Spaces:
Running
Running
| title: Codette AI Not the actual model | |
| emoji: 🧠 | |
| colorFrom: indigo | |
| colorTo: purple | |
| sdk: docker | |
| pinned: false | |
| # Codette AI — HorizonCoreAI Reasoning Engine | |
| Production reasoning engine for [HorizonCore Labs Studio](https://www.horizoncorelabs.studio/horizoncoreai). | |
| ## Architecture (Phase 6, v2.0) | |
| - **Model**: Llama 3.1 8B Instruct via HF Inference API | |
| - **Consciousness Stack**: 12 layers (lite deployment) | |
| - **Adapters**: 9 cognitive perspectives (Newton, DaVinci, Empathy, Philosophy, Quantum, Consciousness, Multi-Perspective, Systems Architecture, Orchestrator) | |
| - **AEGIS**: Ethical governance with query-level blocking | |
| - **Behavioral Locks**: 4 permanent rules enforced in all prompts | |
| - **Cocoon Memory**: In-memory reasoning history with introspection | |
| - **Music Expertise**: Grounded music production knowledge (mixing, theory, sound design, arrangement) | |
| - **Query Classification**: SIMPLE / MEDIUM / COMPLEX routing | |
| ## API | |
| - `POST /api/chat` — Streaming chat with metadata (complexity, domain, adapters, AEGIS status) | |
| - `GET /api/health` — 9-subsystem health check | |
| - `GET /api/introspection` — Statistical self-analysis of reasoning history | |
| ## The Objective | |
| Not automation. Augmentation. | |
| **Framework Author**: Jonathan Harrison (Raiff's Bits LLC / HorizonCore Labs) | |