How to build a contextual AI agent in Jitterbit Harmony

Introduction

This guide shows how to add memory and context to a reactive AI agent in Jitterbit Harmony using Integration Studio. This agent stores conversation history and maintains context across multiple interactions.

Important

These steps are a continuation from the how-to guide on building a reactive agent.

Tip

For learning purposes, reference the Contextual Agent provided through Jitterbit Marketplace for an implementation of this guide.

Build a contextual AI agent

Open the Integration Studio project that contains your reactive AI agent.
Add a datastore, such as Jitterbit Cloud Datastore or equivalent. Create two key storages:
- Bot_Authorized_Users: To manage authorized sessions.
- Agent_Q_and_A: To store conversation history for context.
Update the main AI logic:
1. Open the workflow Main - AI Agent Tools Logic.
2. Click a new operation drop zone and select New Script. Configure the script to fetch conversation history for the user from the datastore. Include previous questions/answers in the LLM prompt to maintain context. See the script named Main-AI Logic in the Contextual Agent for an example.
Add utility workflows. For each:
1. Click Add New Workflow along the top of the design canvas to create a new workflow.
2. Double-click the default workflow name and rename it. Create two workflows:
  - Utility - Read from Datastore: Fetches conversation history.
  - Utility - Write to Datastore: Saves new messages and responses.
  See the workflow named Utility - Bot Chat Cloud Datastore in the Contextual Agent for an example.
Ensure workflows have stateless entry:
- Slack or API entry workflows only convert user input to internal payload.
- Main logic handles memory retrieval and updates.

Next steps

To build upon the AI agent with memory and context, see Build an AI agent with RAG.