FinnoSQLbot

• 

  Welcome Prompt by the FinnoSQLbot to the user with sample Qs and brief info @ linked the financial data

• 

  Example of SQL generation based on the users selected question

• 

  Partial Code Page example

• 

  SQL + Summarization + Result Table Generation example

• 

  Example of Views created for Database & Schema at the Snowflake Worksheet

Inspiration

My inspiration for FinnoSQLbot came from the desire to democratize access to financial ( or any useful) data exploration. I wanted to create a tool that empowers users, regardless of their SQL expertise, to unlock valuable insights from their data stored in Snowflake.

What it does

FinnoSQLbot acts as your AI assistant for financial data exploration. You can ask questions in plain English, and I will generate a focused SQL query based on your prompt. It can also automatically introduce the reference table and available metrics, and 10 sample questions to get you started, even without a specific question. The responses include the generated SQL query and a clear short summary of the retrieved data, making it easy to understand key financial insights.

How I built it

Building FinnoSQLbot involved integrating Natural Language Processing (NLP) capabilities to understand user questions. I leveraged Snowflake's powerful query engine to execute the generated SQL queries and retrieve relevant data. Additionally, I incorporated the model's data summarization techniques to present the results in a concise and informative way.

Challenges I ran into

One of the challenges I faced was ensuring the accuracy and efficiency of the generated SQL queries. I had to fine-tune the NLP model to handle the nuances of financial data terminology and user intent. Additionally, balancing security with flexibility required careful design to ensure queries operated within the designated table and columns. Other challenges were, creating the databases and schemas for other financial datasets I tried initially because of my less knowledge in sequels and new to snowflake worksheets. Creating a Video with voice presentation in my system. I had to spend last few hours in high tension to record effectively and with my voice over clearly. One tool miserably wasted my time. Uploading the final heavy video ( as Mac generated .MOV with high res ) was pain at final stage. You tube took much longer time than anticipated.

Accomplishments that I'm proud of

I'm proud of creating a user-friendly interface that allows users to explore financial data without needing to write complex SQL code. The ability to automatically introduce the reference table and summarize retrieved data makes FinnoSQLbot valuable for both beginners and experienced users. Even sequel experts also can save valuable time from writing complex SQL commands. Very happy on the fact that I can now utilise Snowflake datasets to create sample apps. But need free trial.

What I've learned

This project has been a journey of wide learning in short time for me. I've deepened my understanding of effectively fine tuning techniques through code, Snowflake's rich datasets and functionalities, and the importance of user-centered design in building an intuitive and helpful AI assistant.

What's next for FinnoSQLbot

The future of FinnoSQLbot is bright! I envision expanding its capabilities to support a wider range of financial data queries and with visualizations. Also will try to add agenticAI techniques further to make it faster, reliable, versatile etc. Additionally, integrating with other business intelligence tools could further enhance its value for data-driven decision making.

Built With

• arctic
• openai
• replicate
• snowflake
• streamlit