-
Notifications
You must be signed in to change notification settings - Fork 0
03 Architecture kernel vs gateway
LAP-CHRIS\chris edited this page May 19, 2026
·
1 revision
The architecture follows a separation of powers principle: the kernel makes decisions and manages state, while the LLM translates natural language into structured actions.
The LLM never decides or writes state directly. It only translates.
| Aspect | Description |
|---|---|
| State | Manages all persistent and working state |
| Memory | Episodic, semantic, working, and emotional memory |
| Safety | Evaluates actions against 20 fundamental rules |
| Policies | Stores and enforces learned policies |
| Evaluation | Assesses signals and computes risk scores |
| Learning | Updates policies and semantic memory from outcomes |
| Aspect | Description |
|---|---|
| Translation | Converts natural language to structured commands |
| Planning | Breaks objectives into execution plans |
| Narration | Converts results back to human-readable text |
| Coordination | Orchestrates the plan-safety-execute flow |
| Feature | Community (Local) | Enterprise (Proxied) |
|---|---|---|
| Storage | SQLite | MySQL + Qdrant |
| Vectors | SQLite vector store | Qdrant HNSW |
| Cache | In-memory | Redis |
| Safety | Full pipeline | Full pipeline + MetaCritic |
| Auth | None (localhost) | JWT + RBAC |
| Scale | Single user | Multi-tenant |
Krnl-AI Community — MIT License