Lumen Docs

Lumen Engineering Docs

Lumen is a self-hosted AI knowledge base platform. You upload documents to projects, and an AI assistant answers questions strictly from those documents with citations. Zero third-party data dependency — the only outbound call is the LLM inference API.

What lives where

| Path | Purpose | |---|---| | apps/web/ | Next.js 15 frontend (App Router, Tailwind 4) | | apps/api/ | Hono on Bun — REST API, auth, chat orchestration | | apps/embedder/ | Python FastAPI — multilingual-e5-small + BGE-reranker | | apps/worker/ | Python — BullMQ consumer: parse → chunk → embed | | apps/docs/ | This site (Next 15 + MDX) | | packages/shared/ | Shared TypeScript types | | scripts/ | Utility scripts | | docker-compose.yml | 6-service prod stack | | AGENTS.md | Project rules for AI agents working on the repo |

Entry points

Stack summary

| Layer | Tech | |---|---| | Frontend | Next.js 15 App Router · Tailwind 4 · SWR · TypeScript | | Backend | Hono · Bun runtime · Prisma · Zod | | Database | PostgreSQL 16 + pgvector | | Queue | BullMQ + Redis | | Embedder | multilingual-e5-small (384-dim, 100+ langs) | | Reranker | BGE-reranker-v2-m3 | | LLM | Configurable via /engineer/providers (OpenAI-compatible) | | Auth | Custom JWT (15m access + 7d refresh) | | Deployment | Docker Compose on Dokploy, Traefik routing |

Prod infra (single tenant — dev phase)