GenGraphy

Inspiration

The idea was to democratize the art of data storytelling and make it easier for organizations to get though their day to dat ad-hoc analysis with as least amount of hassle as possible.

What it does

To put it simply, it allows you to tell it what to do. Say you want to find how many of your vendors are in Texas. You just type our ask out on the friednly user interface. Then the LLM generates your queries, puts it through the database tp providesyou with an interactive and intuitive dashbaord filled with graphs ranging from simple bar charts to complex regional heatmaps, time-based or sequential data representations. The chatbot allows for a wide range of asks and greater flexibiltiy with filters than that of a typical PowerBI dashbaord. It is also speeds up workflows in a corporate setting.

How we built it

First , we built the chatbot functionality. Given our approach of trying to generate SQL queries to translate the asks of the user, we had to implement LLMs to enable natural language understanding.

The 2 models we experimented with for LLMs

SQL Query finetuned LLM - Defog SQLcoder - performs better than gpt 4 for sql queries
Google Gemini After that, we intergrated plotly graphs to make it more interactive . Then we built the web application using streamlit to bring our chatbot and dashboard together.

Challenges we ran into

Working with open source hugging face models. Making the dashboard flexible enough to accommodate a wide array of requests.

Accomplishments that we're proud of

Finishing our minimum viable product within time. Picking up streamlit on the go.

What we learned

Our journey with GenGraphy provided invaluable lessons in Streamlit utilization and Hugging Face model training, enriching our skill set and knowledge base.

What's next for GenGraphy

The future of GenGraphy holds exciting possibilities. We aim to enhance flexibility, enabling seamless transition into sales analytics dashboards, P&L statements, and beyond. Incorporating more robust LLM models will further elevate query/report accuracy, ensuring GenGraphy remains at the forefront of data-driven decision-making.