Hello HN! I'm Sai Sudheer, and I built CinePrompt because I was tired of wasting 30 minutes scrolling Netflix just to end up watching nothing.

  Here's what kept happening: I'd know exactly what vibe I wanted — something tense but not horror, or uplifting but grounded — but I had no title in mind. Netflix's keyword search was useless. Their recommendations? Just
  "what's trending" or "because you watched X."

  I'd read dozens of descriptions, watch trailer after trailer, check ratings, and still feel paralyzed by choice. My movie night would be ruined before it even started.

  So I built CinePrompt. You describe your exact mood in plain English, and it finds movies that match.

  Real examples that work:
  - "My boss is driving me insane, need something about workplace revenge"
  - "Something that feels like The Office but a movie"
  - "Just got ghosted, need dark comedies about dating disasters"
  - "Feeling like an imposter at work, want movies about faking it till you make it"

  Each recommendation comes with AI analysis explaining why it fits your vibe and what makes it worth watching — no spoilers, just enough context to help you decide with confidence.

  What started as a weekend side project turned into something I couldn't stop building. I kept adding features based on early user feedback: voice search for couch browsing, streaming platform data so you know where to watch,     
  AI explanations so you're not guessing if a movie actually fits.

  Under the hood, it's built with RAG architecture using vector embeddings for semantic search, Groq inference (with Gemini fallback), and Redis caching for sub-second responses.

  This is my first full-stack product built solo. Recent grad from Northeastern University, and CinePrompt is the most ambitious thing I've shipped.

  Try it now: 5 free searches/day, no login needed. Sign up for 15/day.

  Live: https://cineprompt.vercel.app
  GitHub: https://github.com/SudheerNaraharisetty/Cineprompt

  I'm actively looking for full-time opportunities (AI Engineering, Data Analytics, Product Management, Full-stack).

  I'd genuinely appreciate your feedback:
  - Does this solve a problem you've experienced with movie discovery?
  - What search prompts would you try? Any edge cases I should handle better?
  - What features would make this more useful for you?
  - How accurate are the recommendations for your taste?

I know there's a ton of room for improvement, and I'd be grateful for any technical critique or suggestions. This community has taught me so much, and I'm here to learn from your insights on what I could do better.

  Thanks for taking the time to check this out!

Interesting, nice work on the small details!

Do you happen to have a technique post on what goes behind this tech?