Week 2

Milestones

• Create cQube specific prompt to fetch details like related table and prompt to get sql query based
• Create module to get subset of schema from the sql file
• Integrate cQube datasets
• A standalone application (chat interface) that connects to the cQube that allows for querying of data from cQube and returns the response as a table/big number/chart depending on user request.
• An autocomplete to suggest users sample queries that they can do.

Screenshots / Videos

Video

• cQubeChat Demo

Contributions

• Project Plan

• cQubeChat Implementation details

• Add autocomplete feature

• Onboard cQube database in Text2SQL

• Create a prompt which LLM can use to identify dimension columns, metrics and time dimension for a query

• Create a module to figure out the relevant schema

• Allow user to provide chart type in the NLQ and render a chart for the data accordingly

• Added project files for chatbot

Learnings

• ChatGPT 3 and 3.5 model has a small token limit as compared to ChatGPT 4.0.
• For chatbot the temperature of the models needs to be set in a way such that the results are deterministic.
• The prompt engineering determines how good/bad the result produced by chatGpt will be.