llama_deploy

Why we included this project

Anyone who has strung together a multi-step agent pipeline with LlamaIndex knows the jump from a workflow that runs on their laptop to one that actually serves requests is the hard part. llama_deploy addresses that by adding a message queue and coordination layer, letting individual agents or workflow steps run as separate services that talk to each other reliably. The honest caveat is that the project has been deprecated, and the maintainers point to llama-agents (workflows-py) as the actively maintained successor for serving workflows. Treat it as an architectural reference rather than a foundation: it is worth reading to see how the authors approached durable, distributed agent serving, but it is not where you would start a fresh production build. For teams already in the LlamaIndex ecosystem, it is a useful way to understand the design space before moving on to the successor.