Memory Model

The core Memory class is the fundamental building block of Reeflect, representing a discrete piece of information that can be stored, retrieved, and reasoned with.

Memory Structure

@dataclass
class Memory:
    # Core fields
    id: str = field(default_factory=lambda: str(uuid.uuid4()))
    content: str = ""  # The actual content of the memory
    namespace: str = "default"  # Organizational namespace
    memory_type: MemoryType = MemoryType.PERSISTENT
    status: MemoryStatus = MemoryStatus.ACTIVE
    
    # Metadata and tracking
    created_at: datetime = field(default_factory=datetime.now)
    updated_at: datetime = field(default_factory=datetime.now)
    last_accessed: Optional[datetime] = None
    access_count: int = 0
    
    # Semantic understanding
    embedding: Optional[List[float]] = None  # Vector embedding
    tags: List[str] = field(default_factory=list)  # Categorical tags
    
    # Intelligence attributes
    importance: float = 0.5  # Importance score (0.0-1.0)
    confidence: float = 1.0  # Confidence in accuracy (0.0-1.0)
    metadata: Dict[str, Any] = field(default_factory=dict)
    
    # Relationships
    relations: List[MemoryRelation] = field(default_factory=list)
    
    # Access control
    permissions: MemoryPermissions = field(default_factory=lambda: 
        MemoryPermissions(owner="system"))

Memory Types

Type	Description	Use Cases
SESSION	Temporary memory for current interaction	Current conversation, Temporary context
PERSISTENT	Long-term stored memory	User preferences, Important facts, Recurring information
SYNTHETIC	AI-generated or inferred memory	Summarized knowledge, Inferred preferences, Generated insights

Memory Relationships

Memories can be connected through relationships, forming a knowledge graph structure that enables more sophisticated reasoning and retrieval.

@dataclass
class MemoryRelation:
    """Represents a relationship between memories."""
    
    target_id: str  # ID of the target memory
    relation_type: MemoryRelationType  # Type of relationship
    confidence: float = 1.0  # Confidence in the relationship (0.0-1.0)
    metadata: Dict[str, Any] = field(default_factory=dict)

The following relationship types are available:

• SUPPORTS - Memory supports or reinforces another memory
• CONTRADICTS - Memory contradicts or conflicts with another memory
• EXTENDS - Memory extends or elaborates on another memory
• PRECEDES - Memory comes before another in a sequence
• FOLLOWS - Memory comes after another in a sequence
• RELATED - General relationship without specific type

Best Practice

Use memory relationships to build a rich knowledge graph that enables more sophisticated reasoning and retrieval. Relationships help the system understand how different pieces of information connect and influence each other.

Next Steps

Explore Storage Systems to learn how Reeflect persists memories for long-term use and efficient retrieval.