DATAROOM.NETWORK
Knowledge Graph Runtime — Local, Scriptable, AI-Native
Scroll to Explore
01 / 05
Knowledge as a Living Graph

A single binary that thinks
in three dimensions

DATAROOM is a self-contained knowledge graph platform shipped as a single Rust binary. Markdown notes become interactive 3D nodes. Relationships become traversable edges. The entire corpus — files, state, embeddings, computation — lives locally, with no cloud dependency and no configuration overhead.

1
Binary. No dependencies.
3D
Live graph rendering
Node type extensions
02 / 05
Interactive 3D Knowledge Graph

Your knowledge base,
spatially rendered

A Three.js-powered 3D arena where every markdown file is a node and every wikilink is an edge. Drag nodes into formation, author content in an embedded CodeMirror editor, define custom GLSL shaders per node type, and navigate the full corpus through a VSCode-style command palette — all in one unified interface.

Wikilink Edges
[[double-bracket]] notation creates live graph edges with full autocompletion and backlink resolution
Custom Node Types
Scaffold new types with backend.js, frontend.js, shader.glsl, and a custom iframe UI — no recompile needed
DAG Enforcement
Optional strict acyclic mode with petgraph-based cycle detection and real-time warnings
Command Palette
Ctrl+Shift+P surfaces every graph operation, search, and navigation action — keyboard-first workflow
03 / 05
Embedded AI — No Cloud Required

Four ML models.
Zero external calls.

Four ONNX models are compiled directly into the binary via include_bytes! — MiniLM embeddings, DistilBERT QA, DistilRoBERTa sentiment, and BERT-NER. Inference runs entirely on-device. A built-in vector database handles similarity search across your corpus with cosine, euclidean, and manhattan metrics.

Embed
MiniLM 384-dim
Search
Vector similarity
Extract
BERT-NER entities
Answer
DistilBERT QA
Classify
k-NN + TRM
04 / 05
Reactive Computation Engine

Nodes that compute.
Graphs that execute.

Each node runs a QuickJS JavaScript backend with a full scripting API — persistent state, typed message passing, shell execution, HTTP fetch, and file I/O. Timer nodes fire on schedules persisted across restarts. The graph becomes a live reactive system: connect nodes, route events, build autonomous workflows without leaving the interface.

Message Passing
BFS-routed typed events traverse edges with configurable depth limits and automatic cycle detection
Persistent Timers
SQLite-backed timer rows survive server restarts — schedule node firings with millisecond precision
Shell Integration
cl() executes system commands with streaming output, timeouts, and callbacks from within any node script
LLM Tool Registry
JSON-schema validated tool definitions expose graph operations to any LLM via the structured REPL API
05 / 05
Early Access

A new interface for
thinking with machines

DATAROOM is available now as an early-access release. A single binary. No accounts, no cloud, no configuration. Your knowledge graph, your compute, your data — entirely local and fully scriptable. Join the list for release announcements and early builds.

Confidential  —  Enterprise  —  By Invitation