Urgent.News

What's breaking now, across thousands of outlets.

Tech

OpenIntent Protocol: Um Protocolo Declarativo para Comunicação entre Agentes Baseada em Intenções

À medida que ecossistemas de agentes se tornam mais distribuídos, um dos principais desafios deixa de ser a implementação da lógica de negócio e passa a ser a interoperabilidade entre agentes desenvolvidos por diferentes organizações, linguagens e infraestruturas. O OpenIntent Protocol (OIP) propõe uma abordagem declarativa onde uma intenção é tratada como um contrato formal de execução,…

The OpenIntent Protocol is a declarative protocol for communication between agents based on intentions. As distributed agent ecosystems become more complex, a major challenge shifts from implementing business logic to interoperability between agents developed by different organizations, using various languages and infrastructures.

The OpenIntent Protocol proposes an approach where an intention is treated as a formal contract of execution, describing not just the desired capability, but also how this interaction should occur. This eliminates the need for agents to know communication details, allowing different implementations to be interoperable from a common specification.

Currently, the OpenIntent Protocol is in development as part of the H2A2H architecture (v0.3), with its specification still evolving and potentially changing until version 1.0. Currently, when two services or agents need to communicate, they typically need to define protocol, authentication, message serialization, error handling, discovery of the destination, and streaming or events, which tightly couples application logic to infrastructure.

For example, a typical agent communication chain is: Agent ↓ gRPC Client ↓ mTLS ↓ ProtoBuf ↓ Load Balancer ↓ Server. If communication shifts to NATS or QUIC, the agent may need to be modified. The OpenIntent Protocol aims to eliminate this coupling. The OpenIntent Protocol is based on the principle that an agent should declare what it wants to achieve, not how it should communicate.

The infrastructure already knows the intention contract. The agent only produces or consumes intentions. The protocol is divided into four main parts: Intent, Capability, Channel, and Contract. Intent defines the semantic operation, such as validating a payment, consulting inventory, issuing an invoice, or reserving a hotel. Capability defines the necessary capacity to execute the intention, such as PaymentValidation or InventoryReservation.

Each intention may have one or more mandatory capabilities. Channel formally defines how the intention should be transmitted, with the protocol specifying rather than the agent choosing. Contract describes inputs, outputs, version, compatibility, policies, security, and more. The OpenIntent Protocol supports different communication models like Request/Response, Streaming, Event-based communication, and Fire-and-Forget.

The SDK automatically generates clients, serializers, authentication, connections, handlers, and validation based on the declarative definition of the OpenIntent Protocol. When an intention is executed or published, the infrastructure is materialized by the SDK, ensuring deterministic behavior across different implementations of the protocol.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in Tech

Monitor Website Response Time with cURL

Discover how to measure your website's response time using the curl command, and how it can help you increase traffic and performance of your webpage.

  • cURL is used to measure website response time
  • Command curl -s -o /dev/null -w %{timestarttransfer} https://www.enstai.fi/ is employed
  • Response time of 0.152016 seconds is below Google's optimal threshold of 0.2 seconds

How to build a tiny 1.5B text-to-SQL model that beats a 7B

I wanted to build something with an LLM using my own hands. Not wire an API into a wrapper, but take a base model, train it, measure it, break it, and serve it over HTTP. Why?

  • Author builds tiny 1.5B text-to-SQL model to outperform 7B model
  • Model achieves 44.6% accuracy with held-out evaluation, 49.7% with RL reward
  • 1.5B model still inferior to 7B but answers queries eight times to vote on most common result

More from Friday 4 September →