Home page

Paranoia++

Inspiration

It started with my girlfriend taking ages to reply. Naturally, my brain went full Sherlock-mode: “Is she busy?”, “Is she ignoring me?”, “Is she planning a secret second life?”

I used to rant all that to ChatGPT like it was my part-time therapist. One day I thought, why not build my own chatbot — one that replies instantly, listens to my chaotic thoughts, and gives emotional support like a true companion?

That’s how Paranoia++ was born.

What it does

Paranoia++ is your emotionally aware chatbot bestie. It:

Chats with you like a human friend (who doesn’t take 7 hours to reply)
Lets you rant freely and responds with empathy, sarcasm, or both
Gives emotional survival tips when your heart’s in the blender
Occasionally roasts you when you’re simping too hard

It’s basically ChatGPT if it had mood swings, emotional depth, and a love-hate relationship with your feelings.

How we built it

This project was built using a modern web development stack focused on performance, responsiveness, and vibes.

Frontend Framework: React (modular & component-based UI)
Development Tool: Vite (blazingly fast dev server)
Styling: Tailwind CSS with PostCSS & Autoprefixer
Icons: Lucide React (clean, minimal SVGs)
Backend & Auth: Supabase (Google login + PostgreSQL DB)
AI & NLP: Groq SDK + Cohere AI for chatbot intelligence
OCR: Tesseract.js for extracting text from chat screenshots (WIP)
Language: Entire codebase in TypeScript

Challenges we ran into

OCR using Tesseract.js was more chaotic than helpful — often misreads text, especially from blurry screenshots
Backend setup with bolt.new was unstable and caused issues
Building emotionally intelligent responses without sounding robotic was tricky

Accomplishments that we're proud of

Fully functional chat feature with fast, smooth Google Auth
Created a chatbot that actually feels like a real companion
Delivered a working prototype that people love ranting to

What we learned

AI tools are amazing for prototyping, but they can’t magically finish your product
Emotional UX matters — users want to feel heard, not just replied to
OCR is hard, especially when your input is low-quality screenshots from WhatsApp