Skip to main content

GrokAgent

What it is​

A thin wrapper around naas_abi_core.services.agent.IntentAgent that instantiates a “Grok” intent-based agent with a predefined system prompt, metadata, and a set of intents that route to call_model.

Public API​

  • Constants

    • NAME: Agent display name ("Grok").
    • DESCRIPTION: Human-readable description.
    • AVATAR_URL: URL to an avatar image.
    • SYSTEM_PROMPT: Long-form system prompt that defines behavior/style.
    • SUGGESTIONS: Empty list placeholder.

  • Functions

    • create_agent(agent_configuration: Optional[AgentConfiguration] = None, agent_shared_state: Optional[AgentSharedState] = None) -> IntentAgent
      • Builds and returns a configured GrokAgent.
      • Loads the chat model from naas_abi_marketplace.ai.grok.models.grok_4.
      • Creates a fixed set of Intent entries whose intent_target is "call_model".

  • Classes

    • class GrokAgent(IntentAgent)
      • No additional behavior; inherits all functionality from IntentAgent.

Configuration/Dependencies​

  • Dependencies

    • naas_abi_core.services.agent.IntentAgent:
      • AgentConfiguration, AgentSharedState, Intent, IntentAgent, IntentType
    • Chat model import (required at runtime):
      • from naas_abi_marketplace.ai.grok.models.grok_4 import model

  • Default configuration

    • If agent_configuration is not provided:
      • AgentConfiguration(system_prompt=SYSTEM_PROMPT)
    • If agent_shared_state is not provided:
      • AgentSharedState(thread_id="0")

  • Tools/Agents/Memory

    • tools: empty list
    • agents: empty list
    • memory: None

Usage​

from naas_abi_marketplace.ai.grok.agents.GrokAgent import create_agent

agent = create_agent()

# Interact using IntentAgent's interface (methods depend on IntentAgent implementation).
# Example placeholder:
# response = agent.call_model("search news about space missions")
# print(response)

Caveats​

  • This module does not implement any custom logic in GrokAgent; all behavior depends on IntentAgent and the imported model.
  • The listed intents route to the target string "call_model"; actual dispatch behavior depends on IntentAgent’s implementation.