DEV Community

Discussion on: This is How I Automate the Tech Discovery Using LLMs

View post
Collapse
 
chariebee profile image
Charles Brown •

Nice build—using n8n as the orchestrator makes a lot of sense. A few ideas:

  • Extra GitHub signals: commit frequency (30/90 days), release cadence, issue/PR velocity, bus factor, CI status, Dockerfile/Compose presence, license.
  • More sources: npm/PyPI trending, arXiv-with-code, Hacker News/Reddit mentions, Awesome lists.
  • Prompt tweaks: ask for maturity level, infra needs (GPU/RAM), deployment paths, gotchas; enforce a JSON schema with scores and confidence.
  • Messy repos: fall back to OpenGraph images, package.json homepage, README first image, or a headless-browser snapshot.
  • Ranking: compute a weighted score and let users adjust weights.

Would be happy to try it and share feedback as you iterate.

Collapse
 
hibuno profile image
Muhibbudin Suretno •

Thanks so much for the detailed and insightful feedback, @chariebee! These are all fantastic ideas, and I appreciate you taking the time to write them out. 🔥

I've also been thinking along similar lines, and for the next set of features, I plan to focus on your suggestions for Prompt tweaks and handling Messy repos. I've already started implementing some logic for the latter, like parsing all relevant images from a README file while filtering out logos and icons.

The other suggestions are excellent and definitely on the long-term roadmap. For now, I'm prioritizing these areas to manage operational costs.

I would be thrilled to have you try it out and share more feedback as I iterate. Thanks again for the encouragement!

Let's take a look the project at spy.hibuno.com 🙏🏻