Local-first AI companion prototype for children: ukbe-core safety layer (crisis detection + output validation) around a local LLM, with an Ed25519-signed parent journal. v0.1 — honest prototype, not a product. Apache-2.0
Truth Vector Engine — 6-pillar manipulation detection (emotional, logical, authority, urgency, social, framing). Python. BRIDGRAI Foundation.
Dual-model anti-confabulation engine: two models must concur, REAI equations gate assert vs ask; on disagreement it says so instead of inventing. Universal OpenAI-compatible. AGPL-3.0/commercial.
Ghost sensitivity peaks in adaptive phase-coupled systems: a disguised Adler/SNIC bifurcation, independently reproduced, with a derived calibration rule
Add a description, image, and links to the bridgrai topic page so that developers can more easily learn about it.
To associate your repository with the bridgrai topic, visit your repo's landing page and select "manage topics."