AI Processor Types

AI Processors are categorized by where they operate in the workflow. The following types are currently available:

Pack Pipeline

The Pack Pipeline consists of three sequential stages that generate, validate, and finalize configurations:

CONFIG

The CONFIG stage generates initial implementation-ready configurations based on service item declarations, such as JSON configuration files for network devices, infrastructure-as-code templates, policy definitions, and rule sets. This is where the primary transformation occurs — converting high-level service requirements into vendor-specific configuration files.

VERIFY

The VERIFY stage validates and reviews generated configurations. It can be used for generating validation test cases, reports, identifying potential issues, performing security assessments, and recommending configuration improvements. This optional stage provides quality assurance before configurations are applied to production infrastructure.

The VERIFY stage uses an independent context window separate from the CONFIG stage, and the LLM Model for each stage can be the same or different. This separation ensures that the validation is performed with a fresh context, improving accuracy and avoiding potential bias from the generation process.

EXECUTION

The EXECUTION stage performs the final testing, quality assessment, and preparation of configurations for deployment. This optional stage provides the final layer of confidence before production deployment.

For a deeper look at how these stages chain together and how data flows between them, see Pack Pipeline.

Change Instance Validator

When a Consumer submits a declaration and change instances enter a pending state, the Service Owner must approve or reject them — a step known as validation. This can be done manually, through scripting, or by using an AI Processor as a Change Instance Validator.

The Change Instance Validator passes the consumer's requested declaration and the service's configuration to an LLM, which evaluates whether the request should be approved or rejected. This is useful for enforcing validation rules that go beyond what JSON Schema or regex validation can express. While NetOrca already validates consumer submissions at the submission level using JSON Schema, some business logic, policy constraints, or cross-field dependencies require a deeper level of assessment.

The Change Instance Validator acts as an automated validation layer between the consumer's submission and the actual processing of their intent on the Service Owner's infrastructure.

For setup instructions and configuration options, see Change Instance Validator.

Pack Optimizer

The Pack Optimizer is an AI Processor type that optimizes prompts to reduce the number of retriggers in self-healing loops. It will be available when the Universal Executor is enabled.