AGPL-3.0 · SELF-HOSTED
DATAMART
A structured information platform you run yourself — schema, access control, workflows and an admin UI in one binary, with your entries stored as files you own.
- ≤300MEntries Supported
- 4Client SDKs
- ~50 MBSingle Binary, No Runtime
- 106 MBFull Stack, Idle
DMART in 30 seconds
-
A space organizes your data into folders. A schema describes a valid entry, and dmart checks every write against it.
-
JSON entries, files and tickets all go in. Each one lands as a plain file on disk and as an indexed row in PostgreSQL or SQLite.
-
Roles and permissions are checked on every request, down to the folder, the action and the field.
-
One query language searches everything you stored: filter on any field, sort, and page through the results.
-
A REST API, SDKs for Python, TypeScript, Dart and C#, an admin UI, a catalog, and an MCP server for AI agents, all from one binary.
Core Pillars
Unified Data
Entries, attachments, metadata — all in one model. Structured and unstructured data, seamlessly.
Powerful Search
Full-text search, filtering, aggregation. Powered by SQL full-text blazing performance.
Access Control
Role-based access, granular permissions down to folder and entry level. Secure by default.
Extensible
Plugin architecture, webhooks, custom workflows. Adapt DMART to any use case.
Runs Where a Backend Normally Can’t
One self-contained binary, compiled ahead of time. No runtime to install, no container required, no separate frontend to deploy — the admin UI ships inside it.
Measured on a Raspberry Pi Zero 2 W with 416 MB of usable RAM, serving with PostgreSQL alongside it on the same board: 106 MB resident for the whole stack, ~14 ms warm reads, and four concurrent logins served with a bounded peak that is fully reclaimed. A 10-hour continuous write soak held memory flat — a plateau, not a slope — with no OOM and no thermal throttling. The same binary runs on a server.
- 106 MB
- dmart + PostgreSQL, idle, of 416 MB
- ~14 ms
- warm read, 5,000-entry fixture
- 0.6 GB/day
- written under sustained load — ~460 yr on a 100 TBW card
Your Data Stays Yours
Files are the source of truth
Entries live as plain files on disk. The SQL store is a rebuildable index over them — copy the folder, rebuild the index anywhere, and you have your system back.
Standard formats, not a vendor schema
Payloads are JSON, constrained by JSON Schema Draft 7 and evaluated by a standard library. There is no dmart-specific type system to migrate out of.
Self-hosted, AGPL-3.0
Run it on your hardware, in your network, air-gapped if you need to. No per-seat pricing and no hosted service you depend on.
How It Works
- 01
Define Your Space
Create spaces and organize data into hierarchical folders — just like a file system.
- 02
Store Anything
Create entries with JSON payloads, attach files, add metadata. Schema-validated or free-form.
- 03
Query & Discover
Full-text search, filter by any field, sort, paginate. All through a consistent REST API.
- 04
Integrate Everywhere
Use official SDKs for Python, TypeScript, Dart, and C#/.NET. Or call the REST API directly.
Explore
Features
Unified data management, collaboration tools, and advanced search capabilities.
Why DMART?
Transform data from a liability into an asset. Own your data, no vendor lock-in.
Technical
ASP.NET Core on .NET, PostgreSQL-backed. A single Native-AOT binary — built for simplicity and speed.
Drivers
Official client libraries for Python, TypeScript/JavaScript, Dart/Flutter, and C#/.NET.