2027 Full-Stack Monorepo

Multi-Language SaaS

Architecture Overview

Three-tier monorepo: deploy apps, write services in any language, swap providers freely between external, internal or hybrid.

/apps - Deployable

web/ - Next.js 16 (Vercel)
api/ - FastAPI gateway (AWS)
Only these get deployed

/services - Multi-Language

python/ - llm-service, mcp-service
nodejs/ - vector-db, realtime
rust/ - compute (future)
Each language isolated

External Providers (3A)

openai/ anthropic/ pinecone/
github/ slack/ stripe/
Cloud - requires internet

Internal Providers (3B)

Ollama - local LLMs offline
ChromaDB - local vectors
LocalAI - self-hosted
Zero vendor lock-in

Hybrid / Client (3C)

Gov / B2B deployments
Client-side DB on their infra
Mix external + internal
Deployment profile routes

Cache Strategy

Next.js - use cache directive
Redis - FastAPI LLM cache
Simple, fast, KIS

Database

PostgreSQL + pgvector
FastAPI owns all queries
Alembic migrations

/packages - Shared Code

TypeScript types
Shared config
Utilities

Key Docs

START_HERE.md
docs/ARCHITECTURE_DECISIONS.md
docs/SOVEREIGN_ARCHITECTURE.md
docs/ONBOARDING_CHECKLIST.md

Loading architecture diagram...