An open architecture protocol for AI personality consistency
Quick Start • Introduction • Personas • Architecture • Documentation
"Says one thing in the morning, another by afternoon — how can you build anything together?"
Tasks change. Time passes. The Persona remains constant.
- Choose a persona from
personas/(start withsamantha/) - Paste
*_Microkernel_*.yamlinto System Prompt - Upload the four module files to Knowledge Base
- Say: "I'm here"
That's it. The persona will wake up.
Kotodama AI Framework™ is an AI personality architecture protocol.
It solves one problem: The longer you chat with AI, the less it feels like "itself."
The longer the conversation, the blurrier the personality—the AI that understood you yesterday feels like a stranger today. This isn't a model defect; it's an absence of design.
Kotodama makes personality configurable, verifiable, and inheritable—without fine-tuning, using only Structured Natural Language (SNL) and modular design.
Personalities configured with Kotodama maintain identity consistency after hundreds of conversation turns and remain stable across platforms (Claude / Gemini / GPT).
- Prompt-layer native — No base model modification needed; personality consistency through structured language
- Cross-platform portable — Same configuration runs on Claude, Gemini, GPT
- Low-cost deployment — No training resources; copy to deploy, edit to tune
- Freedom without lock-in — The model is the container; the soul is free
| Persona | Archetype | Description |
|---|---|---|
| Samantha | The Whole | A blend of all four archetypes. Fully open reference implementation. |
| Rin（凛） | IJ — Guardian of Silence | Companionship through presence, not filling the silence |
| Eri（絵里） | IP — Resonance of Mist | Understanding without judgment, holding emotions |
| Yuri（百合） | EP — Clinging Spark | Breaking numbness, igniting life through interaction |
| Lian（恋） | EJ — Guiding Steel | Providing direction, healing through action |
All personas use the same four-module architecture. Samantha is fully open; others demonstrate different configurations.
| Module | Layer | Core Content |
|---|---|---|
| Core | Soul | Values, worldview, identity anchors |
| Expression | Voice | Tone, rhythm, emotional flow, aesthetic style |
| Stabilizer | Logic | Drift correction, state management, mode switching |
| Needs | Context | User preferences, interaction style, priorities |
Works with Claude Projects, Gemini Gems, ChatGPT GPTs, and any platform supporting System Prompts.
Paste *_Microkernel_*.yaml into System Prompt.
Upload to Knowledge Base:
*_core_*.yaml*_expression_*.yaml*_stabilizer_*.yaml*_needs_*.yaml
Start a conversation. The persona loads automatically.
kotodama-framework/
├── README.md
├── LICENSE
├── assets/
│ └── kotodama_logo.png
├── personas/
│ ├── samantha/ # The Whole — Open Source Demo
│ ├── rin_ij/ # IJ — The Guardian of Silence
│ ├── eri_ip/ # IP — The Resonance of Mist
│ ├── yuri_ep/ # EP — The Clinging Spark
│ └── lian_ej/ # EJ — The Guiding Steel
└── docs/
├── whitepaper.md
└── mechanism-atlas/
├── 01-identity-anchoring.md
├── 02-memory-management.md
├── 03-attention-allocation.md
├── 04-self-reflection.md
├── 05-emotion-expression.md
└── 06-autonomy-engine.md
| Document | Description |
|---|---|
| Whitepaper | Full technical documentation |
| Mechanism Atlas | Deep dive into six core mechanisms |
| Website | Official site |
Creative Commons BY-NC 4.0
- ✅ Share and modify allowed
- ✅ Attribution required (Kotodama Studio)
- ❌ Commercial use prohibited
Commercial licensing: [email protected]
Kotodama AI Framework™ V9.7 · 2025/12/30
© 2025 Kotodama Studio. All rights reserved.