entered in Comments
• Using DDL in Perform SQL Query by Natural Language
• Using the [LLM] tag

Perform Semantic Search using AI

• Traditional FileMaker Find mode vs. Semantic Search
• How AI Semantic Search works in FileMaker
• Using vector embeddings to allocate numerical values to words, phrases and images
• Understanding the use of Cosine similarity to define similar, not similar and opposite meanings
• Turning text into vector embeddings
• Using tokens and embeddings
• The trade-off between performance and efficiency
• The principle of using Cosine similarity to measure the angle between two vectors

The Practical Process of Semantic Search

• Performing semantic finds
• Adjusting Cosine similarity
• Advanced semantic find features
• Performing hybrid searches
• Monitoring the process
• Working with token counts
• AI call logging