While the knowledge chatbot is good at generating text, it cannot analyze data.
For analyzing data, BGT offers a data analysis AI agent tool to perform analytical tasks such as searching records, summing up, finding max values, and more.
The way the AI Agents work is by understanding your data and creating a Python action plan to retrieve your data.
It is also equipped with other tools, such as searching the web and using AI to analyze data, like classifying values (to be documented later)
The Data analysis works best on a well-defined DB with friendly, clear names of tables and columns and foreign keys connecting tables.
To get better results, you can add a description on the DB level with any information to assist the agent in answering the questions, such as the connection between tables and explaining specific fields.
In the following versions, BGT will offer the ability to add a description and friendly name for each field and table separately.
Select a DB or table and click the “Data analysis” button to perform data analysis.
Best practices in Sharing chats for end users
When you DB contains several tables related to different topics you can consider one of the following approaches:
General-based chats - having one chatbot connected to all tables - is simple for the user but might confuse the AI
Topic-based chats - having a separate chat for each topic- Less comfortable for the user but better results
Let's take, for example, this list of tables of a music shop
These are very well-designed, with clear names and foreign keys.
However, you could consider having a chat for invoices containing all the finance-related data (invoice/customer employee) and a chat related to music (Tracks/playlist/albums etc)
To implement the topic chats approach go to the list of tables using show content , select therelevant tables and create a shared chat
Then, share it with an appropriate name:
In this case, the end user will see two chats:
For the General chat do the same but selecting the DB level:
