Recommendation

The recommendation widget provides automated, conversational analysis of your dataset using the power of AI. This feature helps you understand what's happening in your data without needing to write queries, surfacing plain-English insights directly within your widget.

The analysis updates automatically as your data updates, ensuring the recommendations reflect the most current state of your dataset.

Overview

The Recommendations engine works like a built-in data analyst. It reviews the results of the widget and generates a summary in natural language, highlighting patterns, trends, or anomalies. You can interact with it by entering a custom prompt or question, just like you would with a chat-based AI.

Use simple prompts to get quick answers:

• What are the main trends this month?

• Which regions are underperforming?

Or add more complexity to guide deeper analysis:

• Identify the top 5 contributors to growth and explain why they performed well.

• Summarize patterns in customer churn by segment and region.

How to Use Recommendations

From the widget Settings menu, select the "Recommendation" visualization type.

Enter your prompt in the 'Prompt' field under Display & Formatting and select your model.

Choosing the AI Engine

You have the option to choose which language model powers your recommendations:

• Internal SLM = Knowi’s Small Language Model: A fast, lightweight model optimized for embedded dashboard analysis.

• OpenAI Integration: If you prefer to use a more advanced model, you can connect your OpenAI account by entering your API key in your user settings. Once configured, your prompts and data will be processed through your OpenAI account. (For a limited time, you can use Knowi's OpenAI's account once you enable it. But we recommend generating your own key at https://openai.com/api/.)

To connect your OpenAI key:

1. Go to User Settings

2. Open the AI Settings section

3. Enter your OpenAI API key

Widget Menu