Ember Nexus: Knowledge Graph API

Ember Nexus is a self-hosted backend for connected data. You store elements - nodes and the relations between them - and get a complete REST API with user accounts, permissions, file storage and powerful search already built in.

No schema to design up front. No permission system to write. No search integration to maintain. One API, running on your own infrastructure, with your data staying yours.

Ember Nexus is source first and ships as a single Docker image for AMD and ARM.

What people build with it

• Knowledge bases and note apps: Notes, tags, backlinks and attachments are a graph. Store them as one instead of fighting a relational schema.
• Content platforms and CMS: Pages, authors, categories and media, with per-user and per-group publishing rights.
• Private clouds and archives: Files and metadata side by side, shared deliberately, searchable instantly.
• Anything multi-user: Every element is private to its owner until access is explicitly granted. Collaboration is built into the data model, not bolted on.

How it feels

Create a node — any type, any properties, no migration:

curl \
  -H 'Authorization: Bearer secret-token:PIPeJGUt7c00ENn8a5uDlc' \
  -H 'Content-Type: application/json' \
  -d '{
        "type": "Note",
        "data": {
          "title": "Graph databases",
          "tags": [
            "db",
            "graph"
          ]
        }
      }' \
  https://api.example.com/

Connect it to another element — relations are data too:

curl \
  -H 'Authorization: Bearer secret-token:PIPeJGUt7c00ENn8a5uDlc' \
  -H 'Content-Type: application/json' \
  -d '{
        "type": "LINKS_TO",
        "start": "<note-uuid>",
        "end": "<other-uuid>"
      }' \
  https://api.example.com/

Search it — full-text and graph traversal in a single request:

curl \
  -H 'Authorization: Bearer secret-token:PIPeJGUt7c00ENn8a5uDlc' \
  -H 'Content-Type: application/json' \
  -d '{
        "steps": [
          {
            "type": "elasticsearch-query-dsl-mixin",
            "query": {
              "match": {
                "title": "graph"
              }
            }
          },
          {
            "type": "element-hydration"
          }
        ]
      }' \
  https://api.example.com/search

Results only ever contain elements the requesting user is allowed to see — even though the query itself is freely written by the user.

Three ideas that make it different

Everything is a graph — including permissions. Ownership and access rights are relations in the same graph as your data. Owning a node means owning its subtree; sharing with a group shares with all its members, nested groups included. Access can be queried and audited like any other data. See the access model.

Users can write real queries, safely. Clients may send full Elasticsearch Query DSL and graph path queries directly to the API. A restricted query grammar and an enforced access filter make this safe even for untrusted users — power that normally requires custom backend code, available out of the box. See search.

Search is a pipeline. Chain search steps across databases: find candidates via full-text, expand them along graph relations, hydrate the results — one request, permissions enforced at every step.

What you need

Ember Nexus runs on your hardware via Docker and uses proven databases under the hood: Neo4j, Elasticsearch, MongoDB, Redis, RabbitMQ and S3-compatible storage. Plan for at least 8 GB of RAM — Neo4j and Elasticsearch are memory-hungry. A complete Docker Compose setup is part of the installation guide.

Get started

• Installation — running in minutes with Docker Compose
• First steps — create a user, get a token, store your first data
• Search — write your first search pipeline
• Live demo — explore a running instance with example data, including tokens that never expire
• GitHub — releases, issues, and the source itself
• Discord — ask questions, share what you build