​
What are Agents in OVADARE?

Agents in OVADARE represent autonomous AI entities already defined within multi-agent platforms like AutoGen or CrewAI. OVADARE does not orchestrate these agents but instead monitors their interactions to detect and resolve conflicts. In OVADARE, agents:
  • Interact with other agents as part of their workflows.
  • Operate under defined policies to avoid conflicts.
  • Trigger conflict detection mechanisms when actions are misaligned or violate boundaries.
Agents are central to OVADARE’s functionality. By monitoring their activities and enforcing policies, OVADARE ensures that multi-agent systems operate harmoniously.

​
How OVADARE Works with Agents

​
1. Agent Monitoring

OVADARE tracks agent activities using the AgentRegistry. This registry captures interactions between agents and enables real-time conflict detection. 
from ovadare.agents.agent_registry import AgentRegistry
from ovadare.agents.agent import Agent


registry = AgentRegistry()

# Register agents
agent_a = Agent("AgentA")
agent_b = Agent("AgentB")
registry.register_agent(agent_a)
registry.register_agent(agent_b)

# Log an interaction
interaction = {
    "from_agent": "AgentA",
    "to_agent": "AgentB",
    "action": "share_data",
    "details": {"data_size": "100MB"}
}

registry.log_interaction(interaction)

​
2. Conflict Detection

When agents perform actions that violate policies or create conflicting workflows, OVADARE detects these issues through the ConflictDetector. 
from ovadare.conflicts.conflict_detector import ConflictDetector

# Detect conflicts
conflict_detector = ConflictDetector()
conflicts = conflict_detector.detect("AgentA", interaction)

if conflicts:
    print("Conflicts detected:", conflicts)
else:
    print("No conflicts detected.")

​
3. Policy Compliance

Agents in OVADARE operate under defined policies. These policies ensure that agents adhere to boundaries set by administrators and prevent issues like misaligned actions or resource conflicts.

​
Key Differences from Orchestration Frameworks

OVADARE is not an orchestration tool. It does not:
  • Define agent roles, tasks, or workflows.
  • Assign tasks or manage agent capabilities.
Instead, OVADARE complements orchestration frameworks by:
  • Observing agent interactions.
  • Enforcing policy compliance.
  • Detecting and resolving conflicts.
For example, if an orchestration framework like AutoGen instructs two agents to work on a task and a conflict arises (e.g., one agent overwriting another’s output), OVADARE flags the issue and suggests a resolution.

​
Learn More