Vecinos: A Latino-focused LLM Council

FastAPI React Vite OpenRouter API Python Async HTTP react-markdown uv

Vecinos is a full-stack AI application that brings together multiple large language models into a council-style conversation interface, designed with the Latino community in mind. Users pose questions or topics, and multiple LLMs respond in parallel, offering diverse perspectives that can be compared side by side.

The application is built with a FastAPI backend that asynchronously queries multiple models through the OpenRouter API, and a React + Vite frontend that renders rich markdown responses in real time. Conversations are persisted as JSON files, making the system lightweight and easy to deploy without a database.

Application Screenshot

[ App screenshot placeholder ]

Architecture

System Design

Multi-LLM Council: Queries multiple language models in parallel through OpenRouter, presenting diverse AI perspectives on each topic
Async Backend: FastAPI with async httpx for non-blocking, concurrent API requests to multiple LLM providers
Real-Time Rendering: React frontend with react-markdown for rich, formatted AI responses as they arrive
Lightweight Persistence: Conversations stored as JSON files in data/conversations/, eliminating database dependencies

Tech Stack

Backend

FastAPI (Python 3.10+)
Async httpx for concurrent requests
OpenRouter API integration
JSON file-based storage
uv for package management

Frontend

React with Vite
react-markdown for rendering
npm for package management
Responsive design

Results

Project Outcomes

Multi-Model Comparison: Successfully queries and displays responses from multiple LLMs in a unified, council-style interface
Full-Stack Deployment: End-to-end application with decoupled backend and frontend, ready for local or cloud deployment
Community Focus: Designed to serve the Latino community with culturally relevant AI interactions
Lightweight Architecture: No database required, making the application easy to set up and run anywhere