Intent Engine - Natural-Language Intent Translation - Architecture for Intent-Driven Orchestration in the Compute Continuum

Sep 15, 2026

Microservice placement in edge–cloud compute continua relies on precise service-level objectives (SLOs), yet requiring metric-level constraints creates adoption barriers and increases misconfiguration risk. Direct conversion from natural-language intents to orchestration-consumable SLO artifacts is unreliable for large language models, leading to unsupported constraints, incorrect grounded values, and schema violations that propagate into placement logic and yield infeasible or wrong decisions. Intent Engine introduces a validated intent-to-SLO construction layer using schema-constrained extraction, retrieval-grounded value building from monitored infrastructure state, and constraint validation. Evaluation on a 716-record edge–cloud dataset shows strong improvements over prompting baselines and non-LLM parsing, including higher total F1 and substantially reduced hallucinations and downstream placement failures, while not performing runtime QoS optimization.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages24

Reading Time1 h 0 min

AccessFree to download

Introduction

Compute continuum and microservice placement

Service-level objectives and misconfiguration risks

Intent-driven orchestration and existing limitations

Intent Engine approach preview

Highlights Intent Engine: Natural-Language Intent Translation for Intent-Driven Orchestration in the Compute Continuum Koushikur Islam,Rodrigo N. Calheiros • Translates natural-language service placement intents into orchestration-consumable Service-level Objective (SLO) artifacts. • Provides retrieval-augmented and schema-bounded LLM-based intent-to-specification construction. • Reduces the need for specialized distributed-systems expertise in expressing service placement goals. • Supports adaptable integration with intent-driven placement algorithms and orchestration frameworks. arXiv : 2608 .20388v 1 [cs .CL] 30 Jun 2026 Intent Engine: Natural-Language Intent Translation for Intent-Driven Orchestration in the Compute Continuum Koushikur Islama,∗ , Rodrigo N. Calheirosa a Western Sydney University, Penrith, NSW, Australia | ARTICLE INFO | | AB STRACT | | | --- | --- | --- | --- | | Keywords:
Intent-Driven Orchestration Service Placement Microservices
Service-level Objectives (SLOs) Edge-Cloud Continuum
Large Language Model (LLM) | | Microservice placement in the compute continuum is driven by low-level Service-level Objectives (SLOs), but requiring users to specify metric-level constraints creates an adoption barrier and increases misconfiguration risk. Although large language models (LLMs) can interpret naturallanguage intents, direct generation of orchestration-consumable SLO artifacts remains unreliable due to unsupported constraints, incorrect grounded values, and schema violations. These errors can propagate to downstream placement logic and produce infeasible or incorrect placements. This paper presents Intent Engine, a natural-language intent translation architecture that constructs validated SLO artifacts for compute-continuum service placement. Intent Engine acts as an intent acquisition and SLO construction layer for existing intent-driven orchestration and placement frameworks; it does not perform placement or runtime QoS optimization. The architecture combines schema-constrained extraction, retrieval-grounded value construction from monitored infrastructure state, and validation against supported constraints before emitting the final SLO artifact. We evaluate Intent Engine using a 716-record intent-to-SLO dataset derived from an edge–cloud testbed, including valid and invalid intents. Across GPT-4.1 mini, Claude Sonnet 4.5, and DeepSeek V4-Flash, Intent Engine outperforms prompting baselines and a non-LLM rule-based parser. With GPT-4.1 mini, it achieves 0.941 total F1 Score and reduces aggregate hallucination by 85.1%, while lowering downstream placement failure from 30 . 8% to 2 . 1% . | | | 1. Introduction
Software applications have driven a paradigm shift toward microservices, enabling improved scalability, flexibility, resilience, and faster delivery. By decomposing software into independently deployable components, microservices support dynamic scaling, rapid rollout, and reduced latency [1] . Modern microservice-based systems primarily rely on cloud infrastructures to achieve scalability, cost efficiency, and availability; however, centralized data centers introduce network latency due to limited geographic proximity to users [2] . This limitation has led to the emergence of edge computing.
Edge computing [3] mitigates latency by processing data closer to its source in real time. Nevertheless, resource constraints at the edge restrict scalability, computational capacity, and availability compared to cloud environments. To balance these trade-offs, the compute continuum integrates edge and cloud resources, enabling microservice deployment across heterogeneous infrastructures.
The compute continuum spans endpoints, edge, and cloud layers, distributing services to reduce latency while improving throughput and scalability [4] . Within this environment, microservice placement becomes critical for meeting user requirements such as low latency, high availability, stora

FAQs

What problem does Intent Engine address in compute-continuum microservice placement?

It addresses the gap between user-friendly natural-language intents and the need for correct, orchestration-consumable SLO artifacts. Without reliable translation, misconfigurations can propagate into placement logic and cause infeasible or incorrect placements.

Why is direct natural-language generation of SLO artifacts unreliable with LLMs?

Direct generation can produce unsupported constraints, incorrect grounded values, and schema violations. These errors can then propagate downstream and lead to wrong or infeasible placement decisions.

What does Intent Engine do when constructing the final SLO artifact?

It performs schema-constrained extraction, builds retrieval-grounded values from monitored infrastructure state, and validates against supported constraints before emitting the final SLO artifact.

Sep 15, 2026

Upload by Seraphina