> ## Documentation Index
> Fetch the complete documentation index at: https://multiwoven-74-feature--update-docker-compose-docs.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Models

> AI/ML Models in AI Squared allows users to define how to gather the required data for input variables, for the AI/ML models they connect to, using AI/ML Source.

<img className="block" src="https://res.cloudinary.com/dspflukeu/image/upload/f_auto,q_auto/v1/DevRel/models" alt="Hero Light" />

## AI/ML Model Creation

Users can select an existing **AI/ML Source Connector** and use predefined **harvesting strategies** to dynamically populate input variables. This harvesting process is especially useful for utilizing **Data Apps** through no-code integration. After the AI/ML Model is created, users can develop visualizations and create **Model Cards** via Data Apps for further insights and analysis.

***

## Harvesting

Harvesting retrieves input parameters from business tools, essential for real-time machine learning model execution. Currently, we support harvesting input variables from sources such as **CRM systems** (e.g., Salesforce, Dynamics 365) and **custom web applications**.

#### Harvesting Strategies:

* **DOM (Document Object Model) Element Extraction**: Harvest data from specific web page elements.
* **Query Parameter Extraction**: Extract data from URL query parameters.

#### Data Types Supported:

* **Text**: Such as Customer names, product descriptions, etc.
* **Images**: Used for image recognition.
* **Dynamic Variables**: Values that change with user interactions.

### Integration with Model Creation

Harvesting is integrated into the model creation process, allowing users to define what data should be collected, ensuring real-time processing during model invocations.

***

## Preprocessing

**Preprocessing** is an important step that occurs after data harvesting and before model inference. It transforms and formats the harvested data to meet the specific input requirements of the machine learning model.

### Key Preprocessing Tasks:

* **Parsing Specific Attributes**: Extract and format the required data fields.
* **Resizing Images**: Ensure images are resized appropriately for image processing models.

These preprocessing steps ensure the data is properly prepared, optimizing accuracy and efficiency during real-time model inference.

***