> For a complete index of all SignalWire documentation pages, fetch https://signalwire.com/docs/llms.txt

# set_params

> Configure AI model parameters such as temperature, timeouts, and speech recognition settings.

[ai-params]: /docs/swml/reference/ai/params

[swml-ai-params-reference]: /docs/swml/reference/ai/params

[ref-agentbase]: /docs/server-sdks/reference/python/agents/agent-base

Set multiple AI parameters at once. Merges into any previously set parameters.

These parameters map to the SWML [`ai.params`][ai-params] object.
See the [SWML AI params reference][swml-ai-params-reference] for the full list of
supported fields.

## **Parameters**

Dictionary of parameter name/value pairs.

## **Returns**

[`AgentBase`][ref-agentbase] -- Returns self for method chaining.

***

## AI Parameter Reference

Parameters set via `set_params()` control the AI model, speech recognition, timing,
and agent behavior. The fields below list commonly used parameters by category.

Default values shown below are **server-side defaults** applied by the SignalWire
platform. The SDK itself sends no defaults -- only parameters you explicitly set
are included in the SWML document.

### LLM Parameters

Output randomness. Range: `0.0` -- `2.0`. Lower values produce more deterministic responses.

Nucleus sampling threshold. Range: `0.0` -- `1.0`. Alternative to temperature for controlling randomness.

Repetition penalty. Range: `-2.0` -- `2.0`. Positive values reduce repetition of token sequences.

Topic diversity. Range: `-2.0` -- `2.0`. Positive values encourage the model to explore new topics.

Maximum response tokens. Range: `1` -- `16385`.

AI model to use (e.g., `"gpt-4o-mini"`, `"gpt-4.1-mini"`, `"nova-micro"`, `"nova-lite"`).

### Timing Parameters

Silence duration in milliseconds to detect end of speech. Range: `250` -- `10000`.

Idle delay in milliseconds before the AI re-prompts the caller. Range: `0` -- `600000`.

Inactivity delay in milliseconds before the call is automatically disconnected. Range: `10000` -- `3600000`.

Maximum speech duration in milliseconds before the input is finalized.

### Behavior Parameters

Wait for the caller to speak first before the AI begins talking.

Safety enforcement. When enabled, the AI applies content safety filters.

Transparent barge-in mode. When enabled, caller speech interrupts the AI naturally without discarding context.

Persist the conversation summary after the call ends.

### Audio Parameters

AI voice volume adjustment. Range: `-50` -- `50`.

URL of an audio file to play as background audio during the conversation.

URL of hold music or a tone string (e.g., `"tone:440"`).

## **Example**

```python {5}
from signalwire import AgentBase

agent = AgentBase(name="support", route="/support")
agent.set_prompt_text("You are a helpful assistant.")
agent.set_params({
    "temperature": 0.7,
    "end_of_speech_timeout": 1000,
    "attention_timeout": 10000,
    "wait_for_user": True
})
agent.serve()
```