​
How does YAFAI-Core work?

1

Prepare the workspace configuration

This step involves setting up the workspace configuration, which includes defining the agents, their roles, and the environment they will operate in.
2

Spin up the core engine

In this step, the core engine is initialized and started. This involves loading the configuration and preparing the system to handle agent interactions.
3

Connect to gRPC link and talk to Agents

After the core engine is up, it exposes a gRPC service to communicate with the agents. This allows for real-time interaction and coordination among agents.

​
How YAFAI handles the multi Agentic flow?

1

Client connects and sends a message

The client initiates the interaction by connecting to the YAFAI-Core and sending a message. This message typically contains a request or command for the agents.
2

Orchestrator analyses the message and responds

If the intent is out of scope orchestrator resists from responding
If the intent is a general chat like greetings or summarising existing results, orchestrator responsds with a chat
If the user request needs an agent invoke, orchestrator requests an agent invoke and dispatches an agent go routine
1

Agent Analyses the task given by orchestrator

The orchestrator identifies the appropriate agent based on the request and invokes it. This involves sending the request to the agent and waiting for a response.
2

Agent processes the request

Agent fetches the available tools and integrations, if a tool call is needed, the call is made to yafai-skill server.
3

Agent sends back response to orchestrator

Once the agent has processed the request, it sends the response back to the orchestrator.
3

Orchestrator sends back final answer to the client

If the retries are over and a final answer was not found, orchestrator closes the loop.
if the user intent is resolved, the final answer is shared.