Learning Path Generator

1 Goal

Help people learn faster and more efficiently to reach their learning goal. LearnAnything's mission is to facilitate the consolidation of the world's knowledge sharing by creating a place where users can share their learning experiences and knowledge with others.

2 Overview

Why did we build LearnAnything? You might ask. Well, we want you to think of LearnAnything as more than just a learning path generator, but as an entry point for YOU to begin your learning journey on any topic that you could possibly imagine. Just as Mahatma Gandhi once said, "Live as if you were to die tomorrow. Learn as if you were to live forever."

3 How to Run

3.1 Local (without Docker)

• Backend
  • cd server
  • uv sync
  • Set OPENAI_API_KEY in your shell or .env
  • Optional: set CORS_ORIGINS=http://localhost:3000 (comma-separated list supported)
  • Optional: set CACHE_BACKEND=firestore to enable learning path caching (default: noop — no caching)
  • Optional: set SHARE_BACKEND=firestore to enable immutable unlisted share links. Identical snapshots reuse the same content-addressed URL (default: noop — sharing disabled)
  • Start API: uv run uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
• Frontend
  • cd client
  • npm install
  • npm run dev
  • App runs at http://localhost:3000
• API base URL config
  • client/.env.development defaults to VITE_API_BASE_URL=http://localhost:8000
  • Local FE+BE flow should call http://localhost:8000/v1/lp/{term}

3.2 Frontend Production Env (Vercel)

• The browser never calls Cloud Run directly. client/.env.production sets VITE_API_BASE_URL=/api (browser → Vercel serverless proxy in client/api/).
• In Vercel Project Settings, set BACKEND_BASE_URL=https://<cloud-run-url> and BACKEND_API_KEY=<backend API_KEY> (server-only; the proxy adds X-API-Key).
• See DEPLOY.md for the full setup.

3.3 With Docker Compose

• Run from repo root: docker compose up --build
• Frontend served on http://localhost:3000
• Backend served on http://localhost:8000

3.4 Backend Deployment (Cloud Run)

• Cloud Build triggers deploy automatically when server/** changes: dev → lp-backend-staging, main → lp-backend-prod (server/cloudbuild.yaml). See DEPLOY.md for env vars and rollback.
• The container serves FastAPI with:
  • uvicorn app.main:app --host 0.0.0.0 --port $PORT
• Health check endpoint:
  • GET /health

4 SEO & Discoverability

The client includes per-route SEO optimizations so search engines and AI crawlers can discover and cite the site:

• Per-route <head> tags — each page sets its own <title>, <meta description>, canonical URL, OpenGraph, and Twitter Card tags via a shared Seo component (client/src/seo/Seo.jsx). Uses React 19 native <head> hoisting (no extra library).
• JSON-LD structured data — Home page emits WebSite, WebApplication, and Organization schemas (with a SearchAction). About page emits AboutPage, Person, Organization, and BreadcrumbList.
• Crawl files — robots.txt (allows search + AI search bots, blocks training-only crawlers) and llms.txt are served from client/public/; sitemap.xml is generated into dist/ at build time from the indexable prerendered routes.
• Pre-rendered SEO — npm run build runs a post-build script (client/scripts/prerender-seo.mjs) that writes static HTML with <title>, meta tags, JSON-LD and readable body content for /, /about, /topics and the 24 reviewed /learn/* pages, plus noindex shells for /feedback, /learningpath and /share, so bots that don't execute JS still see the critical signals.
• noindex on generator — /learningpath?term=... is marked noindex,follow to prevent infinite indexable URLs.
• Site constants — SITE_URL, SITE_NAME, etc. are centralized in client/src/config/site.js (reads VITE_SITE_URL with a prod fallback).

5 Resources

• GPT model output comparison tool
• OpenAI Best Practices for Prompt Engineering
• OpenAI documentation - model details
• OpenAI example prompts and use cases
• OpenAI Production Best Practice
• FE Utilities
  • WebGradient Templates
  • ShareGPT - Design reference
• Similar ideas
  • Similar idea: Learn from anyone
  • Roadmap.sh - open source learning groadmap project
  • Learn Anything xyz
  • Learney - closest example
  • Map of Reddit by Anvaka (graph lover)
• Miscellaneous
  • Github Copilot code assistant
  • Fine tuning the model (GPT-3) - Tutorial by Weights & Biases (to watch)