The enterprise software engineering platform for complex codebases

KAVIA provisions a dedicated, customer-isolated VPC for your organization within KAVIA-managed AWS infrastructure. KAVIA connects only to repositories authorized by your organization and creates temporary working copies within that isolated environment for code ingestion, system understanding, workflow execution, and artifact generation.

KAVIA encodes and writes durable project knowledge and generated artifacts to repositories designated by your organization in its SCM. Temporary customer content is retained only for the period configured by the customer, never longer than seven days, and is then deleted.

KAVIA provides access to approved models and automatically selects and routes them according to the task, required quality, performance, and reliability. Inference is performed through approved external model-serving endpoints governed by enterprise no-training and zero-data-retention controls.

This configuration uses the same dedicated, customer-isolated KAVIA Cloud architecture as Managed Model Routing. KAVIA processes approved repository content within a dedicated VPC, writes durable project knowledge and generated artifacts to repositories designated by the customer, and deletes temporary customer content after the configured retention period, which may not exceed seven days.

Customer administrators and authorized users gain additional control over approved model selection, compute allocation, consumption policies, and usage reporting. Teams may select models for particular workflows or use KAVIA-managed routing while retaining visibility into model use and allocation.

The application-processing and inference path is the same as Managed Model Routing. This configuration changes model and usage governance. It does not create a different hosting or security boundary.

Persistent repository clones, durable outputs, and selected processed artifacts are stored in customer-controlled AWS storage, such as Amazon EFS. KAVIA-managed agents and services access approved source code during active workflows to perform ingestion, code understanding, knowledge creation, orchestration, and artifact generation.

KAVIA provides access to approved models and manages model selection and routing according to the selected plan. Inference is performed through KAVIA-approved external model-serving endpoints governed by enterprise no-training and zero-data-retention controls.

This is a split-boundary architecture: persistent storage remains within the customer's AWS account, while active KAVIA application processing and model inference are delivered through KAVIA-managed services.

Persistent repository clones, durable outputs, and selected processed artifacts are stored in customer-controlled AWS storage. KAVIA-managed services perform active ingestion, system understanding, workflow orchestration, and artifact generation against the approved project scope.

The customer selects, contracts for, and governs the model endpoint used for inference. The endpoint may be hosted within the customer’s environment or provided through a customer-contracted cloud model platform. KAVIA connects to that endpoint through an approved interface and routes the context required to complete the selected task.

This configuration gives the customer direct control over the model provider and inference endpoint, while KAVIA continues to operate the application-processing layer.