Prompts
NucleusIQ includes a prompt-engineering module with a factory-based API.
PromptFactory and PromptTechnique
Use PromptFactory to create prompt strategies without changing agent code.
from nucleusiq.prompts import PromptFactory, PromptTechnique
prompt = PromptFactory.create_prompt(PromptTechnique.ZERO_SHOT).configure(
system="You are a precise analyst.",
user="Answer clearly and include assumptions.",
)
Built-in techniques
| Technique | Enum value |
|---|---|
| Zero-shot | PromptTechnique.ZERO_SHOT |
| Few-shot | PromptTechnique.FEW_SHOT |
| Chain-of-thought | PromptTechnique.CHAIN_OF_THOUGHT |
| Auto chain-of-thought | PromptTechnique.AUTO_CHAIN_OF_THOUGHT |
| Retrieval augmented generation | PromptTechnique.RETRIEVAL_AUGMENTED_GENERATION |
| Prompt composer | PromptTechnique.PROMPT_COMPOSER |
| Meta prompting | PromptTechnique.META_PROMPTING |
Use with Agent
from nucleusiq.agents import Agent
from nucleusiq.prompts import PromptFactory, PromptTechnique
from nucleusiq_openai import BaseOpenAI
agent = Agent(
name="researcher",
role="Research assistant",
objective="Find and synthesize relevant information",
llm=BaseOpenAI(model_name="gpt-4o-mini"),
prompt=PromptFactory.create_prompt(PromptTechnique.ZERO_SHOT).configure(
system="Be concise and accurate.",
user="Use evidence from available context and tools.",
),
)
Extending with custom techniques
Register your own prompt class if you need a custom strategy:
from nucleusiq.prompts import PromptFactory, PromptTechnique
# PromptFactory.register_prompt(PromptTechnique(...), CustomPromptClass)
See also
- Agents — Agent configuration and lifecycle
- Execution modes — How prompts interact with mode behavior
- OpenAI provider guide — Provider-specific behavior