Mozg

⚠️ Work in Progress — This project is experimental and under active development. APIs, schema, and behaviour may change without notice. Not recommended for production use.

A "slow by design" cross-database query layer with a single GraphQL endpoint.

Mozg (Polish for "brain") is an experiment in thinking differently about data: instead of centralizing knowledge, it teaches systems how to connect it.

At its core, Mozg is a "slow by design" cognition layer — a single GraphQL endpoint that doesn't store knowledge, but reasons across it. Rather than building yet another database, Mozg acts as a unifying language for describing relationships between entities that already exist: rows in PostgreSQL, nodes in Neo4j, documents in ArangoDB, pathways in BioCyc, or even resources behind REST APIs.

Where most systems optimize for speed through precomputation and duplication, Mozg embraces a different trade-off: freshness and flexibility over latency. Every query is executed live, traversing connections across heterogeneous data sources at runtime. The result is not just data retrieval, but on-demand knowledge synthesis.

The single /graphql endpoint becomes a kind of neural interface:

• You bring your own credentials
• Mozg maps the relationships
• The system resolves meaning across boundaries

And because not every user wants to think in GraphQL, Mozg includes a browser-based query builder — a visual way to compose complex, cross-database queries without writing a single line of code.

In essence, Mozg is less like a database and more like a brain-shaped abstraction layer: it doesn't store memories — it connects them.

---

Features

• 🔌 One endpoint – /graphql queries any supported database or REST API
• 🔑 Credentials in the query – connection parameters are part of the GraphQL input
• 🔗 Relation traversal – hasMany, hasOne, belongsTo across tables / endpoints
• 🗺 Schema introspection – discover tables & columns of SQL and graph databases; REST sources are described from their catalog entry
• 🌐 Web UI – visual query builder at / with examples dropdown
• 🧪 GraphiQL – full IDE at /graphql
• 🦉 OWL ontology ingestion – parse Turtle, RDF/XML, OWL/XML and Manchester Syntax ontologies via ingestOntology mutation

---

Quick start

npm install
npm start          # http://localhost:4000
# or
npm run dev        # restarts on file changes

---

Open in GitHub Codespaces

Click the button below to open a fully configured dev environment in your browser – no local setup required.

The Codespace will automatically:

1. Start a Node 20 container
2. Run npm install
3. Start the dev server (npm run dev)
4. Forward port 4000 and open a preview

---

Public Databases & APIs

The following free, public endpoints can be used directly — no sign-up required.

General

Name	Driver	Host	Database	User	Password	Notes
RNAcentral	postgres	hh-pgsql-public.ebi.ac.uk	pfmegrnargs	reader	(public read-only — see RNAcentral docs)	⚠️ Currently degraded: the public reader account has no USAGE on the rnacen schema, so introspection returns almost nothing and queries fail with permission denied for schema rnacen
RFAM	mysql	mysql-rfam-public.ebi.ac.uk	Rfam	rfamro	(empty)	RNA families
Neo4j Movies	neo4j	demo.neo4jlabs.com	movies	movies	(public demo — see Neo4j Labs)	Movie graph (scheme: neo4j+s)
JSONPlaceholder	rest	–	https://jsonplaceholder.typicode.com	–	–	Fake REST data
RestCountries	rest	–	https://restcountries.com/v3.1	–	–	Country data
Open-Meteo	rest	–	https://api.open-meteo.com/v1	–	–	Weather (no key)
PokéAPI	rest	–	https://pokeapi.co/api/v2	–	–	Pokémon
Jikan	rest	–	https://api.jikan.moe/v4	–	–	Anime (MyAnimeList)

Bioinformatics

All entries below are freely accessible without registration unless noted. Pass connection parameters as shown; entity names and where filters follow each catalog entry's documentation.

Name	Driver	Catalog key	Base URL	Notes
NCBI E-utilities	rest	ncbi	https://eutils.ncbi.nlm.nih.gov/entrez/eutils	General access to all NCBI databases; pass db= in where
GenBank	rest	genbank	https://eutils.ncbi.nlm.nih.gov/entrez/eutils	Nucleotide sequences; use where: { db: "nuccore" }
PubMed	rest	pubmed	https://eutils.ncbi.nlm.nih.gov/entrez/eutils	Biomedical literature; use where: { db: "pubmed" }
RefSeq	rest	refseq	https://api.ncbi.nlm.nih.gov/datasets/v2	Reference sequences via NCBI Datasets v2 API
GEO	rest	geo	https://eutils.ncbi.nlm.nih.gov/entrez/eutils	Gene expression; use where: { db: "gds" }
UniProt	rest	uniprot	https://rest.uniprot.org	Protein knowledgebase (Swiss-Prot + TrEMBL)
PDB	rest	pdb	https://data.rcsb.org/rest/v1/core	3-D macromolecular structures
InterPro	rest	interpro	https://www.ebi.ac.uk/interpro/api	Protein families and domains
HMDB	rest	hmdb	https://hmdb.ca	Human metabolome; 200 000+ metabolites
STRING	rest	string-db	https://string-db.org/api/json	Protein–protein interaction networks
Ensembl	rest	ensembl

---

OWL Ontology Ingestion

Mozg can parse OWL ontologies in Turtle, RDF/XML, OWL/XML or Manchester Syntax and extract classes, object properties and data properties.

mutation {
  ingestOntology(input: {
    url: "https://raw.githubusercontent.com/Skitionek/Mozg/main/examples/blog.ttl"
    format: turtle
  }) {
    tripleCount
    classes { iri label subClassOf }
    objectProperties { iri label domain range relationType }
    dataProperties { iri label domain range }
  }
}

You can also supply raw content instead of a URL:

mutation {
  ingestOntology(input: {
    content: "@prefix : <http://example.org/> . :Foo a owl:Class ."
    format: turtle
  }) {
    tripleCount
    classes { iri label }
  }
}

---

Example queries

Introspect a SQLite database

{
  introspect(connection: { driver: sqlite3, database: "path/to/file.db" }) {
    tables {
      name
      columns { name type nullable isPrimaryKey }
    }
  }
}

Query with a relation

{
  query(input: {
    connection: { driver: postgres, host: "localhost", database: "mydb", user: "alice", password: "secret" }
    from: "users"
    limit: 20
    relations: [
      { entity: "posts", foreignKey: "user_id", type: hasMany, select: ["id", "title"] }
    ]
  }) {
    count
    data
  }
}

---

Commit conventions

This project uses Conventional Commits.
GitHub Copilot is configured (.github/copilot-instructions.md + .vscode/settings.json) to generate compliant commit messages automatically.

Type	Purpose
feat	New feature
fix	Bug fix
docs	Documentation only
refactor	Code restructure without behaviour change
chore	Maintenance / dependency updates
ci	CI/CD changes
test	Tests
perf	Performance
build	Build system
revert	Revert a commit

Examples:

feat(api): add introspect query for schema discovery
fix(db): handle null foreign key in belongsTo relation
chore(deps): upgrade mysql2 to 3.9.8