Contributing New LLM Clients

Add support for new LLM providers to MiroFlow by creating a provider class that integrates with the existing client infrastructure.

Client Structure

Each LLM client inherits from LLMProviderClientBase and implements 4 required methods:

• _create_client() - Initialize API client
• _create_message() - Make API calls
• process_llm_response() - Handle responses
• extract_tool_calls_info() - Parse tool calls

---

Implementation Steps

Step 1: Create Provider File

Create src/llm/providers/your_provider_client.py:

Provider Implementation

import dataclasses
from src.llm.provider_client_base import LLMProviderClientBase

@dataclasses.dataclass
class YourProviderClient(LLMProviderClientBase):
    def _create_client(self, config):
        # Initialize your API client
        pass

    async def _create_message(self, system_prompt, messages, tools_definitions, keep_tool_result=-1):
        # Make API call
        pass

    def process_llm_response(self, llm_response, message_history, agent_type="main"):
        # Extract response text, return (text, should_exit)
        pass

    def extract_tool_calls_info(self, llm_response, assistant_response_text):
        # Parse tool calls, return (tool_calls, tool_names)
        pass

Step 2: Create Configuration

Agent Configuration

main_agent:
  llm: 
    provider_class: "YourProviderClient"
    model_name: "your-model"
    your_api_key: "${oc.env:YOUR_API_KEY,???}"
    your_base_url: "${oc.env:YOUR_BASE_URL,https://api.yourprovider.com/v1}"

Step 3: Set Environment Variables

Environment Setup

export YOUR_API_KEY="your-key"
export YOUR_BASE_URL="https://api.yourprovider.com/v1"  # optional if using default

Examples

See existing providers in src/llm/providers/:

• ClaudeAnthropicClient - Direct API
• ClaudeOpenRouterClient - Proxy API
• GPTOpenAIClient - OpenAI API

---

Documentation Info

Last Updated: September 2025 · Doc Contributor: Team @ MiroMind AI