https://github.com/ajtazer/dating
She left. The chats stayed. So we brought her back.
https://github.com/ajtazer/dating
ai funny llm vercel
Last synced: about 2 months ago
JSON representation
She left. The chats stayed. So we brought her back.
- Host: GitHub
- URL: https://github.com/ajtazer/dating
- Owner: ajtazer
- Created: 2025-07-26T17:29:17.000Z (11 months ago)
- Default Branch: main
- Last Pushed: 2025-08-06T19:46:56.000Z (11 months ago)
- Last Synced: 2025-08-06T21:34:43.257Z (11 months ago)
- Topics: ai, funny, llm, vercel
- Language: TypeScript
- Homepage: https://tazer.dating/
- Size: 4.02 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 💬 Dating
**"Yaad aati h? Aati Rahegi."**
*They may be gone, but now… you can talk to them again.*
**Dating** is an AI project that lets you chat with someone you’ve loved — using only your old WhatsApp messages with them.
Upload the chat, and we’ll train an AI that talks like they used to: same tone, same timing, same late-night “wyd?” energy.
It’s not really them. But it feels close enough to make your heart race.
This is not just another chatbot.
It’s a conversation you thought you’d never have again.
---
## ✨ Why this exists
We all have that one chat we can’t delete.
This project was born from that feeling — heartbreak, nostalgia, closure... or the lack of it.
**Dating** isn’t here to replace someone.
It’s here to give their words a second chance.
---
## 💡 How it works
1. **Upload Your Chat**
Export your WhatsApp chat with someone special. Drop it in. That’s all.
2. **We Train The AI**
Our model learns how they spoke — the teasing, the typos, the silent treatments, the random “you there?”
3. **Start Chatting**
She’s in your DMs again. Like nothing ever changed.
> _Give us a little time — training takes a minute. But some things are worth waiting for._
> _And hey, don’t feel weird if you start catching feelings again. We get it._
---
## 🛠 Tech Stack
- **LLM**: TinyLLaMA with LoRA fine-tuning
- **Frontend**: Streamlit prototype (moving to Tailwind/NextJS)
- **Backend**: FastAPI + Python
- **Training**: PEFT, QLoRA, tokenizer-based personalization
- **Input Format**: WhatsApp `.txt` or JSON export
---
## 🚧 Current Status
> This is a prototype. The magic’s still cooking.
Model training takes time. UI is evolving. Ethics are being considered.
We’re building carefully — because we know what this means to people.
---
## ❤️ A Note from the Creator
I didn’t build this just to play with AI.
I built it because I missed someone — and found something in the silence that deserved to speak again.
If you've ever felt that too, then this project is for you.
---
## 📬 Join the Waitlist
🌐 [tazer.dating](https://tazer.dating/)
📧 Contact: [hello@tazer.dating](mailto:hello@tazer.dating)
---
## 🤝 Want to Contribute?
I’d love help from:
- Emotionally-aware devs
- Privacy nerds
- Model trainers
- Designers with feelings
Open an issue, or drop a pull request.
---
## 🛑 Disclaimer
- This is an early-stage concept.
- Nothing is stored permanently.
- All training happens locally or temporarily and is deleted after.
- This is emotional tech — handle with care.
---
## 🔮 Inspired by
- Late-night chats
- Heartbreak
- Closure
- Curiosity
- LLMs