Skip to content

0003-agents.md

Latest commit

 

History

History
260 lines (191 loc) · 5.48 KB

0003-agents.md

File metadata and controls

260 lines (191 loc) · 5.48 KB

Agent APIs Reference

powermem provides comprehensive multi-agent memory management capabilities.

Agent Memory Management

Creating Agent-Specific Memories

from powermem import Memory, auto_config

config = auto_config()

# Create memory instances for different agents
support_agent = Memory(config=config, agent_id="support_agent")
sales_agent = Memory(config=config, agent_id="sales_agent")
tech_agent = Memory(config=config, agent_id="tech_agent")

Agent-Specific Operations

Adding Agent Memories

# Support agent adds memory
support_agent.add(
    messages="Customer prefers email support",
    user_id="customer123",
    agent_id="support_agent",
    metadata={"category": "preference"}
)

# Sales agent adds memory
sales_agent.add(
    messages="Customer interested in AI features",
    user_id="customer123",
    agent_id="sales_agent",
    metadata={"category": "interest"}
)

Agent-Specific Search

# Search only support agent memories
results = support_agent.search(
    query="customer preferences",
    user_id="customer123",
    agent_id="support_agent"  # Filter by agent
)

Cross-Agent Search

# Search across all agents
results = support_agent.search(
    query="customer information",
    user_id="customer123"
    # No agent_id filter - searches all agents
)

Multi-Agent Memory Manager

Using MultiAgentMemoryManager

from powermem.agent.implementations.multi_agent import MultiAgentMemoryManager
from powermem.agent.components.collaboration_coordinator import CollaborationCoordinator
from powermem.agent.components.permission_controller import PermissionController

config = auto_config()

# Create multi-agent manager
manager = MultiAgentMemoryManager(config=config)

# Register agents
manager.register_agent("alice", "developer")
manager.register_agent("bob", "developer")
manager.register_agent("charlie", "qa")

Agent Components

Collaboration Coordinator

Manages agent collaboration and shared memories.

from powermem.agent.components.collaboration_coordinator import CollaborationCoordinator

coordinator = CollaborationCoordinator(config=config)

# Track collaboration
coordinator.track_collaboration(
    agent_ids=["alice", "bob"],
    memory_id="memory_123",
    context="Working on API integration"
)

Permission Controller

Controls access permissions for agent memories.

from powermem.agent.components.permission_controller import PermissionController

permission = PermissionController(config=config)

# Set permissions
permission.set_permission(
    agent_id="alice",
    memory_id="memory_123",
    permission="READ_WRITE"
)

# Check access
can_access = permission.check_access(
    agent_id="bob",
    memory_id="memory_123"
)

Privacy Protector

Manages privacy levels and data protection.

from powermem.agent.components.privacy_protector import PrivacyProtector

privacy = PrivacyProtector(config=config)

# Set privacy level
privacy.set_privacy_level(
    memory_id="memory_123",
    level="PRIVATE"
)

# Check if memory can be shared
can_share = privacy.can_share(
    memory_id="memory_123",
    target_agent_id="bob"
)

Scope Controller

Manages memory scopes (AGENT, USER, GROUP, etc.).

from powermem.agent.components.scope_controller import ScopeController

scope = ScopeController(config=config)

# Set memory scope
scope.set_scope(
    memory_id="memory_123",
    scope="AGENT"
)

# Get memories by scope
agent_memories = scope.get_by_scope(
    agent_id="alice",
    scope="AGENT"
)

Memory Scopes

powermem supports different memory scopes:

  • AGENT: Agent-specific memories
  • USER: User-specific memories
  • GROUP: Group/shared memories
  • SYSTEM: System-wide memories

Setting Memory Scope

memory.add(
    messages="Project status update",
    user_id="alice",
    agent_id="alice_dev",
    metadata={"scope": "AGENT"}
)

Scope-Based Search

# Search only agent-scoped memories
results = memory.search(
    query="project status",
    user_id="alice",
    filters={"scope": "AGENT"}
)

Multi-User Memory Manager

For multi-user scenarios:

from powermem.agent.implementations.multi_user import MultiUserMemoryManager

config = auto_config()

# Create multi-user manager
manager = MultiUserMemoryManager(config=config)

# Add user-specific memory
manager.add_user_memory(
    user_id="user123",
    messages="User preference",
    metadata={"category": "preference"}
)

# Search user memories
results = manager.search_user_memories(
    user_id="user123",
    query="preferences"
)

Hybrid Memory Manager

For hybrid scenarios combining multi-agent and multi-user:

from powermem.agent.implementations.hybrid import HybridMemoryManager

config = auto_config()

# Create hybrid manager
manager = HybridMemoryManager(config=config)

# Add memory with both agent and user context
manager.add(
    messages="Collaborative memory",
    user_id="user123",
    agent_id="agent456",
    metadata={"scope": "GROUP"}
)

Best Practices

  1. Use agent_id consistently: Always specify agent_id when creating memories
  2. Scope appropriately: Use appropriate memory scopes for isolation
  3. Set permissions: Configure permissions for sensitive memories
  4. Track collaboration: Use collaboration coordinator for shared work
  5. Privacy by default: Set appropriate privacy levels

See Also