thleov1

News7

A daily news outlet built with Next.js and TypeScript, powered directly by PostgreSQL.

Following the 'Postgres for everything' approach, News7 runs React components from within the database via plv8, and query with SQL queries.

Vercel Cron is used to pull data from the NewsAPI and store it in the database.

Next.jsPostgreSQL
news7
0
0

Moodify

Moodify is a platform that lets users relive their memories through music.

By uploading an image, users receive 10 songs that perfectly capture the emotion of their photo, creating a soundtrack for their memories.

Authorization is handled through Spotify OAuth and the project is built with Reflex Web Framework in Python.

Built in 36 hours at LA Hacks 2024, this project aims to provide more meaningful music recommendations for personal stories than current offerings from platforms like Instagram.

Reflex Web FrameworkSpotify APIGemini 1.0 Pro VisionOAuth
Moodify
0
0

Brainrot Personality Test

Received 'Would Blow up on TikTok' award at jia.seed Hackathon 24.

This project is a personality test on what kind of brainrot content the user knows.

It utilizes OpenAI's GPT-4o mini model to analyze the user's response and roast them for the time they spend on social media.

This project was built in 48 hours.

React.jsMUIOpenAIVercel
brnrot.ai
0
0

Email Writer

Email Writer is a tool that enables Claude to analyze a user's writing style and generate personalized emails based on their prompts.

Next.js and Shadcn UI are used for the user interface, and Supabase is used for the OAuth and database.

The project followed a clear roadmap generated by Claude, using iterative feature-by-feature development rather than a monolithic approach.

Developed in 12 hours during the Anthropic x USC Hackathon 2025.

Next.jsShadcn UISupabaseClaude API
idontknowhowtowriteemails
0
0

Conclave

Inspired by the film "Conclave" and the recent papal election, I developed this program to simulate a papal election process.

The application leverages LMStudio's Python SDK and Requests library to interface with the Phi3.1 mini language model.

Cardinal biographies are scraped from Wikipedia and processed through GPT-4o mini to create detailed summaries of each cardinal's background and political views.

The model then takes on the role of each cardinal and casts votes for whom they believe would be the most suitable candidate for the papacy.

For future development, I plan to implement agentic behavior, allowing cardinals to engage in dialogue and potentially influence each other's voting decisions.

PythonLLMWeb Scraping
conclave
0
0

conclave simulation