What is an Assistant?

An Assistant (also called Agent) is an intelligent software component designed to autonomously perform tasks and provide assistance by interacting with users and systems. Leveraging artificial intelligence and machine learning, assistants can understand user input, process information, and execute actions based on predefined rules or contextual understanding.

Assistants serve as a bridge between users and complex systems, enabling streamlined interactions, task automation, and personalized responses. They are highly adaptable and can operate in a variety of environments, such as customer service, business process management, or information retrieval.

What is an Assistant Composed Of?

An Assistant is composed of the following components:

1. Prompt

The prompt is the starting input that defines how the assistant behaves. It sets the personality, goals, and boundaries for every interaction.

Key elements:

• Instructions – how the assistant should operate.
• Context – background information for understanding the task.
• Constraints – rules and limits the assistant must follow.

2. Knowledge Base

The knowledge base (KB) is where your assistant’s factual grounding lives. It contains structured or unstructured data such as product sheets, documentation, and FAQs.

Indexation ensures this data is ingested, split, and searchable, so responses are reliable and consistent.

3. Large Language Models (LLMs)

The LLMs are the engines of reasoning in the pipeline. One or more models may be used depending on your setup. They interpret prompts, combine them with KB results, and generate natural language outputs.

• Core role: provide language understanding and generation.
• Flexibility: multiple models can be orchestrated (e.g., one optimized for retrieval, another for summarization or reasoning).
• Link: [See pipeline architecture →]

4. Tools / Function Calling

Assistants use tools and function calling to go beyond text generation. Functions let the assistant interact with external systems and trigger workflows.

Typical uses:

• Fetch data from APIs or business apps.
• Perform operations like sending emails or updating records.
• Enrich conversations with real-time or domain-specific information.

This makes the assistant dynamic, actionable, and integrated into your processes.