OpenAI provider

The NucleusIQ OpenAI provider (nucleusiq-openai) gives you full access to Chat Completions and Responses API with a simple agent interface.

Quick start

from nucleusiq_openai import BaseOpenAI

llm = BaseOpenAI(model_name="gpt-4o-mini")
agent = Agent(..., llm=llm)

Set OPENAI_API_KEY in your environment.

Three-layer design

NucleusIQ keeps agent creation simple while giving power users full control:

Layer	Who	Control
Agent	90% of users	Model name, basic config
LLM Config	Power users	AgentConfig.llm_params — temperature, max_tokens, etc.
Direct API	Advanced	Per-task execute(..., llm_params=OpenAILLMParams(...)) overrides

Native tools

The OpenAI provider supports native tool types:

• function — Custom Python functions
• code_interpreter — Code execution
• file_search — File search
• web_search — Web search
• image_generation — DALL·E
• mcp — Model Context Protocol
• computer_use — Computer use

Streaming and usage

• Streaming — execute_stream() works with both Chat Completions and Responses API
• Usage tracking — agent.last_usage.display() shows prompt/completion/reasoning tokens

Full guide

For the complete design guide (parameter merge chain, all 40+ parameters, implementation status), see openai_provider_user_guide.md in the repo.