Application

SWE applications enable users to create end-to-end generative AI-based apps that can leverage their organization's data in various formats and sources to create fully managed, scalable deployable applications within minutes. SWE apps can be embedded directly as UI components in iframes or can be used only as the backend engine through the SWE API and SDK. SWE apps empower users with basic technical capabilities to build custom generative AI apps.

Application Types:

AI Assistant Retrieval:

A chain-based application that allows users to connect to a single data source and quickly receive responses specific to the context.

Advanced Agent:

An agent-based application that enables users to connect to multiple tools and perform complex tasks. Users can ask questions related to a specific tool or general inquiries

Application main components:

Model:

Users can select their preferred Large Language Model (LLM) to power the app. SWE provides out-of-the-box options to connect to OpenAI and Google Vertex models, including multiple version support. Users can also register their own private LLM models.

Prompt:

The prompt serves as the interface for app creation, allowing customization of how the app should act and react. Using prompt engineering techniques, SWE users can make the app engine respond in a specific format or tone or perform a particular instruction to accomplish a task.

Tool:

Tools extend SWE 'Advanced Agent' apps to perform more complex operations beyond natural interaction with end users. Tools are essentially functions that the configured LLM model can use to perform an action. The current supported type of tool is a DB tool, and VectorDB tool enabling the LLM model to run queries against databases as necessary to collect relevant information according to the end user's request.

Context

Context allows you to connect your 'AI Assistant Retrieval' application to a single source of data. The context can be an SQL DB or VectorDB and Knowledge(coming soon)