Crusoe

Large Language Model services using Crusoe Cloud’s Managed Inference OpenAI-compatible API

Overview

CrusoeLLMService provides chat completion capabilities using Crusoe Cloud’s Managed Inference API with OpenAI-compatible interface. It supports streaming responses and function calling.

Crusoe LLM API Reference

Pipecat’s API methods for Crusoe integration

Example Implementation

Function calling example with Crusoe

Crusoe Documentation

Official Crusoe documentation

Crusoe Cloud

Access models and manage API keys

Installation

To use Crusoe LLM services, install the required dependencies:

uv add "pipecat-ai[crusoe]"

Prerequisites

Crusoe Account Setup

Before using Crusoe LLM services, you need:

1. Crusoe Account: Sign up at Crusoe Cloud
2. API Key: Generate an API key from your account dashboard
3. Model Selection: Choose from available models (default: zai/GLM-5.2)

Required Environment Variables

• CRUSOE_API_KEY: Your Crusoe API key for authentication

Configuration

api_key · str · required

The API key for accessing Crusoe’s Managed Inference API.

base_url · str · default: "https://api.inference.crusoecloud.com/v1/"

The base URL for the Crusoe API. Override if using a different endpoint.

settings · CrusoeLLMService.Settings · default: "None"

Runtime-configurable model settings. See Settings below.

Settings

Runtime-configurable settings passed via the settings constructor argument using CrusoeLLMService.Settings(...). These can be updated mid-conversation with LLMUpdateSettingsFrame. See Service Settings for details.

Parameter	Type	Default	Description
model	str	"zai/GLM-5.2"	Crusoe model identifier. Check Crusoe Cloud for available models.
temperature	float	NOT_GIVEN	Sampling temperature (0.0 to 2.0). Lower values are more focused, higher are creative.
max_tokens	int	NOT_GIVEN	Maximum tokens to generate.
top_p	float	NOT_GIVEN	Top-p (nucleus) sampling (0.0 to 1.0). Controls diversity of output.
frequency_penalty	float	NOT_GIVEN	Penalty for frequent tokens (-2.0 to 2.0). Positive values discourage repetition.
presence_penalty	float	NOT_GIVEN	Penalty for new topics (-2.0 to 2.0). Positive values encourage new topics.

NOT_GIVEN values are omitted from the API request entirely, letting the Crusoe API use its own defaults. This is different from None, which would be sent explicitly.

Usage

Basic Setup

import os
from pipecat.services.crusoe import CrusoeLLMService

llm = CrusoeLLMService(
    api_key=os.getenv("CRUSOE_API_KEY"),
)

With Custom Settings

import os
from pipecat.services.crusoe import CrusoeLLMService

llm = CrusoeLLMService(
    api_key=os.getenv("CRUSOE_API_KEY"),
    settings=CrusoeLLMService.Settings(
        model="zai/GLM-5.2",
        temperature=0.7,
        max_tokens=1000,
    ),
)

Updating Settings at Runtime

Model settings can be changed mid-conversation using LLMUpdateSettingsFrame:

from pipecat.frames.frames import LLMUpdateSettingsFrame
from pipecat.services.crusoe.llm import CrusoeLLMSettings

await worker.queue_frame(
    LLMUpdateSettingsFrame(
        delta=CrusoeLLMSettings(
            temperature=0.3,
        )
    )
)

Notes

• OpenAI Compatibility: Crusoe’s API is OpenAI-compatible, allowing use of familiar patterns and parameters.
• Function Calling: Supports OpenAI-style tool/function calling format.
• Streaming: Supports streaming responses for real-time interaction.
• Developer Role: Crusoe does not support the developer message role. Messages with this role will be automatically converted to system messages.

Event Handlers

CrusoeLLMService supports the following event handlers, inherited from LLMService:

Event	Description
on_completion_timeout	Called when an LLM completion request times out
on_function_calls_started	Called when function calls are received and execution is about to start

@llm.event_handler("on_completion_timeout")
async def on_completion_timeout(service):
    print("LLM completion timed out")