tierwatch: Gemma 4 reads the free-tier pages I depend on, and has to quote every number
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle What I Built tierwatch is a small CLI that watches the free-tier and pricing pages your projects depend on and tells you when a limit changes. It reports changed facts , not changed HTML. I keep a list of free-tier limits for a side project, and I have had to correct that list after the fact more than once.…
tierwatch is a small command-line interface (CLI) application that monitors the free-tier and pricing pages of software projects you depend on. It notifies you when a limit changes, presenting changed facts rather than changes in HTML. The developer maintains a list of free-tier limits for a side project and has had to update this list after the fact on multiple occasions. Two specific instances are Cerebras and Brave Search API.
For Cerebras, the free tier used to have a rate limit of 30 requests per minute and 60K tokens per minute, but has since changed to 5 requests per minute and 30K tokens per minute. The new offering requires a verified payment method to receive $5 in credits. This change is not hidden and can be found on the pricing page that users usually forget to revisit until their applications start returning error codes.
The Brave Search API underwent another change, where it shifted from offering 2,000 free queries per month to providing $5 in monthly credits. The developer should verify if these credits are enough for their site search calls, which are estimated at around 1,500 queries per month.
The application uses Gemma 4, an open-weight model, to extract limits from the pricing pages and verify that each quote from the page is accurately present in the actual text. It compares snapshots of the pages with previous versions to identify changes and presents a summary of the differences. The tool is built using Python, the standard library, and is available under the MIT license on GitHub.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.