> ## Documentation Index
> Fetch the complete documentation index at: https://rag-docs.peakfs.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> How we utilize RAG at Peak Innovations

<img src="https://mintcdn.com/peak/S89t8ZUZoKiNypat/images/rag-framework.jpg?fit=max&auto=format&n=S89t8ZUZoKiNypat&q=85&s=d34ca3c000414a395086a1633a75c64c" alt="RAG Framework" width="3840" height="2160" data-path="images/rag-framework.jpg" />

## Brief Overview

The RAG (Retrieval-Augmented Generation) framework at Peak Innovations is a versatile AI system that processes complex queries by dynamically selecting agents, breaking down tasks, and retrieving information from multiple sources.

Requests can range from real-time user queries to backend tasks, like generating customized learning materials.

## How it works

<Steps>
  <Step title="AI Agent Routing">
    The system directs each request to the most appropriate AI agent, configured
    to handle specific tasks such as financial analysis, knowledge retrieval, or
    personalized learning.
  </Step>

  <Step title="Dynamic Task Pipelines">
    The chosen agent break down the task and constructs a sequence of subtasks,
    forming a dynamic pipeline. These subtasks are linked to specific tools that
    retrieve and process data from sources like APIs, databases, and vectorized
    content.
  </Step>

  <Step title="Tool Utilization">
    Each tool processes its assigned subtask, using relevant data sources to
    deliver specific outputs. Tools range from language models for response
    generation to database access tools for retrieving structured data.
  </Step>

  <Step title="Synthesis and Response">
    Once all subtasks are completed, the framework synthesizes the results into
    a coherent, contextually accurate response, leveraging the combined data to
    deliver precise answers tailored to the user’s needs.
  </Step>
</Steps>

This flexible, modular approach enables the RAG framework to handle diverse tasks with precision, adapting seamlessly to the unique requirements of each application.