Agent Data Analyzer

Step-by-step technical guide for Agent Data Analyzer

Login

  1. Go to the Web URL:
  2. Enter your assigned User and Password.
  3. If you are unable to login, ask for help from Manhattan Team.

Initialize Agent

  1. Launch Agent Foundry from the Hamburger Menu.
  2. Click on “Create Agent from Scratch”.
  3. Go to Agent Configuration by clicking the “Gear” icon, next to the Agent Name.
  4. Change the AgentId and Agent Name as you see fit.
  5. Scroll down to set the parameters “Conversational” and “Allow Attachment from Chat” flag as provided in the Agent Details section.

Important Note

  • To add a new Task, click the default task dropdown, then click on Add New.
  • To add an Action, click on the + sign on the Initial Screen or at the end of the current action on the screen.

Agent Details

FieldValue
NameMO26-Agent-DataAnalyzer
DescriptionTakes in a CSV and lets the user ask questions about it.
Conversationaltrue
allowAttachmentFromChattrue

Screenshot

Actions

  • Add an action by clicking plus.

Task: default

Action 1 — createResource (createResource)

Action Details:

  • source: objectstore
  • mimeType: text/plain
  • namespace: {:Attachment_From_Chat.0.FileNamespace}
  • filePath: {:Attachment_From_Chat.0.FileLocation}
  • fileName: {:Attachment_From_Chat.0.FileName}
  • outputVariableName: InputFileResource

Action1

Action 2 — readFile (readFile)

Action Details:

  • target: table
  • fileName: InputFileResource
  • outputVariableName: InputFileData

Action2

Action 3 — storeInLocalDB (storeInLocalDB)

Action Details:

  • inputVariableName: InputFileData
  • outputVariableName: InputFileData

Action3

Action 4 — Plan & Execute (callGenAIService)

Action Details:

  • Instruction:
    <PURPOSE>
    The user has uploaded the CSV file: "{:Attachment_From_Chat.0.FileName}".
    Your goal is to answer any questions they may have around this file and its content.
    The content is stored in a local DB table called: "InputFileData".
    </PURPOSE>
    
    <RULES>
    1. Anytime you are referencing the data in the CSV, you should return a segment of the CSV in tabular format.
    </RULES>
    
  • User Input: {:userInput}
  • Parameters:
    • builtInTools: queryLocalDB

Action4

Task: greetingTask

Action 1 — addTextResponse (addTextResponse)

Action Details:

  • textResponse: Please upload a CSV file.

Greeting