Building a Tabular Data Assistant on AI Studio

In this process, you will learn how to set up a tabular data assistant by adding data sources, configuring databases, and customizing your AI chatbot. These steps include uploading structured files, setting up databases, and deploying AI tools to enhance the user experience. Follow these instructions to successfully create and manage your tabular data assistant.

Step 1

Begin by adding a data source for your tabular data assistant. Navigate to Knowledge Management to add a new knowledge.

Select Add Knowledge under the Knowledge Management panel.

Step 2

Click the plus sign next to Tabular Data to upload structured files or databases.

Enter a descriptive name for your new knowledge source.

To use an internal PostgreSQL database, select the default one from the dropdown menu.

Step 3

Alternatively, to configure and organize an external PostgreSQL database, switch to the Manage Database tab. Please Note that this step is only applicable for Super Admins. If you wish to add one, please ask the super admin to add a external database and share with required groups.

If you have an externally deployed PostgreSQL database, select Add Database to create a new secure database for your team.

Enter the necessary credentials to establish the connection.

Step 4

Once the DB source is decided, either internal or external database, then proceed with uploading your files.

Then allocate the required resources based on the file size.

Step 5

Review existing tabular data to manage or update your uploaded data sources. Ensure that the knowledge has been successfully added.

Step 6

Access the connector details to view and adjust available data connections.

Step 7

Visit Knowledge Objects to manage structured items within your structured dataset. Monitor indexing attempts to track the progress and health of your data imports. Additional documents can be attached from here.

Step 8

Switch to Projects to start building your tabular data assistant. Create a new project and utilize premium AI tools in one seamless workflow.

Choose the private option to keep your project secure and team-focused. Select shared if you wish for multiple collaborators to build your chatbot together.

Step 9

Choose tabular Knowledge source to empower chatbots that understand your structured data. Select the database from the available options.

Step 10

Select the data added earlier through Knowledge Management and proceed with it.

Step 11

Select an appropriate model to run your tabular chatbot. For example: Select gpt-4o for optimal performance. To add a model navigate to LLM Management tab in Global setting to add a model for usage.

Step 12

Proceed to chatbot configuration to personalize your assistant's interactions. Expand advanced settings to unlock deeper customization options. Add custom chat reasoning rules for a unique conversational style.

Step 13

Navigate to prompt personalization to tailor your AI's responses with custom prompts. Explore the variety of templates available under the Prompt Personalization tab.

Step 14

Add or update system prompts to guide your chatbot's core behavior. Expand the content filter prompt for precise control over conversation content. Add a conversation name prompt for clear and relevant chat session labels.

Step 15

Edit follow-up question prompts to enhance engagement and user experience. Establish answer validation prompts to ensure the accuracy of chatbot replies. Configure organization policy prompts for compliance.

Step 16

Open the application card to personalize your chatbot project's appearance and details. Enter or update the application name for a strong first impression.

Step 17

Decide whether to keep your application public, private, or share it with collaborators. Move to app brand customization to tailor branding elements, bot personality, and user interaction experience.

Step 18

Click Next and select the most suitable resource size for your project.

Adjust the temperature setting to enhance response creativity.Click Deploy to launch the application.

Step 19

Once running, view its logs by clicking on Logs button.

Click Usage to access detailed insights and efficiently manage your AI project data.

Step 20

Open the API panel to access endpoints and integration options.

Step 20-22 are optional steps, they can be followed if API integration of the chatbot is required.

Step 21

Utilize the provided API URL links with specific endpoints. Click the link to access the model prediction API endpoint.

Generate your API token and set its expiry date.

Step 22

The token will be needed to run the deployed API. Test API endpoints directly from the documentation by clicking on View Documentation button.

Visit the documentation page to use your new API token for endpoint testing. Start by clicking on Authorise button to Enter the token. Here you can test the endpoints seamlessly before integration to production.

Step 23

Navigate to Projects screen to start testing the chatbot.

Step 24

Test your chatbot instantly by clicking Try This Chatbot.

Once the application window opens, type your request or question into the message box. For Example, if the knowledge selected contains sales transactions with details like product, quantity, and closing price, then ask a related query to the knowledge.

Step 25

Query the chatbot and visualize your results by selecting the chart icon.

Download your generated chart.Click the download button to save instantly.

Step 26

Click copy button to copy results.

Use the left arrow option to ask a follow-up question.

Step 27

Click Stop to complete your session once you have finished using the application.