NovaNet

Native content generation engine powered by Neo4j knowledge graphs

Generate culturally-authentic content across 200+ locales — not translation, but true localization from semantic concepts.

Part of SuperNovae Studio

---

Table of Contents

• Philosophy
• Why NovaNet?
• How It Works
• Use Cases
• Comparison
• Features
• Architecture
• Tech Stack
• Status
• Links

---

Philosophy

We don't translate. We generate natively.

This isn't a marketing slogan. It's a fundamental architectural choice that shapes every decision in NovaNet.

❌ Traditional:  Source → Translate → Target
                         ↓
              Context lost, meaning diluted

✅ NovaNet:     Concept (invariant) → Generate natively → L10n
                         ↓
              Context preserved, meaning amplified

Traditional localization treats content as strings to be converted. NovaNet treats content as semantic concepts to be expressed — each locale gets content that was born in that language, not translated into it.

Read more: VISION.md

---

Why NovaNet?

Traditional translation pipelines lose context. They translate words, not meaning.

NovaNet builds a knowledge graph of your content — understanding relationships, terminology, and cultural context — then generates native-quality content that preserves meaning across languages.

Problem	NovaNet Solution
Lost context in translation	Knowledge graph preserves semantic relationships
Inconsistent terminology	Centralized concept mapping with locale-specific expressions
Generic AI translations	Domain-aware generation with cultural knowledge layers
Manual locale management	Automated multi-locale orchestration (200+ locales)

---

How It Works

NovaNet uses a 3-layer architecture separating invariant concepts from locale-specific generation:

%%{init: {'theme': 'base', 'themeVariables': {'lineColor': '#64748b'}}}%%
flowchart LR
    subgraph INVARIANT[" Invariant Layer "]
        C[Concept]
        P[Page]
        B[Block]
    end

    subgraph KNOWLEDGE[" Locale Knowledge "]
        L[Locale]
        V[Voice & Culture]
        E[Expressions]
    end

    subgraph OUTPUT[" Generated Output "]
        CL[ConceptL10n]
        PL[PageL10n]
        BL[BlockL10n]
    end

    C --> CL
    P --> PL
    B --> BL
    L --> CL
    L --> PL
    L --> BL
    V --> CL
    E --> BL

    classDef invariant fill:#6366f1,stroke:#4f46e5,color:#fff
    classDef knowledge fill:#10b981,stroke:#059669,color:#fff
    classDef output fill:#8b5cf6,stroke:#7c3aed,color:#fff

    class C,P,B invariant
    class L,V,E knowledge
    class CL,PL,BL output

    style INVARIANT fill:#e0e7ff,stroke:#6366f1,color:#312e81
    style KNOWLEDGE fill:#d1fae5,stroke:#10b981,color:#064e3b
    style OUTPUT fill:#ede9fe,stroke:#8b5cf6,color:#4c1d95

Loading

1. Invariant Layer — Define concepts, pages, and blocks in a language-neutral way
2. Locale Knowledge — Each locale has its own voice, cultural references, expressions, and formatting rules
3. Generated Output — Content is generated natively for each locale, not translated

---

Use Cases

Use Case	Description
Multilingual SaaS	Generate native content for 200+ locales from a single concept graph
SEO Localization	Domain-specific keyword generation with locale-aware expressions
AI Content Orchestration	Claude-powered generation with cultural knowledge context
Brand Consistency	Maintain voice and terminology across all markets
Knowledge Graph Visualization	Explore semantic relationships in interactive 2D/3D

Target Application

NovaNet powers QR Code AI — generating native content across 200+ locales for the world's most advanced QR code platform.

---

Comparison

Aspect	Traditional Translation	NovaNet
Approach	Source → Target	Concept → Native Generation
Context	Lost between languages	Preserved in knowledge graph
Quality	Robotic, generic	Native, culturally-aware
Terminology	Inconsistent across locales	Centralized concept mapping
Scalability	Linear (per locale)	Parallel (all locales)
Cultural Fit	Surface-level adaptation	Deep locale knowledge layers

---

Features

Knowledge Graph (Neo4j)

Capability	Description
35 Node Types	Concepts, Pages, Blocks, Locales, Expressions, and more
50+ Relationships	Semantic links with temperature for spreading activation
3 Scopes	Global (shared knowledge), Shared (cross-project), Project (specific)
~19,000 Nodes	Production graph for QR Code AI

Content Generation

Capability	Description
200+ Locales	Full international coverage with native generation
Orchestrator Pattern	Dispatches tasks to specialized sub-agents
Locale Knowledge	Voice, culture, market, lexicon per locale
SEO & GEO	Search and generative engine optimization

NovaNet Studio

Capability	Description
2D/3D Visualization	React Flow + force-graph with toggle (V key)
AI Chat	Natural language → Cypher with Claude API (⌘J)
40+ Shortcuts	Full keyboard navigation
10 Filter Presets	Quick views via number keys

---

Architecture

%%{init: {'theme': 'base', 'themeVariables': {'lineColor': '#64748b'}}}%%
flowchart TB
    subgraph MONO["NovaNet Monorepo (novanet-dev)"]
        direction TB
        CORE["@novanet/core v8.1.0\nTypes · Schemas · Filters"]
        DB["@novanet/db v1.0.0\nDocker · Seeds"]
        CLI["@novanet/cli v1.0.0\nValidation · Generation"]
        STUDIO["@novanet/studio v0.1.0\nNext.js · React · 2D/3D"]
    end

    NEO4J[("Neo4j 5.26\n~19,000 nodes")]

    CORE --> STUDIO
    CORE --> CLI
    DB -.-> NEO4J
    STUDIO --> NEO4J

    classDef pkg fill:#06b6d4,stroke:#0891b2,color:#fff
    classDef cli fill:#f59e0b,stroke:#d97706,color:#fff
    classDef app fill:#8b5cf6,stroke:#7c3aed,color:#fff
    classDef db fill:#018bff,stroke:#0284c7,color:#fff

    class CORE,DB pkg
    class CLI cli
    class STUDIO app
    class NEO4J db

    style MONO fill:#f8fafc,stroke:#64748b,color:#334155

Loading

Packages

Package	Version	Description
@novanet/core	8.1.0	TypeScript types, Zod schemas, NovaNetFilter API, Cypher generators
@novanet/db	1.0.0	Neo4j Docker infrastructure, Cypher seeds, migrations
@novanet/cli	1.0.0	Validation and generation tools (in development)
@novanet/studio	0.1.0	Interactive graph visualization with AI-powered queries

---

Tech Stack

Layer	Technologies
Runtime	React 19, Next.js 16, TypeScript 5.9, Tailwind 3.4
Database	Neo4j 5.26 Community + APOC
State	Zustand 5, Zod 3.24
Visualization	React Flow, react-force-graph-3d
AI	Claude API (Anthropic)
Build	Turborepo 2.8, pnpm 9

---

Status

Development	Private — Active development by SuperNovae Studio
This Repo	Public showcase & documentation
Production	Powering QR Code AI with 200+ locales

Ecosystem

Repository	Description	Access
novanet	Public showcase (this repo)	Public
novanet-dev	Turborepo monorepo (main development)	Private

Legacy repositories (novanet-core, novanet-studio, novanet-infra) have been migrated to the monorepo.

Interested?

This project is currently in private development. For inquiries:

• Star this repo to follow updates
• Visit SuperNovae Studio
• Contact via supernovae.studio

---

Links

Resource	Description
Vision	Product vision and philosophy
Ecosystem	Repository structure and packages

---

Built by Thibaut MÉLEN & Nicolas CELLA at SuperNovae Studio