YAFAI-Core Documentation
Configuration
Core Engine for Multi Agent Orchestration
YAFAI-Core takes in a directory of configuration file in the below format.
Copy
# Unique name for this configuration
name: "<project_name>" # e.g. "plane_project_management"
# High-level purpose or scope of this configuration
scope: "<overall_scope>" # e.g. "managing projects and issues within the Plane tool"
orchestrator:
name: "<orchestrator_agent_name>" # e.g. "project_manager"
description: "<orchestrator_description>" # e.g. "Oversees project creation, issue tracking, and workflow management."
scope: "<orchestrator_scope>" # e.g. "Help users manage their projects and issues effectively."
model: "<llm_model_name>" # e.g. "llama-3.3-70b-versatile"
provider: "<llm_provider>" # e.g. "groq"
goal: "<orchestrator_goal>" # e.g. "Streamline project execution and issue resolution."
team:
<agent_1_name>:
name: "<agent_1_name>" # e.g. "project_creator"
capabilities: "<agent_1_capabilities>" # e.g. "create new workspaces and projects"
description: "<agent_1_description>" # e.g. "Manages the lifecycle of projects."
model: "<agent_1_model>" # e.g. "deepseek-r1-distill-llama-70b"
provider: "<agent_1_provider>" # e.g. "groq"
goal: "<agent_1_goal>" # e.g. "Enable users to organize work."
depends: "<dependency_agent>" # e.g. "project_manager"
responds: "<responds_to_agent>" # e.g. "project_manager"
<agent_2_name>:
name: "<agent_2_name>" # e.g. "issue_tracker"
capabilities: "<agent_2_capabilities>" # e.g. "create, update, and track issues"
description: "<agent_2_description>" # e.g. "Manages the lifecycle of issues."
model: "<agent_2_model>" # e.g. "deepseek-r1-distill-llama-70b"
provider: "<agent_2_provider>" # e.g. "groq"
goal: "<agent_2_goal>" # e.g. "Ensure all issues are logged and up-to-date."
depends: "<dependency_agent>" # e.g. "project_creator"
responds: "<responds_to_agent>" # e.g. "project_manager"
Implemented LLM providers as of v0.1 are :
groq
Suggested models are,
llama-3.3-70b-versatile
for orchestratordeepseek-r1-distill-llama-70b
for agents
Assistant
Responses are generated using AI and may contain mistakes.