What's going on here?
Too much information. Too little time.
The main issue that prompted this tool is the overwhelming amount of information that I am faced with when navigating places like congress.gov and ballotopedia. It is impractical to understand what cadidates are actually doing by looking at the raw text of the bills the put their names on due to the dense legal jargon and high word count of the bills.
Step 1 Scrape Members
Members of 119th Congress are scraped from the official congress.gov website, and saved to a local database
Step 2 Scrape the Legislation text, sponsor, and cosponsors
Each legislation is scraped from congress.gov and saved to the database
Step 3 Summarize the legal impacts of each bill
Now we loop through each bill in the database and ask an AI to make a summary of the legal impacts of the bill. The idea here is to shrink long bills down to use fewer words/tokens and setup the next pass to focus on what the bills do ideally.
Step 4 Highlight impacts that show up
Finally we loop through all of the legislators, grab all of the summaries of bills that they either sponsored or cosponsored, and ask for trends between the bills. The bills are sorted from most cosponsors to least cosponsors. This is important because the most recent text in a prompt is the highest weighted in the response, so we are valuing the bills that make our legislator unique higher.
