Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 10 Current »

Demo Best Practices

The demo covers the following three features of the PrivateAI:

1- AI Knowledge Chatbot (RAG)

2- Data Analysis

3- Semantic search

AI Knowledge Chatbot (RAG)

Show the customer that you can upload and connect a variety of different data sources, including: Documents, Sites, email, Micorsoft Onedrive and Sharepoint, Google Drive and more.

Add a Collection

Show them that you can also create a collection of documents.

Steps to Create a Collection:

  1. Login to the Private AI dashboard.

  2. Navigate to Collections.

  3. Select +New.

  4. Name the collection (e.g., "Medical Records Demo").

  5. Download the zip file here. Extract all the files and upload them to the new collection of documents.

Do not upload the zip file, it will not be ingested.

Ask Questions on the Collection

Select the Collection and click on Chatbot.

image-20241117-080747.png

Below are example questions you can use.

  • Basic Query Examples:

    • "What medications were prescribed to John Doe after his myocardial infarction?"

    • "What are the recommendations for Jane Doe's anxiety management?"

  • Advanced Queries:

    • "Compare the discharge instructions for John Doe's appendectomy and myocardial infarction."
      Note: The answer requires analytical reasoning, not just finding the answer in a document.

    • "List all medications prescribed to John David Smith for hypertension."
      Note: The answer requires understanding which medications are related to hypertension and which are not

Point out that the data source that the answer is based on is available under the answer.

image-20241117-115340.png

Also point out that the Chatbot can be congigured to get answers based on both Public and Private data. The user can choose to sleect only public or Private data.

image-20241117-115938.png

Data Analysis

Connecting to the Chinook Database

The Chinook database demonstrates Private AI's ability to interact with structured data.

Steps to Connect:

  1. In the Private AI dashboard, navigate to Data Sources.

  2. Select Database.

  3. Select +New

  4. Name of Database: Chinook Database

  5. Usage should be set to: For analysis

  6. Use the following Connection String : Driver={ODBC Driver 17 for SQL Server};Server=172.31.10.251 ,1433;Database=Chinook_AutoIncrement;Uid=bganalysis;Pwd=dev*112233;MultipleActiveResultSets=True;App=BGService; Select Check, and you will see the tables of the database below.

  7. Select All the Tables and Save.

To view the table in the dashboard - select the Chinuk database and click “Show content.”

To understand the structure please review the following ERD:

ChinukDBSchema-20241117-111703.png

Ask Questions on the Database

Select the Database and click on Data Analysis.

image-20241117-081907.png

Below are example questions you can use.

  • What are the top 5 customers by purchase amount?

  • List all tracks purchased in 2023.

    • Follow-up question: create a chart from the top 7 tracks by revenue

  • What is the most purchased album in the database?

  • What are the top three genres with the highest revenue generated, and which artists contributed most to those revenues?

  • For each customer, determine the average number of days between their purchases and identify those with irregular purchasing patterns (standard deviation of days between purchases).

Semantic Search

Semantic search enables intuitive, meaning-based searches across unstructured text.

Steps to Perform a Semantic Search

  1. Navigate to the Collections section in the dashboard.

  2. Select the collection created earlier ("Medical Records Demo") and select “Semantic Search

image-20241117-095356.png

  1. Use the search bar to perform queries.

Query Examples

  • "Show records mentioning mental health issues."

  • "Which patients have a history of hypertension?"

Lite Mode

You can create a chatbot and share it so that the end user will see it in lite mode.

Lite mode is the non-admin view. The user will not have access to auditing and the firewall.

They will also only see the chats that have been shared with them.

Sharing a chat to Lite Mode

Go to a data source and click on Share Chatbot.

You then have an option to share as Lite Mode

A LiteMode chatbot requires authentication and is only relevant for internal employees. Permissions typically restrict the chat sources. The LiteMode chats are available on the BusinessGPT site or mobile app.

image-20241117-101902.png

Lite Mode View

You can view the lite mode as admin by going to account settings >> Lite Mode >> View Lite Mode, or sign

in with a non-account manager

Any user that is part of the group specified in the Lite Mode group will see lite mode.

The account admin always logs into the regular dashboard.

image-20241117-112830.png

Lite Mode looks like this. It is a simplified BusinessGPT app. The end user has the option to ask on company data that has been shared to them and public data.

image-20241117-102106.png

Show the result in mobile using F12 of the Chrome an chat

image-20241117-112949.png

Show the mobile menu:

image-20241117-113024.png

  • No labels