Introduction

This section of the documentation covers the usage of the AI Support appjar from an end user perspective. The appjar provides two distinct levels of access: regular users interact with the AI-powered chat interface, while administrators manage the configuration that drives the AI behavior, including assistants, language models, documents, prompts, and integrations.

The AI Support appjar is designed around a conversational model. Users ask questions or request assistance through a chat interface, and the system responds using a configured AI assistant backed by a large language model (LLM). The quality and relevance of those responses depend on the documents, prompts, and model configurations that administrators maintain behind the scenes.

How AI Support Works

At the core of the appjar is the concept of an Assistant. An assistant is a configured combination of a language model, a system prompt, and a set of document categories. When a user starts a conversation, the assistant uses Retrieval-Augmented Generation (RAG) to search through the relevant documents, find contextual information, and generate a response grounded in that knowledge base.

A typical workflow looks like this: an administrator creates one or more LLM connections, uploads documents that the AI should reference, writes prompts that guide the AI behavior, and assembles these components into an assistant. Users then interact with that assistant through the chat interface without needing to understand the underlying configuration.

User Features

Regular users access the Chat view, where they can start new conversations, continue existing ones, and review past sessions. The chat supports rich text messages, file attachments, user mentions, and real-time streaming responses from the AI. Users can also rate conversations after they are closed, providing feedback that administrators can review.

For more details, see User Features.

Administrator Features

Administrators have access to a comprehensive set of management views that control every aspect of the AI Support system. These include managing assistants, language models, documents, prompts, categories, and channel integrations (WhatsApp, Facebook, and Instagram). Administrators can also use the LLM Inspector to review the detailed exchange between the system and the language model for debugging and quality assurance purposes.

For more details, see Administrator Features.