Roadmap

Where the platform is, and where it's going.

LumIntelligence is developed in the open with our own products first, every capability below either ships in a Lumi native app today, or is being built toward that bar.

Now

Shipping and validated

LLM-agnostic architecture

Gemini, OpenAI, or fully offline with Ollama, swap providers without touching your data model.

Text-based ingestion

Free-form written content, notes, transcripts, documents, is the current ingestion surface.

General-purpose parsing engine

Node Discovery (NoD), Autonomous Data & Heuristics Discovery (ADHD), and the Intention Engine (IE) work together to find entities and route queries without hand-tuned rules per domain.

Supports Lumi native apps

LumiContacts runs in production on this foundation today, proving the graph under a real, opinionated product.

Next

In active development

Multi-modal ingestion

Extending beyond text to images, audio, and other source formats as first-class ingestible units.

Expanded file format & storage support

Broader native support for the formats and storage backends teams already use, without a custom pre-processing step.

Additional third-party platform integrations

Direct connections into the tools where source content already lives.

Higher workflow autonomy, RSVP

The Recursive Survey & Validation Protocol lets the pipeline revisit and self-check its own parsing, enabling more purposeful downstream processing with less manual tuning.

Multi-knowledge-network workflows, TGS

The Twin Galaxy System runs parallel, specialized knowledge networks side by side, for proprietary or domain-specific knowledge that shouldn't blend into a general graph.

Open LumIntelligence as a development platform

Opening the platform so outside developers can build their own applications on the same engine that powers the Lumi native apps.

A quick glossary

A few terms above are shorthand for internal systems. Directional definitions, expect these to sharpen as each capability ships.

Node Discovery

Identifies entities mentioned within raw source text.

Autonomous Data & Heuristics Discovery

Surfaces structure and patterns in unstructured input without hand-written per-domain rules.

Intention Engine

Interprets what a query is actually asking, to guide retrieval and validation.

Recursive Survey & Validation Protocol

A self-checking pass that revisits its own parsing before committing results downstream.

Twin Galaxy System

Parallel, specialized knowledge networks that run alongside the general graph for proprietary domains.

Want to build on where this is headed?

Connect on LinkedIn