dotgov | Devpost

mockup (based off real response)
n8n workflow
curl simulation of query
webhook view in n8n

Inspiration

It is for businesses or individuals interested in keeping up to date on regulations or legislation (courtesy of the federal register), offering insights for business decisions and updates for political enthusiasts/activists.

What it does

It is an agent knowledgeable on recent regulation changes and regulation proposals published on the federal register. It is also able to analyze the sentiment of comments in regulation proposals in order to gauge public feelings, indicating its likelihood of passing.

How we built it

The backend is mostly built with n8n (a workflow automation software) and AWS. Sentiment analysis was made possible with AWS comprehend. The backend's reasoning is powered with RAG, using the openAI turbo 3.5 model. The n8n agentic workflow is hosted on a Tiber VM-SPR-SML instance in a Docker container. Government publications are fetched from https://www.federalregister.gov/ with their APIs.

Challenges we ran into

Using Tiber kubernetes clusters for the first time with Docker containers and networks to create an agentic system was challenging. Using n8n helped the process with n8n nodes, python scripts, and langchain. I also ran into anti-scraping countermeasures the federal register had and had to fall back to working within the scope of their API. Time was a challenge and was not able to connect the backend with the frontend.

Accomplishments that we're proud of

Learning how to use Tiber, n8n, and the federal register's APIs. Experimenting with Intel Tiber Kubernetes cluster with Docker containers was fun too.