Deployment

How to run cuvis-ai remotely and integrate it into other services. Deployment is the umbrella; gRPC sits inside it as the wire protocol that powers the deployment story.

If you only need to run a pipeline locally against a .cu3s file, use restore-pipeline instead — no deployment required.

Start here

• Architecture

  ---

  What the cuvis-ai gRPC service is, how it's organised, and the typical request/response lifecycle.

• Deployment Guide

  ---

  Prerequisites, install, TLS, Docker, Kubernetes, monitoring, security, troubleshooting.

• gRPC Workflow

  ---

  Operational walk-through: starting the server, connecting from a client, training and inferring remotely.

API reference

• Session Management

  ---

  Creating and tearing down sessions.

• Configuration API

  ---

  Server- and pipeline-side configuration.

• Pipeline API

  ---

  Load, inspect, and run pipelines remotely.

• Training & Inference

  ---

  Train pipelines and run inference over gRPC.

• Types & Errors

  ---

  Wire types and the error taxonomy.

Client patterns

• :material-pattern: Connections & Sessions

  ---

  Connection management and session lifecycle patterns.

• Workflows & Error Handling

  ---

  Configuration, training, inference, and error-handling patterns.

• Sequence Diagrams

  ---

  Visual workflows for the major operations.