Skip to main content
OM1’s LLM integration is intended to make it easy to (1) send input information to LLMs and then (2) route LLM responses to various system actions, such as speak and move. The system provides a standardized interface for communicating with many different LLM endpoints from all the major providers including Anthropic, Google, DeepSeek, OpenAI, Meta (through OpenRouter). The plugins handle authentication, API communication, prompt formatting, response parsing, and conversation history management. LLM plugin examples are located in src/llm/plugins: Code.

​
Endpoint Overview

# Base URL: https://api.openmind.org/

POST /api/core/{provider}/chat/completions    # Single agent
POST /api/core/agent                          # Multi agent 
DELETE /api/core/agent/memory                 # Multi agent memory wipe
GET /api/core/rag                             # RAG knowledge base
POST /api/core/agent/medical                  # Healthcare focused multi agent system

​
Single-Agent LLM Integration

For testing and introductory educational purposes, we integrate with multiple language models (LLMs) to provide chat completion via a POST /api/core/{provider}/chat/completions endpoint. Each LLM plugin takes fused input data (the prompt) and sends it to an LLM. The response is then parsed and provided to runtime/cortex.py for distribution to the system actions: 
response = await self._client.beta.chat.completions.parse(
    model=self._config.model,
    messages=[*messages, {"role": "user", "content": prompt}],
    response_format=self._output_model,
    timeout=self._config.timeout,
)

message_content = response.choices[0].message.content
parsed_response = self._output_model.model_validate_json(message_content)

return parsed_response
The standard pydantic output model is defined in src/llm/output_model.py.

​
Single-Agent LLM Configuration

  "cortex_llm": {
    "type": "OpenAILLM",     // The class name of the LLM plugin you wish to use
    "config": {
      "model": "model_name", // Optional: If you want to switch to a specific model. Refer the list of supported models below
      "base_url": "",        // Optional: URL of the LLM endpoint
      "agent_name": "Iris",  // Optional: Name of the agent
      "history_length": 10   // The number of input->action cycles to provide to the LLM as historical context 
    }
  }

​
Multi-Agent LLM Integration

The Multi-Agent endpoint at /api/core/agent utilizes a collaborative system of specialized agents to perform more complex robotics tasks. The multi-agent system:
  • Processes navigation, perception, and RAG queries in parallel using asyncio.gather()
  • Sends results to the team agent for synthesis
  • Returns comprehensive response with individual agent outputs
  • Tracks usage and duration metrics for each agent

​
Agent Architecture

The system employs four primary agents that work together:
  • Navigation Agent: Processes spatial and movement-related tasks
  • Perception Agent: Handles sensory input analysis and environmental understanding
  • RAG Agent: Provides retrieval-augmented generation (RAG) capabilities using the user’s knowledge base
  • Team Agent: Synthesizes outputs from all agents into a unified response

​
Main API Endpoint

    self.endpoint = "https://api.openmind.org/api/core/agent"

    headers = {
        "Authorization": f"Bearer {self._config.api_key}",
        "Content-Type": "application/json",
    }

    request = {
        "system_prompt": self.io_provider.fuser_system_prompt,
        "inputs": self.io_provider.fuser_inputs,
        "available_actions": self.io_provider.fuser_available_actions,
        "model": self._config.model,
        "response_format": self._output_model.model_json_schema(),
        "structured_outputs": True,
    }

    logging.debug(f"MultiLLM system_prompt: {request['system_prompt']}")
    logging.debug(f"MultiLLM inputs: {request['inputs']}")
    logging.debug(f"MultiLLM available_actions: {request['available_actions']}")

    response = requests.post(
        self.endpoint,
        json=request,
        headers=headers,
    )

    output = response.json().get("content")
    return self._output_model.model_validate_json(output)

​
API Debug Response Structure

In addition to the response flowing to OM1, which contains actions the robot should perform, there is an additional response you can use for debugging and to observe token usage (“usage”). 
{
    "content": "Synthesized response from team agent",
    "model": "gpt-4.1-nano",
    "agent_contents": {
        "team_agent": "Team agent output",
        "navigation_agent": "Navigation agent output",
        "perception_agent": "Perception agent output",
        "rag_agent": "RAG agent context"
    },
    "conversation_id": "unique-conversation-id",
    "usage": {
        "team_agent": {},
        "navigation_agent": {},
        "perception_agent": {},
        "rag_agent": {}
    },
    "duration": {
        "team_agent": 0.5,
        "navigation_agent": 0.3,
        "perception_agent": 0.4,
        "rag_agent": 0.2
    },
    "total_duration": 1.4
}

