API Reference
Technical reference for developers building on SingularCore.
Architecture Layers
| Layer | Responsibility |
|---|---|
| Brain | Reasoning, decision-making, task orchestration |
| Muscle | Execution, sandboxing, high-performance operations |
| Memory | Storage and retrieval across time horizons |
Memory Tiers
| Tier | Purpose | Persistence |
|---|---|---|
| Working | Active context | Session |
| Episodic | Conversation history | Long-term |
| Semantic | Knowledge base | Permanent |
Agent System
Multiple specialized agents handle different responsibilities. Each agent operates within defined security boundaries and communicates through a secure internal message system.
Skill Interface
class Skill:
name: str # Unique identifier
description: str # Human-readable description
version: str # Semantic version
impact_tier: int # Security tier (0-3)
async def execute(self, context: Context) -> Result:
"""Execute the skill with given context."""
passAudit Ledger
All operations are logged to an immutable audit trail. Access the ledger to review:
- Action timestamps and outcomes
- Agent decisions and reasoning
- Skill executions and results
- Security tier classifications
License
SingularCore is free and open source under AGPL-3.0. Third-party skills may use any license.
See License for details.