Agentic AI Schema Mapping and Geocoding Template (Gemini)

# Agentic Schema Mapping (AI Assistant Schema Mapping)

## Overview

This template highlights the development of an FME Workspace that can be configured as either a Workspace App or Automation tailored for schema mapping and address data standardization from various data providers. These solutions demonstrate how FME can seamlessly handle schema mapping and data preparation for ingestion into downstream processes, using Agentic AI (Gemini) to expedite the process without the need to build a look up table for schema mapping or manually define schema definitions. This works with any AI provider that supports structured outputs.

### Key Features

1. **Schema Mapping Address Data**:

- Accepts incoming files with varied schemas.

- Converts the data to CSV for prompt engineering.

- Submits the data to Gemini for schema mapping.

- Outputs results in:

- **CSV** for tabular data.

2. **Automation**:

- Watches a designated directory for incoming files.

- Processes files automatically and writes outputs to another directory.

- Enables a fully hands-off workflow for schema mapping and data preparation.

## Sample File Scenarios

The workspace apps are equipped to handle varying schemas. Below are the provided sample schemas and the standardized schema output.

### Input Schemas:

- **Schema 1**:

- `streetAddress, locality, administrativeArea, country, postalCode`

- **Schema 2**:

- `Street, City, State`

- **Schema 3**:

- `Address`

### Standardized Output Schema:

The input schemas are transformed into the following standardized schema:

- `streetAddress, locality, administrativeArea, country, postalCode`

This schema ensures compatibility for downstream validation and transformation processes, such as:

- Duplicate filtering.

- Filling in missing ZIP codes.

- Additional custom transformations as needed.

## Use Case: Geocoding

This workflow serves as a robust use case for geocoding by ensuring that address data is consistently mapped and ready for spatial or tabular processing, reducing manual intervention and increasing efficiency with the help of AI.

## Availability

This functionality is currently available in **FME 2024.2**. Three sample files have been included within the workspace app descriptions for demonstration purposes.