Create an Agent: Quickstart

Agent Quickstart: Build Your First Intelligent Agent

Overview

Agents in Dataworkz act as orchestrators of data and AI—allowing you to connect and retrieve insights from a wide variety of sources such as RAG applications, SQL databases, JDBC, MongoDB, GraphQL, and REST APIs. Each data source or interaction is enabled via "tools" the agent uses, powered by LLMs.

This guide walks you through creating an Agent that combines multiple data sources using AI capabilities.

---

Use Case Example: Unified Customer Insights Agent

Goal: Build an agent that can:

• Retrieve FAQs from PDF documents via a RAG app.

• Extract product specs from Word documents via another RAG app with a custom prompt.

• Query customer purchase history from a Snowflake table.

• Access support ticket data via a REST API.

The agent will use these tools to provide a unified response to customer inquiries—combining structured and unstructured data sources.

---

Step-by-Step: Creating an Agent

1. Create the Agent

1. Navigate to the Agents screen in Dataworkz.

2. Click "Create Agent".

3. Give your agent a name and optional description in the chat box. Use this space to describe the agent’s role or logic (e.g., “Fetch customer insights across PDFs, Word docs, Snowflake, and REST APIs”).

4. Click Create.

---

2. Define Agent Behavior

Each agent in Dataworkz is structured around three key sections that control how it operates:

• Description: A brief overview of what the agent is for.

• Planning Instructions: Written in plain English, this section defines how the agent should use the tools you assign to it.

• Instructions: Also written in plain English, this section tells the agent how to format or present the output.

Use these sections to guide the agent’s logic and ensure consistent responses.

Example configuration:

• Description: Support assistant agent that provides insights from documents, databases, and support APIs.

• Planning Instructions: When asked about a product or customer, first check relevant RAG apps (FAQs and product specs). Then query the customer’s purchase history from Snowflake and retrieve any related support tickets using the REST API. Combine these data points to answer the query.

• Instructions: Respond in complete sentences with a clear summary. Reference documents when needed and list any relevant customer data in a concise format. Use bullet points if returning multiple items.

---

3. Add Tools to the Agent

Now define the tools your agent will use.

RAG Tool – FAQs from PDFs

1. Create a RAG App (via “Create with Default Settings”).

2. Upload your PDF file with FAQs.

3. Configure the LLM and prompt as needed.

4. Add this RAG App as a tool to the agent.

RAG Tool – Product Specs from Word Docs

1. Repeat the RAG creation steps for the Word documents.

2. Customize the prompt for extracting structured specs.

3. Add this RAG App as another tool to the agent.

SQL Tool – Snowflake Purchase History

1. Create a RelationalDB Tool for Snowflake.

2. Test your query (e.g., SELECT * FROM purchase_history WHERE customer_id LIKE '${UserInput}').

3. Add this SQL tool to your agent.

API Tool – Support Tickets

1. Add a REST API tool with your endpoint (e.g., https://api.mycompany.com/support-tickets).

2. Define any headers or auth tokens needed.

3. Add this as the final tool to your agent.

---

4. Test the Agent

1. Go to the agent’s chat interface.

2. Ask a sample question, such as:

   “What issues has customer Jane Doe reported recently, and what product did she purchase?”

3. The agent should respond by:

   • Extracting product data from the Word doc.

   • Checking Jane’s purchases in Snowflake.

   • Summarizing support tickets from the REST API.

   • Referencing FAQs if needed.

---

Final Notes

• You can continuously improve your agent’s logic by updating its instructions.

• Tools can be reused across different agents.

• Use prompt engineering to fine-tune how each tool behaves with the LLM.