About the Gen AI Tool

The Savant Gen AI Tool, currently in beta, offers generative artificial intelligence capabilities within your Savant analysis. This tool leverages the power of Large Language Models (LLMs) to enhance your data analysis process.

Adding a Gen AI Tool

Savant supports LLMs such as OpenAI ChatGPT and Google Gemini.

To integrate the Gen AI Tool into your analysis, follow these steps:

1. Configure Your OpenAI Account:

Initially, the Generic API Tool includes a default Savant AI account for minimal testing. However, for more robust usage, configure your own LLM service provider. You have two options: OpenAI and Azure.

1. Navigate to the Systems page.

2. Click "New System."

3. Search for OpenAI and select it.

4. Choose the environment (OpenAI or Azure).

5. Provide the API token from the selected environment and click Authenticate.

6. Rename and describe your OpenAI provider.

7. Confirm the setup.

Please contact us to join our private beta for Google Gemini.

2. Adding the Gen AI Tool:

1. Go to your analysis.

2. Click this icon.

3. Click Gen AI.

Configuration

Setting up the Gen AI Tool involves these configurations:

Selecting the Provider: Choose your LLM service provider. For instance, you can select the Savant AI provider to try the tool on (with reduced throughput).
Defining the Prompt: Craft a well-structured prompt that outlines your desired analysis. You can include specific instructions for the LLM provider to generate relevant insights from your data. For example:
1. Prompt: “What country is this city found in? Return just the name of the country with no other words. If you cannot guess the country, return nothing.”
2. Field to Transform: BillingCity
After defining your prompt, click "Apply" to finalize the setup.

Output

Upon applying the configuration, you'll notice the following changes to your dataset:

In development mode, Savant will calculate 5 new records every time you click "Apply".
- Each you click Apply multiple times, Savant we calculate 5 additional records until 1000 records have calculated with that prompt and data.
- If you run the entire bot and calculate on all of your data, Savant will show 1000 records from that run in your development mode preview.
- This is designed to conserve your LLM service credit usage.
The AI Answer field is added to your dataset with the response from the LLM service.

Referencing Columns in Streaming Mode

In the advanced configuration, there is an option to use streaming mode. This is designed for very small data bots - typically less than 100 rows - that want faster run times.

This mode also gives you the option to inject one or many fields dynamically into your prompt. To do this, simply write ${field name} any where in your prompt.

For a record where City = "Paris" and Country = "Japan", this prompt for data validation "Is ${City} in {$Country}?" would ask the LLM service "Is Paris in Japan"?

Importance of Adding Specific Instructions

To maximize the accuracy and relevance of the LLM service's generated insights, provide clear and specific instructions. By being explicit in your prompt, you can direct the LLM provider to return precise information. Avoid ambiguity and opt for pinpoint directives, ensuring the generated responses align with your analytical needs.