Skip to main content

OntologyEngineerAgent

What it is​

An Agent factory and thin wrapper class that configures an “Ontology Engineer” agent specialized in BFO (Basic Formal Ontology) guidance and a text-to-ontology workflow by delegating work to sub-agents.

Public API​

  • create_agent(agent_shared_state: Optional[AgentSharedState] = None, agent_configuration: Optional[AgentConfiguration] = None) -> Optional[Agent]
    • Creates and returns an OntologyEngineerAgent instance.
    • Responsibilities:
      • Loads the default chat model via naas_abi.models.default.get_model().
      • Applies a default AgentConfiguration(system_prompt=SYSTEM_PROMPT) when none is provided.
      • Creates an AgentSharedState when none is provided.
      • Instantiates and attaches two sub-agents:
        • naas_abi.agents.EntitytoSPARQLAgent.create_agent()
        • naas_abi.agents.KnowledgeGraphBuilderAgent.create_agent()
  • class OntologyEngineerAgent(Agent)
    • Subclass of Agent with no additional methods/attributes (placeholder for the configured agent type).

Configuration/Dependencies​

  • Depends on naas_abi_core.services.agent.Agent:
    • Agent, AgentConfiguration, AgentSharedState
  • Uses the default model provider:
    • naas_abi.models.default.get_model
  • Composes two other agents:
    • naas_abi.agents.EntitytoSPARQLAgent
    • naas_abi.agents.KnowledgeGraphBuilderAgent
  • SYSTEM_PROMPT defines operating guidelines, including:
    • Educational BFO Q&A behavior.
    • Delegation to Entity_to_SPARQL for mapping text to SPARQL.
    • Delegation to Knowledge_Graph_Builder for triplestore insertion (with a confirmation step described in the prompt).

Usage​

from naas_abi.agents.OntologyEngineerAgent import create_agent

agent = create_agent()
print(type(agent).__name__)  # OntologyEngineerAgent

With explicit configuration/state:

from naas_abi.agents.OntologyEngineerAgent import create_agent, SYSTEM_PROMPT
from naas_abi_core.services.agent.Agent import AgentConfiguration, AgentSharedState

cfg = AgentConfiguration(system_prompt=SYSTEM_PROMPT)
state = AgentSharedState()
agent = create_agent(agent_shared_state=state, agent_configuration=cfg)

Caveats​

  • The OntologyEngineerAgent class itself adds no behavior; all behavior is inherited from Agent and driven by configuration (SYSTEM_PROMPT) and attached sub-agents.
  • No tools are registered (tools = []); only sub-agents are attached.