Final matrix summary
- Genes / rows
- Awaiting preview
- Samples
- Awaiting selection
- Identifier mode
- Awaiting inspection
- Value domain
- Expression values; NQC validates raw-count compatibility
Scientific AI architecture
Zomniverse: A Modular Scientific AI Environment
Zomniverse is a structured AI environment for scientific reasoning, explainable computation, and inspectable knowledge systems. It combines deterministic modules, live research retrieval, curated concept registries, dictionary support, system-state explanation, and local language-model synthesis inside a unified architecture.
Rather than presenting AI as a single black box, Zomniverse organizes intelligence into visible, specialized cities , where retrieval, computation, explanation, and synthesis each have a defined role.
01
Architecture
How Zomniverse Works
Zomniverse is built as a modular system of cities, each responsible for a distinct layer of reasoning or execution. Some cities are deterministic, some are source-based, and some provide AI synthesis.
- The frontend provides a unified interface for structured, inspectable answers.
- A central orchestration layer routes requests to the appropriate city.
- Deterministic modules handle exact tasks such as computation, lexical definitions, and system-code explanations.
- ZAR AI adds a synthesis layer powered by a locally hosted model.
- Responses can combine multiple cities while preserving visible structure and source identity.
This architecture supports traceability, modular growth, and trust-aware interaction , making it possible to scale capabilities without collapsing everything into one undifferentiated model response.
02
System components
Cities in the Current System
L
Documentation
Learn City
Learn City is the documentation and knowledge layer of Zomniverse. It organizes structured system writing, scientific explanation, architecture notes, and long-form reference material in a format designed for clarity and navigation.
Z
Research and synthesis
ZAR · Zomniverse AI Research
ZAR is the active research and reasoning environment of Zomniverse. It integrates Research City, Concept City, Meaning City, System City, Compute City, and the Zomniverse Synthesis Layer to produce hybrid answers that are readable, structured, and expandable.
G
Scientific workflow
GeneBean · MARSS
GeneBean provides governed scientific workflows for dataset inspection, row identity, expression-matrix canonicalization, normalization, quality control, and downstream analysis preparation.
03
Product distinction
What Makes Zomniverse Different
Zomniverse is not designed as a generic chatbot with attached tools. It is designed as a modular scientific AI environment where different answer types come from different execution paths.
01
Research retrieval remains separate from language synthesis.
02
Computation can be deterministic rather than guessed.
03
Dictionary and concept layers can remain source-aware.
04
System behavior can be explained directly within the interface.
05
The local model is one component of the system, not the entire system.
This separation helps preserve scientific trust boundaries while still allowing flexible, human-readable answers.
04
Design principles
System Philosophy
At the core of Zomniverse is the idea that advanced AI systems should be structured, inspectable, and explainable . A useful system should not only produce outputs. It should also make its reasoning layers, authorities, and boundaries more visible to the user.
Zomniverse therefore treats explanation, modularity, and execution design as part of the product itself, not as secondary documentation.
Any system that can fail should also be able to explain how it works.
