Important: This GitHub edition is a significantly limited public demo intended for review, discussion, and investor interest. It is not the full private practitioner platform.
An NLP session tool for areas that NLP can responsibly support, and a future engine for psychotherapists, psychologists, and NLP practitioners to build their own pathways from their own knowledge and experience for use with VR.
This version demonstrates the core direction of Lumina:
- staged, rapport-based session flow
- local knowledge database as baseline rulebook
- Gemini-assisted question sequencing
- evidence-gated mapping instead of overconfident assumptions
- careful wording around camera observations
- a future path toward practitioner-led treatment design, media generation, and VR-assisted sessions
- detects browser language and defaults to Danish or English
- runs a session in stages instead of jumping to a final script
- uses a local JSON knowledge base as minimum baseline knowledge
- asks smarter follow-up questions based on user intent
- separates motivations from triggers and body signals
- demonstrates ethical guardrails for camera-assisted self-observation
Compared with the private packages, this GitHub version is intentionally restricted.
- no server-side encrypted practitioner memory
- no production authentication stack
- no file generation or download workflows
- no private treatment template management
- no production-grade MediaPipe observation pipeline
- no practitioner supervision tools
- no export, sound generation, video generation, or VR generation
Lumina is designed as a structured support tool and practitioner platform concept. The current baseline includes:
- rapport building
- staged intake and exploration
- outcome clarification
- motivation mapping
- meta-program inspired questioning
- resource elicitation
- calm integration and reinforcement
It also includes strong limits:
- no lie detection claims
- no certainty claims about hidden emotions
- no webcam heat sensing from ordinary cameras
- no replacement of clinician judgment
This repository is meant to communicate the product direction safely and clearly to:
- potential investors
- advisors
- collaborators
- technical reviewers
The full private roadmap includes encrypted memory, practitioner accounts, editable treatment templates, richer multi-modal observation, and eventually audio/video/VR-assisted sessions — but those are not fully exposed in this public demo.
npm install
cp .env.example .env.local
npm run devThen open:
http://localhost:3000Create .env.local from .env.example and add your Gemini key.
app/
components/
data/
lib/
public/A blank public/ folder is included so you can later add logos, OG images, demo assets, and other static files.
This repository already uses the correct modern config:
typedRoutes: true
It does not use the deprecated experimental.typedRoutes setting.
The current name can still change. For the next branding round, prefer names that signal:
- practitioner support
- guided change
- calm intelligence
- future VR compatibility
Examples worth exploring further:
- PraxisFlow
- FrameShift VR
- MindStage
- Rapport Engine
- Pathweave
These should still be checked directly with a registrar before you commit to domains.
For a Danish version, see README.da.md.
The long-term vision is practitioner-led rather than AI-led.
Future versions should support:
- practitioner-written or uploaded basis scripts
- practitioner observation notes as the primary source of direction
- AI suggestions only where useful, never as a replacement for professional judgment
- sound-file generation
- video-file generation
- later VR sessions with sound and visual framing
- reframing, change, removal, optimization, or strengthening workflows under practitioner control
The system should not act smarter than the psychotherapist. It should support, structure, suggest, and help operationalize the practitioner's own expertise.
See LICENSE. This public repository is evaluation-only.