> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zeroruntime.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Cerebras

> Use the Cerebras LLM plugin in a Zero Runtime pipeline for fast inference. Setup, options, and usage in Python, JavaScript, and Go.

Cerebras is an **LLM** plugin built for fast inference. It takes the transcribed
conversation and generates the reply, running open models such as Llama and Qwen on
Cerebras hardware.

## Setup

Set your Cerebras API key in the worker environment. Generate a key from the [Cerebras cloud](https://cloud.cerebras.ai/):

```bash theme={null}
export CEREBRAS_API_KEY=<key>
```

## Usage

Import the plugin and pass it to the pipeline's `llm` slot.

<CodeGroup>
  ```python Python theme={null}
  from zrt.plugins import CerebrasLLM

  llm = CerebrasLLM(
      model="llama3.3-70b",
      temperature=0.7,
  )

  # Pipeline(llm=llm, ...)
  ```
</CodeGroup>

## Configuration Options

*Constructor parameters for the Python SDK (`CerebrasLLM`). The JavaScript and Go SDKs expose equivalent options where supported, in their idiomatic form (camelCase / `LLMOptions` struct).*

| Parameter               | Type    | Default          | Description                                                                                   |
| ----------------------- | ------- | ---------------- | --------------------------------------------------------------------------------------------- |
| `api_key`               | `str`   | `None`           | Cerebras API key. Falls back to the `CEREBRAS_API_KEY` environment variable when unset.       |
| `model`                 | `str`   | `"llama3.3-70b"` | Cerebras model used to generate replies (e.g. `gpt-oss-120b`, `zai-glm-4.7`, `llama3.3-70b`). |
| `temperature`           | `float` | `0.7`            | Sampling randomness; lower is more deterministic.                                             |
| `tool_choice`           | `str`   | `"auto"`         | How the model decides when to call tools (`"auto"`, `"none"`, or `"required"`).               |
| `max_completion_tokens` | `int`   | `None`           | Caps the length of each reply.                                                                |
| `top_p`                 | `float` | `None`           | Nucleus-sampling probability cutoff.                                                          |
| `seed`                  | `int`   | `None`           | Seed for reproducible sampling.                                                               |
| `stop`                  | `str`   | `None`           | Stop sequence that halts generation.                                                          |
| `user`                  | `str`   | `None`           | End-user identifier passed to the API.                                                        |

## Import paths

| SDK        | Import                                | Constructor                                 |
| ---------- | ------------------------------------- | ------------------------------------------- |
| Python     | `from zrt.plugins import CerebrasLLM` | `CerebrasLLM(...)`                          |
| JavaScript | `from '@zrt/js-sdk/plugins/cerebras'` | `CerebrasLLM({ ... })`                      |
| Go         | `.../zrt-golang-sdk/plugins/cerebras` | `cerebras.NewLLM(cerebras.LLMOptions{...})` |

The reply is streamed to the [text-to-speech](/plugins/tts/cartesia) plugin, which
synthesizes the agent's voice.