Find answers from the community

Updated 6 months ago

I have one question regarding llama-

I have one question regarding llama-agent. I looked into the code. It looks like the orchestration and the agent services are tightly coupled. So, if I have to use the Agent Service in aother orchestration, I will have to duplicate the Agentservice and the processing of the agent. is that understanding correct?

9 comments

LLogan M

I don't think I fully understand

RRajib

I mean, if I have to deploy and develop, I need to touch both the orchestrator and the agent code. the orchestrator is not a seperate component

RRajib

launcher = ServerLauncher(
[agent_server_1, agent_server_2],
control_plane,
message_queue,
additional_consumers=[human_consumer]
)

RRajib

like in this code, the orchestrator is part of control plane

RRajib

In a typical microservices, you have a atomic layer and the composite layer

RRajib

each are decoupled from each other, the cpmposite layer has the orchestration logic

RRajib

the atomic layer has the atomic services logic

LLogan M

the composite layer here is the control plane and message queue I suppose

The atomic layer is all the services

You can launch each independently, and add/remove services over time as well by registering/deregistering with the control plane

LLogan M

so the orchestration and services are technically coupled, but not that strongly either

Add a reply