Start the A2A server.

Start the Ollama server if not already running.

Start the A2A-Minions server.

Start the web server.

Stop the A2A-Minions server.

Stream chat responses using SGLang.

Summarizes the metadata for a given query using the local model Args: query: The query to summarize the metadata for

Test multi-turn conversation with context from a file.

Test the summarization of older conversation turns.

Convert text to speech using the MLX Omni Server. Args: text: The text to convert to speech model: The TTS m

Convert to dictionary for serialization.

Context manager to track request metrics.

Context manager to track task execution metrics.

Type keystrokes into a specific application using AppleScript.

Validate Azure OpenAI API key by checking if it's not empty.

Validate HuggingFace API token by checking if it's not empty.

Verify an Azure Attestation JWT token cryptographically. This function: 1. Retrieves signing certificates from the Azure Attestation

Verify an SNP attestation token. Local client verifies it is talking to a machine that is using a SNP TEE. Args: snp_token (

Warn if key is missing and return True if key exists

Run the Minion protocol to answer a task. Args: task: The task/question to answer context: List of c

Run the RLM protocol. Args: task: The task/question to answer. context: List of context strings (will be joi

Main method to execute the Deep Research Minions protocol Args: query: The original user query firecrawl_api_

Run the Arch Router Minion protocol to answer a task. Args: task: The task/question to answer context: List

Execute a development task using the DevMinion protocol. Args: task: The development task description

Run the tunable minion protocol with specified cost sensitivity.

Run the MinionCUA protocol, returning a result dictionary.

Run the minions protocol with MCP tools available

Run the minions protocol to answer a task using local and remote models. Args: task: The task/question to answer doc_

Run the minion protocol to answer a task using local and remote models. Args: task: The task/question to answer conte

Run the minion protocol to answer a task using local and remote models. Args: task: The task/question to answer conte

Run the secure minion protocol to answer a task. Args: task: The task/question to answer context: List of context str

Cleanup on object destruction.

Cleanup when the client is destroyed

Cleanup on destruction.

Initialize the Agno-based local client. Args: model: Agno model instance (HuggingFace, Ollama, OpenRouter, etc.)

Initialize the MinionAgent. Args: remote_client: Client for the remote/supervisor model (e.g., OpenAIClient)

Initialize MinionRLM. Args: remote_client: The powerful remote model client (controller/code writer). local_

Initialize the DeepResearchMinions class Args: local_client: Local client for the Minions protocol remote_cli

Initialize Arch Router Minion. Args: remote_client: Remote supervisor client local_clients: Dictionary of av

Initialize the DevMinion. Args: local_client: Client for the local model (e.g. OllamaClient) remote_

Initialize CostAwareMinion with cost sensitivity level. The CostAwareMinion intelligently balances between local and remote models

Initialize the MinionCUA with local and remote LLM clients.

Initialize MCP config manager Args: config_path: Path to MCP config file. If None, will look in default locations

Initialize with a SyncMCPClient

Initialize SyncMinionsMCP with local, remote LLM clients and MCP.

Initialize the Minion with local and remote LLM clients. Args: local_client: Client for the local model (e.g. OllamaClient)

Initialize the Minion with local and remote LLM clients. Args: local_client: Client for the local model (e.g. OllamaClient)

Initialize the Minion with local and remote LLM clients. Args: local_client: Client for the local model (e.g. OllamaClient)

Initialize the Arch Router model. Args: model_name: HuggingFace model identifier device: Device to load mode

Initialize a conversation history with a maximum number of turns. Args: max_turns: Maximum number of turns to st

Initialize workspace manager. Args: workspace_dir: Path to the workspace directory backup_enabled: W

Initialize the MultiModalEmbedder. :param model_name: name of the local ollama model used to embed text, images, etc.

Initialize a ChromaDB client and collection. :param embedding_model: name of the local ollama model used to embed text, images, etc.

Initialize a Qdrant client and collection. :param embedding_model: name of the local ollama model used to embed text, images, etc.

Initialize the parameter optimizer Args: local_client: Local client for Minions (will create OllamaClient if Non

Initialize the power monitor. :param mode: "mac" to use powermetrics, "nvidia" to use nvidia-smi, or "auto" to detect. :para

Initialize the inference estimator. Args: model_name: Name of the model from the model_to_params table is_qu

Initialize the evaluator with teacher and student models.

Initialize the PII extractor with regex patterns and NLP model.

Initialize the NousResearch client. Args: model_name: The model to use (default: "Hermes-3-Llama-70B")

Initialize the MiniMax client. Args: model_name: The model to use (default: "MiniMax-M2"). Only "MiniMax-M2" is c

Initialize the Vercel AI Gateway client. Args: model_name: The model to use (e.g., "anthropic/claude-sonnet-4", "openai/gpt-4o-mi