Retvest Deploy / Deployment
Your organisation chooses the infrastructure. Retvest Deploy brings the company AI environment and included Retvest Intelligence Layer to the agreed server and model setup.
Compare the environments
| What to plan | On premises | Your company cloud |
|---|---|---|
| Infrastructure | Company-owned Linux hosts and storage | Linux VMs and storage in your company’s cloud account |
| Company boundary | Your premises, network and access arrangements | Your cloud account, selected network and region |
| Model placement | On the application host or a reachable dedicated model/GPU server | On your cloud model host or another explicitly connected company endpoint |
| Operating responsibility | Agree hardware, network, patching, backups and recovery owners | Agree account access, VM costs, network, patching, backups and recovery owners |
| The Deploy package | Employee tools, agreed integrations, agents and RIL | The same package, with the cloud profile agreed for your pilot |
| External connections | Document any licence, certificate, download or optional service traffic | Document provider services, model paths and optional outbound connections |
These are deployment-planning paths, not a claim of a qualified installation, managed hosting service or air-gapped operation. Cloud, model and infrastructure charges belong to the chosen environment.
When the model needs its own hardware
The company application and model endpoint can run on separate hosts. Plan this around GPU capacity, latency, network access and the information sent between them.
A split on-premises/cloud arrangement requires an agreed data path and operating profile. It does not automatically establish an approved hybrid or air-gapped deployment.
Model connections and data paths
A deployment is an operating environment
Identify the Linux hosts, HTTPS address, persistent storage, model endpoint and permitted users. Define the workload before sizing.
Agree the tools, knowledge sources, integrations and RIL record required for the pilot. Name the people who accept the outcome.
Assign responsibility for access, retention, backups, upgrades and recovery. Qualify the hardware and workload before committing capacity.
Planning your installation
Define these responsibilities with your infrastructure team.
The application requires a dedicated Linux Docker host and persistent storage. Model memory and GPU needs depend on the model and workload; plan them separately from application requirements.
Infrastructure requirementsAgree HTTPS, user access, model connectivity and approved outbound traffic. A company-server installation does not by itself establish an air-gapped environment.
Model endpoint configurationName the people responsible for operating the environment. Keep configuration and state outside release directories, and agree the backup and recovery plan for your pilot.
Operating requirementsThe web workspace, terminal and developer connections sit alongside RIL, the included system of record for AI work.
Explore DeployRetvest Deploy / Walkthroughs & pilots
Bring one workflow. We’ll discuss the environment, model and work record it needs.