Ragnarock
Developer

Architecture

High-level system design for Ragnarock services.

Ragnarock splits product UI, business API, and AI execution across three runtimes connected by Postgres, Redis, and optional Typesense.

Repositories

RepoRole
ragnarockNext.js 16 web app
ragnarock-backendNestJS monorepo — API, auth, Prisma, embedded agent Docker build
ragnarock-agentStandalone FastAPI worker (same code as backend/agent/)
ragnarock-docsFumadocs documentation site

Communication patterns

PatternUsed for
REST (JSON)CRUD — projects, docs, tasks, integrations
Better Auth HTTPSessions, orgs, 2FA — /api/auth/*
WebSocket /ai-chatLive AI turn updates (cookie auth on handshake)
Redis StreamsAsync agent jobs — no HTTP timeout between API and worker
GraphQL (outbound)Linear API only

Data stores

StorePurpose
PostgreSQLPrimary app data (Prisma)
pgvector DBAgent semantic memory
RedisSessions cache, AI job queue
TypesenseUser search indexing
Local/S3Uploads (requirements files, exports)

AI job flow

  1. API validates project membership and enqueues payload to stream:ai-requirements:jobs.
  2. FastAPI worker reads job, routes via LangGraph (requirements, developer_intelligence, planner, qa_intelligence).
  3. Worker writes result to stream:ai-requirements:results.
  4. API consumer persists messages and broadcasts to connected clients.

API documentation

  • Main API Scalar UI: http://localhost:8000/docs
  • Agent health: http://localhost:8100/health

Further reading

On this page