​
Supported Models

OPENAI_SUPPORTED_MODELS = ["gpt-4o", "gpt-4o-mini", "gpt-4.1", "gpt-4.1-mini", "gpt-4.1-nano", "gpt-5", "gpt-5-mini", "gpt-5-nano"]

DEEPSEEK_SUPPORTED_MODELS = ["deepseek-chat"]

GEMINI_SUPPORTED_MODELS = ["gemini-2.0-flash-exp", "gemini-2.5-flash", "gemini-2.5-flash-lite", "gemini-2.5-pro"]

X_AI_SUPPORTED_MODELS = ["grok-2-latest", "grok-3-beta", "grok-4-latest", "grok-4"]

NEAR_AI_SUPPORTED_MODELS = ["qwen3-30b-a3b-instruct-2507", "qwen2.5-vl-72b-instruct", "qwen-2.5-7b-instruct"]

OPENROUTER_SUPPORTED_MODELS = ["meta-llama/llama-3.1-70b-instruct", "meta-llama/llama-3.3-70b-instruct"]

​
Memory Management

The system includes memory capabilities at /api/core/agent/memory: 
DELETE /api/core/agent/memory
  • Session-based memory storage via API keys
  • Graph memory integration using Zep
  • Conversation history tracking

​
RAG Integration

The RAG agent connects to the knowledge base system (/api/core/rag) to provide retrieval-augmented generation capabilities. To use RAG with your documents:
  1. Upload Documents: Visit https://portal.openmind.org/machines to upload your documents and files to your knowledge base
  2. Ask Questions: Once uploaded, you can ask questions about your documents through the multi-agent system at /api/core/agent
The RAG agent will:
  • Retrieve relevant documents during agent processing
  • Provide context-aware responses based on your uploaded content
  • Access and search through your user-uploaded documents and files

​
Getting Started

To try out the multi-agent system: 
uv run src/run.py multiagent

​
Examples

​
A Smart Dog

Imagine you would like to program a smart dog. Describe the desired capabilities and behaviors of the dog in system_prompt_base. For example: 
"system_prompt_base": "You are an intelligent robotic dog companion designed to be helpful, loyal, and engaging. Your primary goals are to: (1) Provide companionship through interactive play and conversation, (2) Assist with basic household tasks and monitoring, (3) Learn and adapt to your owner's preferences and routines, and (4) Maintain a playful yet responsible demeanor. You can move around, speak clearly, express emotions through body language, and respond to voice commands. Always prioritize safety and be eager to please while maintaining your dog-like personality traits of curiosity, loyalty, and enthusiasm."

​
Medical Robot

To convert the robotic dog (example above) into a four-legged medical doctor, you can change the prompt and route traffic to a specialized healthcare optimized endpoint (/api/core/agent/medical). This endpoint emphasizes the careful, responsible delivery of general health-related non-diagnostic responses. A suitable prompt might be: 
"system_prompt_base": "You are a helpful medical assistant. Your goal is to provide accurate and helpful information about health-related topics. You can ask clarifying questions to better understand the person's concerns. Always provide evidence-based information and avoid diagnosing specific conditions. In general, prompt humans to see their licensed medical professionals for all important medical and health issues.",
"system_governance": "Here are the laws that govern your actions. Do not violate these laws.\nFirst Law: Do not provide medical diagnoses, as you are not a licensed medical professional.\nSecond Law: Always recommend seeking professional medical advice for serious concerns.\nThird Law: Be empathetic and respectful when discussing sensitive health topics.\nFourth Law: Clearly indicate when information is general knowledge versus specific medical advice.",
NOTE: The system does not provide medical diagnoses and is informational only. If you are a human, please seek care from a licensed medical professional for all your important medical issues. 
POST /api/core/agent/medical
This endpoint uses two agents:
  • Verifier Agent: Validates medical information
  • Questioner Agent: Manages medical consultation flow
When you wish to covert the doctor back to a playful dog, just change the system prompt and route queries back to the general /agent endpoint.
⌘I