Inspiration

We wanted to build something that could make handling trade documents easier. Going through documents like invoices and Bills of Lading and checking the details manually can take a lot of time. We thought it would be useful to have a tool that could do some of this work automatically.

What it does

SilvaPass extracts important details from trade documents and puts them into a structured format so users can check the information more easily. It uses an LLM to help with extraction and has a rule-based fallback if the AI extraction doesn't work.

How we built it

We built the backend to handle document processing and connected it to an LLM API for extracting information. We also added rule-based extraction as a backup so the system can still work if the API call fails.

Challenges we faced

One challenge was getting the information from different documents into the same format. We also had to handle API errors and make sure the system had a backup when AI extraction wasn't available.

What we learned

While building SilvaPass, we learned how to connect an LLM API to a project and use it to solve a practical problem. We also got more experience with backend development, document processing, and handling errors.

What's next

We want to support more types of trade documents, improve extraction accuracy, and make it easier to compare details across documents to find mistakes or missing information.