Yogi Buddy: Yoga with me

Yogi Buddy: Yoga with Me is a free, no-login, no-sign-up yoga companion designed to help those who spend long hours at a desk. Created for friends who felt intimidated by yoga videos and unsure of safe poses for stiff backs, the app asks users about their feelings, available time, and any pain or injuries. Based on this information, it suggests personalized routines for stress relief, headache relief, better sleep, back and neck pain, or weight loss.

The app provides step-by-step guidance and breathing exercises while adhering to safety rules by not acting as a doctor, inquiring about pregnancy and injuries first, and recommending professional help for serious conditions. Built using open-weight models from Hugging Face Inference Providers, the backend communicates with Hugging Face's OpenAI-compatible router, employing meta-llama/Llama-3.1-8B-Instruct.

Initially, models were run locally with Ollama, then moved to a hosted open-model backend, eliminating the need for users to install anything. A proxy manages the Hugging Face token, limiting the chat endpoint to prevent abuse. The model name and token are stored in environment variables, allowing quick model swaps without altering code.

The frontend is constructed with React + Vite and deployed on Vercel, while the backend resides on Render. Open innovation enabled the creation of a personalized app without being tied to a single vendor or pricing model. Demonstrating the project's flexibility, Yogi Buddy was initially built free and locally with Ollama, allowing the creator to test the entire experience without cost.

When a chosen model was unavailable, switching to another model was as simple as updating an environment variable and redeploying. This open-source approach empowers users to control where the app runs, from a hosted API to their own server, and decide whether to keep it fully local in the future.