Dapr’s microservices runtime now supports AI agents

Back in 2019, Microsoft Open-Toursod DAPRBuilding distributed is a new runTime to simplify the microservies-based applications. At that time, no one was still talking about AI agents, but it appears that the DAPR had some basic building blocks to support built -in AI agents from the beginning. Because one of the main features of DAPR is a Virtual an idea ActorWhich can accept and process the messages independently from all the other actors in the system.

Today, the DAPR team is launching DAPR agents, it helps developers to create AI agents by providing a lot of building blocks to do this.

DAPR co-creator and maintenance Yaron Snyder explained, “Agents are a very good use case for DAPR.” “From a technical point of view, you can use actors as a very light weight to run these agents and can really be able to run a scale with their state-and be resources. It’s all great, but there are many business arguments that you need to write. Its state of state and its orchestration is only a part of. And many people, they can choose a workflow engine or an actor structure, but on the other hand they still need to do a lot of work to write agents argument. There are a lot of agent framework, but they do not have the same level of orchestration and state party that has DAPR. “

Derived from DAPR agents ShakeA popular Open-Source Project that expands DAPR to use this AI agent. While talking to Microsoft AI researcher Roberto Rodriguez, the two teams decided to bring the project under the DAPR umbrella to ensure the continuity of the new agent structure.

“In different ways we see the agent system and the whole terminology” as another word for the systems distributed, “DAPR co-creator and maintenance Mark Fassel said.[…] Instead of calling them microsurvesis, you can call their agents right now, most because you can keep the big language model in them. “

To coordinate these agents efficiently you need an orchestration engine and state party, the team argues – which provides the DAPR. This is because DAPR actors (and closed, their states, their state, is finished) is very skilled and is meant to be able to spin between milliseconds.

At this point, DAPR agents can talk out of the box with most popular model suppliers. These include AWS Bedrock, Open A, ethnographic, mistral and hug face. Support for local LLMs will come soon.

At the top of the conversation with these models, since DAPR agents extend the existing DAPR structure, developers also have the ability to define a list of equipment that can use to fill a given task after the agent.

Currently, DAPR agents support Python, the .net support will be launched soon. Java, JavaScript and Go will follow soon.