Skip to main content

SevenBucketsAgent

What it is​

An ontology-engineering chat agent factory for the BFO “7 Buckets” framework. It builds a SevenBucketsAgent (a thin subclass of Agent) configured with:

  • A large system prompt embedding a Turtle template ontology.
  • A tool to validate and save generated Turtle ontologies to a fixed ontologies directory.

Public API​

  • create_agent(agent_shared_state: Optional[AgentSharedState] = None, agent_configuration: Optional[AgentConfiguration] = None) -> Agent

    • Creates and returns a configured SevenBucketsAgent.
    • Loads a Turtle template from TEMPLATE_ONTOLOGY and injects it into the system prompt.
    • Registers one tool: save_ontology (LangChain tool; return-direct).

  • class SevenBucketsAgent(Agent)

    • No additional behavior beyond Agent (empty subclass).

Configuration/Dependencies​

  • Constants
    • NAME = "7_Buckets"
    • DESCRIPTION = "Converts a process into a valid ontology following the BFO 7 Buckets framework."
    • ONTOLOGIES_DIR = "libs/naas-abi-marketplace/naas_abi_marketplace/domains/ontology_engineer/ontologies"
    • TEMPLATE_ONTOLOGY = <ONTOLOGIES_DIR>/BFO7BucketsProcessOntology.ttl
    • SUGGESTIONS: single canned prompt suggestion.
  • External dependencies
    • rdflib.Graph for parsing/serializing Turtle and counting triples.
    • langchain_core.tools.tool for exposing save_ontology as a tool.
    • naas_abi_core.services.agent.Agent (Agent, AgentConfiguration, AgentSharedState)
    • Chat model import used by create_agent:
      • from naas_abi_marketplace.ai.chatgpt.models.gpt_4_1 import model
  • Filesystem behavior
    • Creates ONTOLOGIES_DIR if missing.
    • Writes .ttl files within ONTOLOGIES_DIR.

Tool: save_ontology​

Registered inside create_agent as a LangChain tool (return_direct=True).

Signature:

  • save_ontology(filename: str, turtle: str, overwrite: bool = False, encoding: str = "utf-8") -> str

Behavior:

  • Sanitizes filename:
    • Must be non-empty, no / or \, only letters/digits/._-, ensures .ttl suffix.
  • Validates turtle by parsing with rdflib (format="turtle").
  • Serializes and writes normalized Turtle to <ONTOLOGIES_DIR>/<filename>.
  • If file exists and overwrite=False, returns a message instead of writing.

Usage​

from naas_abi_marketplace.domains.ontology_engineer.agents.SevenBucketsAgent import create_agent
from naas_abi_core.services.agent.Agent import AgentSharedState

agent = create_agent(agent_shared_state=AgentSharedState(thread_id="demo"))

# The agent exposes a tool named "save_ontology" via its tool system (implementation depends on Agent runtime).
# Typically, the LLM will call the tool after generating Turtle content.

Caveats​

  • create_agent() parses TEMPLATE_ONTOLOGY at import-time execution; it will fail if the template file is missing or invalid Turtle.
  • save_ontology enforces strict filename rules and forbids path separators (cannot write outside ONTOLOGIES_DIR).
  • Turtle content must be valid; invalid Turtle raises during rdflib parsing.