DataConnect

Inspiration

DataConnect focuses on an issue facing many businesses: the data that they need for everyday use is held in a wide array of spreadsheets and other forms, and there is no alternative to sifting through these spreadsheets to find the information that they need. Upgrading to some sort of SQL database to store information is a possibility, but this would require a lot of effort and technical knowledge to convert their data into this format. DataConnect is the solution to this issue.

What it does

DataConnect uses Agentic AI to parse through and create relational databases of information from the wide array of spreadsheets and CSVs that businesses might have, mimicking a human who will be able to see the relations between different spreadsheets. The agent can then turn into a domain expert on the data present in this new database, allowing individuals to chat with the data and ask questions like "how many lots have been on hold for 7+ days" or "where are marketing opportunities that our business isn't focusing on enough" which would have previously required painstakingly searching through multiple spreadsheets and numerous headaches.

DataConnect determines its own understanding and schema towards the databases, and continually improves its ability to make connections between different spreadsheets. It provides the users with complete control and understanding over their data, without any painstaking process to transfer their data to a more usable platform.

Sponsor Technologies Used

Lovable for frontend creation, Gemini for agent creation, Composio for agentic access to Google Drive and Sheets, and Render for SQL database,

How we built it

DataConnect uses mostly Python for its backend and TypeScript for its frontend. The frontend (which was built in Lovable) communicates directly with Composio first, in order to find the necessary files that are to be parsed, and then Composio provides Gemini with the same files to create connections between them and extract necessary insights, which are then routed and displayed in a dashboard format on the front-end while being stored in a Render database. Following this, the user can interact with the Gemini API to "talk" with their data.

What's next for DataConnect

Spreadsheets aren't the only example of "messy" data without a proper schema to allow for parsing, and the team at DataConnect is ready to tackle any sort of problematic examples of data lakes and create parseable versions of data that are actually usable and automatable for organizations.

Built With

• composio
• gemini
• lovable
• render