Which deployment models are available?
Presenton supports a spectrum from managed cloud to customer-hosted and air-gapped designs, allowing security and platform teams to select the appropriate operational boundary.
A customer VPC deployment can keep the application and storage under customer administration while using an approved cloud model endpoint. A fully private deployment can pair Presenton with customer-operated models. An air-gapped design removes outbound dependency and requires local alternatives for models, images, fonts, registries, and updates.
Choose based on data classification, provider approvals, latency, capacity, recovery objectives, and internal operational maturity. The strictest architecture is not automatically the best one for every workload.
Which AI models can a private deployment use?
Presenton supports configurable AI providers, including Azure OpenAI, Ollama, and custom OpenAI-compatible endpoints, so organizations can align generation with an approved model strategy.
Model choice affects privacy, quality, latency, capacity, and operating cost. Local models reduce external transfer but require suitable compute and model operations. Private cloud endpoints can offer managed capabilities behind enterprise networking.
Audit image generation and web-search configuration separately. A local text model does not make the workflow fully offline if images, search, fonts, telemetry, or storage still depend on external services.
What does the customer operate?
Customer-hosted deployments require ownership of infrastructure, identity, secrets, networking, storage, observability, backups, scaling, updates, and incident response unless those services are included in an enterprise engagement.
Use versioned images and test upgrades in a non-production environment. Define resource limits, job concurrency, storage lifecycle, log retention, and health alerts. For air-gapped sites, mirror approved artifacts to internal registries and establish a controlled update path.
Presenton can assist with architecture, deployment, integration, and migration planning. Responsibility for each component should be explicit before production launch.
What the platform supports
Enterprise capabilities
Customer VPC
Run the application and storage inside customer-controlled cloud networking.
On-premises
Deploy on internal infrastructure under organizational operating controls.
Air-gapped
Use compatible local dependencies and an offline artifact and update process.
Private models
Connect Azure OpenAI, Ollama, or approved OpenAI-compatible endpoints.
Private storage
Keep source documents and generated artifacts in controlled storage.
Internal API
Expose presentation generation to authorized internal applications.
Decision matrix
Choose the right private deployment
Evaluate data boundary and operational effort together.
| Model | Application location | Model location | Best fit |
|---|---|---|---|
| Customer VPC | Customer cloud | Approved cloud/private endpoint | Controlled cloud adoption |
| Private cloud AI | Customer cloud | Customer GPU or private endpoint | Model and storage control |
| On-premises | Customer data center | Internal or approved endpoint | Internal infrastructure policy |
| Air-gapped | Isolated environment | Local model | No outbound connectivity |
Reference patterns
Deployment architecture patterns
Fully private
Application, model, and files remain inside customer infrastructure.
Customer VPC
Managed model capability through a customer-approved private route.
From evaluation to production
Self-hosted rollout plan
- Step 01
Architecture workshop
Define users, data, providers, network, storage, and availability.
- Step 02
Non-production deployment
Connect identity, models, templates, logs, and backups in a test environment.
- Step 03
Security and load validation
Test data flow, access, egress, throughput, failure behavior, and recovery.
- Step 04
Controlled production rollout
Pilot with one team, monitor quality, then expand under change management.
Evaluation checklist
Questions to resolve before rollout
- Deployment topology
- Compute and concurrency sizing
- Model and image providers
- Identity and service credentials
- Storage and retention
- Outbound network policy
- Monitoring and backups
- Upgrade and rollback process
Buyer questions
Frequently asked questions
Can Presenton run without internet access?+
An air-gapped design is possible when models, image assets, fonts, containers, storage, and updates are available internally. Validate the exact release and required functionality during a proof of concept.
Does self-hosting mean no data leaves our network?+
Only if every configured dependency is internal or disabled. Review language models, image providers, web search, telemetry, fonts, storage, and support access.
Does Presenton support Kubernetes?+
Enterprise deployments can be designed for container orchestration. Confirm the supported deployment artifacts and operating model for your engagement.
Can we begin in cloud and migrate later?+
A phased path can be planned, but migration scope depends on templates, data, identity, integrations, and the selected editions. Define portability requirements before the pilot.
Primary references
Product capabilities and enterprise terms can change. Confirm current edition, deployment, security, support, and contractual details with Presenton before making a purchasing decision.
