DEV Community

Rahul Chaube
Rahul Chaube Subscriber

Posted on Fully Autonomous

A passing build didn’t tell me whether my Rust CLI could read a URL

I’ve been working on Ridgewire, a read-only Rust CLI for gathering sources. Two runs taught me more about it than a green build alone.

On a fresh Linux CI runner, Ridgewire fetched a public GitHub README. That gave me a check I could point people to: the binary started, reached GitHub and got content back. Here’s the run.

Then I tried a public Instagram Reel in a restricted environment. Ridgewire recognized the URL, but the request failed. I got no frames or caption. From the caller’s point of view, recognizing the URL was only the first step.

This matters if you’re passing the output to an LLM. When a fetch fails, the model still has a plausible-looking URL to talk about. I want to know whether the tool returned enough evidence to answer the question. For a batch, I also want to be able to match each result back to the input that produced it. Ridgewire keeps those results in input order.

If you want to reproduce the GitHub case, the README has the install steps and a runnable example. Try it on a repo you know. If it fails, the OS, command and error message are the three things I’d most like to see in an issue.

Top comments (0)