User Guide¶
Use this guide to select a provider, choose a workflow class, attach provider-compatible tools, and run or stream the result.
Choose a path¶
| Goal | Guide |
|---|---|
| Install, authenticate, and run a first Minion | Getting Started |
| Add hosted search, code, file, or image tools | Provider Tools |
| Use async execution and provider-native streaming | Execution and Streaming |
| Create a reusable workflow specialization | Custom Minions |
The core pattern¶
from minions.gpt import ResearchMinion
minion = ResearchMinion(
model='gpt-5.6-sol',
instructions='Research the question and distinguish evidence from conclusions.',
)
result = minion.run( 'Explain the requested subject.' )
print( result.final_output )
Tools are optional. When tools are used, import them from the same provider pathway as the Minion.
Keep a workflow provider-pure
Do not pass an OpenAI tool to a Gemini Minion, a Gemini tool to a Claude Minion, or otherwise mix provider contracts. Minions preserves provider-native behavior rather than translating schemas.