How to Set Up and Use AI Retrieval?

The AI Retrieval node lets your AI Agent pull relevant information from a knowledge base whenever a user asks a question. It finds and retrieves the most relevant training data, ensuring the response is accurate and contextually appropriate.

How to Setup AI Retrieval?

Follow these steps to set up AI Retrieval:

Step 1: Access AI Retrieval

1. Open the Studio, Go to your Scenario.

2. From AI category, add AI retrieval node.

3. Select the AI Retrieval block.

Step 2: Enter a Search Query

• Input the search query that the AI Agent should use to fetch relevant knowledge base content.

• The AI Agent will retrieve information based on this query.

Step 3: Filter Content by Tags

You can create and add tags to your training data from the training section.

• Use Include or Exclude to refine search results:

  • Include: Matches any document that contains at least one of the specified tags.

  • Exclude: Retrieves all documents except those with the specified tags.

• Click Add Tags to specify filters for more precise retrieval.

Step 4: Set Nodes Limit

• The Nodes Limit slider controls how many pieces of information are retrieved.

• Adjust this setting to balance response depth and efficiency.

Step 5: Choose Response Type

• Raw: Retrieves direct text responses from the knowledge base.

• JSON: Converts raw data into a structured JSON format based on the provided schema.

Step 6: Store Retrieved Data

• Select a variable to store the retrieved information for further processing.

Step 6: Testing AI Retrieval

• Use the Emulator to verify how the AI Agent retrieves and presents responses.

• Adjust search queries, tag filters, and node limits as needed to refine results.

Example: AI Retrieval in a Food Ordering System

Scenario:

A restaurant uses an AI Agent to assist customers with their food orders. The AI Agent retrieves answers from a knowledge base containing menu details, ingredients, and special offers.

Example Implementation:

1. Search Query: A customer asks, "What are the vegetarian options available?"

2. Tag Filtering:

   • Include: vegetarian, vegan, plant-based

   • Exclude: non-vegetarian, seafood

3. Nodes Limit: The AI Agent retrieves up to 5 relevant dishes.

4. Response Type:

   • Raw: Lists vegetarian dishes with descriptions.

   • JSON: Structures the response, categorizing dishes based on availability.

Final Response:

The AI Agent responds with:

"Here are some vegetarian options available today: Paneer Butter Masala, Veg Biryani, and Mushroom Pasta. Let me know if you need more details."