Strategy Guide
NucleusIQ strategy starts with selecting the right execution mode.
Gearbox strategy
| Mode | Best for | Default tool budget |
|---|---|---|
| DIRECT | Fast answers, simple requests | 25 calls / run (configurable) |
| STANDARD | Most production workflows | 80 calls / run (configurable) |
| AUTONOMOUS | Complex/high-stakes workflows | 300 calls / run (configurable) |
from nucleusiq.agents.config import AgentConfig, ExecutionMode
config = AgentConfig(execution_mode=ExecutionMode.STANDARD)
Selection heuristic
- Start with STANDARD for most use cases.
- Use DIRECT when latency and cost are critical and workflow is simple.
- Use AUTONOMOUS for decomposition/verification-heavy tasks.
Production guardrails
Combine mode choice with plugins:
from nucleusiq.agents import Agent
from nucleusiq.agents.config import AgentConfig, ExecutionMode
from nucleusiq.prompts.zero_shot import ZeroShotPrompt
from nucleusiq.plugins.builtin import ModelCallLimitPlugin, ToolRetryPlugin
from nucleusiq_openai import BaseOpenAI
agent = Agent(
name="strategy-guardrails",
prompt=ZeroShotPrompt().configure(system="You are a helpful assistant."),
llm=BaseOpenAI(model_name="gpt-4.1-mini"),
plugins=[
ModelCallLimitPlugin(max_calls=15),
ToolRetryPlugin(max_retries=2),
],
config=AgentConfig(execution_mode=ExecutionMode.STANDARD),
)