How rockie/gurubase works
Project type and purpose
• Gurubase is a full-stack, self-hostable, Retrieval-Augmented Generation (RAG) microservices system that lets you create AI-powered Q&A assistants (“Gurus”) on any topic.
• It consists of a Next.js 14 + TailwindCSS frontend, a Django REST backend with Celery workers, and several supporting services (PostgreSQL, Redis, RabbitMQ, Milvus).
File structure
• Top-level directories:
• .dev – local and self-hosted Docker Compose configs and Nginx setup
• .github – CI/CD workflows for building and publishing Docker images and running linters/tests
• src/gurubase-frontend – Next.js application with components, pages, Redux, hooks, styling, and build configs
• src/gurubase-backend – Django project with apps (accounts, analytics, core), Celery tasks, management scripts, migrations, Dockerfile, and ARCHITECTURE.md
• Root docker-compose.yml and start scripts orchestrate the microservices locally or in staging/self-hosted environments.
README for architectural insights
• It explicitly calls out a microservices architecture deployed via Docker Compose.
• Key services:
• Frontend: Next.js (React SSR/SSG)
• Backend API: Django REST Framework, ASGI for async tasks
• Vector store: Milvus for similarity search
• Message queue: RabbitMQ for Celery tasks
• Cache: Redis for rate-limiting, caching embedding lookups, session data
• Database: PostgreSQL for relational data (users, settings, crawl state)
• External services: LLM API providers, GitHub / YouTube / Web crawlers, OAuth providers, Sentry