Configuration | Smallest AI Docs

Configuration is how you define your agent’s behavior—which LLM it uses and how it responds. Get this right and your agent feels natural. Get it wrong and users will notice.

What is Configuration?

Every agent needs two things configured:

LLM Settings — Which model to use, temperature, streaming behavior
Prompts — System instructions that define personality and constraints

The SDK provides sensible defaults, so you can start simple and tune later.

Key Settings

Setting	What It Controls
Model	GPT-4o, Claude, Llama, or your own
Temperature	0.0 = deterministic, 1.0 = creative
Streaming	Essential for real-time responses (always use `stream=True`)
System Prompt	The personality, rules, and context

What’s Next

Prompts

Craft effective system prompts that define behavior.

LLM Settings

Model selection, temperature, and streaming.

Bring Your Own Model

Run local models with Ollama, vLLM, or custom servers.