LLM Clients Overview
MiroFlow supports multiple LLM providers through a unified client interface. Each client handles provider-specific API communication while maintaining consistent functionality.
Available Clients
| Client | Provider | Model | Environment Variables |
|---|---|---|---|
ClaudeAnthropicClient |
Anthropic Direct | claude-3-7-sonnet | ANTHROPIC_API_KEY, ANTHROPIC_BASE_URL |
ClaudeOpenRouterClient |
OpenRouter | anthropic/claude-3.7-sonnet, and other supported models | OPENROUTER_API_KEY, OPENROUTER_BASE_URL |
GPTOpenAIClient |
OpenAI | gpt-4, gpt-3.5 | OPENAI_API_KEY, OPENAI_BASE_URL |
MiroThinkerSGLangClient |
SGLang | MiroThinker series | OAI_MIROTHINKER_API_KEY, OAI_MIROTHINKER_BASE_URL |
Basic Configuration
Agent Configuration
main_agent:
llm:
provider_class: "ClientName"
model_name: "model-name"
api_key_param: "${oc.env:API_KEY,???}"
base_url_param: "${oc.env:BASE_URL,default-url}"
Quick Setup
- Set relevant environment variables for your chosen provider
- Update your YAML config file with the appropriate client
- Run:
uv run main.py trace --config_file_name=your_config_file --task="task"
Documentation Info
Last Updated: September 2025 · Doc Contributor: Team @ MiroMind AI