# Model Inference

Record: `arch:model-inference`
Kind: capability
Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md)
Severity: contextual
Scope: service, model serving
Canonical: https://banes-lab.com/ontology#arch-model-inference

## Repair

- Refactored by: Add Inference Service, Add Adapter/Contract
- Detected by: raw model calls in business logic
- Violated by: inference without validation/observability/fallback
- Measured by: [latency](https://banes-lab.com/records/arch/latency.md), error rate, output quality
- Enforced by: serving standards

## requires

- [Model Artifact](https://banes-lab.com/records/lex/model-artifact.md)
- [Input/Output Contract](https://banes-lab.com/records/lex/input-output-contract.md)

## reinforces

- [Artificial Intelligence Architecture](https://banes-lab.com/records/arch/artificial-intelligence-architecture.md)

## enables

- [Runtime Prediction/Generation](https://banes-lab.com/records/lex/runtime-prediction-generation.md)

## conflicts-with

- [Training-Time-Only Model Logic](https://banes-lab.com/records/lex/training-time-only-model-logic.md)

## tensions-with

- [Latency/Cost](https://banes-lab.com/records/lex/latency-cost.md)

## tensions

- [Model Inference Latency/Cost](https://banes-lab.com/records/tension/latency-cost-model-inference.md)

## referenced-by

- [Prompt Engineering](https://banes-lab.com/records/arch/prompt-engineering.md)
- [Agentic Architecture](https://banes-lab.com/records/arch/agentic-architecture.md)