Nubase docs

Nubase turns AI-written code into real apps — eight modules (Database · Auth · Storage · Assets · Functions · AI Gateway · Memory · cron) under one self-hostable backend and one Studio.

Where to start

The eight modules — how data, auth, the deploy layer (Assets · Functions · cron), AI Gateway and Memory share one auth model and one tenant DB.
Quickstart — spin up a backend, create a project, write your first memory and table in ~3 min.
Memory guide — what the LLM memory layer does, how it scores, and the curl recipes.

Core modules

Memory

mem0-compatible REST API, vector + BM25 + entity boost fusion retrieval, audit history, pluggable providers (OpenAI / Anthropic / OpenAI-compatible).

Database

Dedicated Postgres per tenant. PostgREST-compatible REST API for every table. Row-Level Security executed with JWT claims.

Storage

S3 / R2-compatible buckets, file metadata in Postgres, signed URLs, RLS-aware ACLs.

Auth

Email + OAuth, JWT issuance with per-tenant secrets, refresh-token rotation. Drop-in for Supabase client SDKs.

Plus the deploy layer and more — Assets, Functions, AI Gateway and cron.

Mental model

Every request carries an apikey header — a JWT identifying the tenant and the role (anon / authenticated / service_role).
A second Authorization: Bearer <jwt> identifies the end user (used for auth.uid() in RLS and for Memory's userId binding).
Each tenant lives in its own physical Postgres database. Memory, auth tables, storage metadata and your business tables all live there together.