​
Brightbot Interactive Analysis

Brightbot’s high level architecture workflow is as follows:

The image illustrates the “Brightbot Interactive Analysis” system’s architecture, showing the data flow and interaction between its components.

​
Query

The query is in the form of a natural language question posed by the user. It serves as the starting point for the entire process.

​
Brightbot API

Provides an interface for interacting with the system, allowing users to send queries and receive responses.

​
Supervisor Agent

Responsible for parsing the query and passing the request to the appropriate agent(s).

​
Retrieval Agent

Implements Retrieval Augmented Generation (RAG) to fetch relevant data from the knowledge graph and data catalog.

​
Knowledge Graph

A structured representation of knowledge that refines the user query and feeds into the Prompt + Context component.

​
Data Catalog

Provides metadata about the knowledge content, also feeding into the Prompt + Context component.

​
Prompt + Context

Receives input from the Knowledge Graph, Data Catalog, and Supervisor Agent, and sends it to the appropriate agent.

​
Analytics Agent

Generates a Jupyter notebook identifying key trends in the data and providing insights.

​
Visualization Agent

Generates visualizations in the form of charts.

​
Engineering Agent

Generates DBT code for data cleaning and transformation.

​
Multimodal Output/Response

Depending on the type of query, the system can produce different outputs, including:

  • Jupyter notebooks (from the Analytics Agent)
  • Charts (from the Visualization Agent)
  • DBT code (from the Engineering Agent)
  • Textual responses (from the Retrieval Agent)

The architecture outlines a closed-loop system where a question initiates the process, leading to the generation of analysis through various data sources and processing by the LLM embedded inside the agents. The system includes a quality assurance step through the a review step in the Retrieval Agents, which ensures the output’s accuracy. The Brightbot API orchestrates the flow of data and interactions across the system.