Planned architecture

Cloud-native, modular, and cost-controlled by design.

ScaleVane is being engineered as a multi-tenant SaaS platform with provider-flexible AI, reliable background processing, controlled media workloads, and measurable unit economics.

This page describes the intended production architecture. Components marked as planned are not represented as already deployed.
FE
Experience layer

Next.js, TypeScript, responsive dashboard, onboarding, workspaces, approvals, and subscription controls.

API
Application services

Secure APIs, tenant isolation, usage metering, background queues, integrations, and audit logs.

AI
Intelligence services

Gemini/Vertex AI primary path, optional specialist models, provider abstraction, prompt and cost logging.

Technical decisions

Infrastructure choices tied to product requirements.

Credits will be used only where the workload requires them, not to create unnecessary cloud complexity.

Web application

Next.js and TypeScript for the SaaS dashboard, with an independently deployable static marketing website.

Data and tenancy

PostgreSQL, workspace-scoped access controls, encrypted connections, backups, and tenant isolation policies.

AI provider layer

A provider abstraction prevents permanent dependence on one model and allows quality-cost routing by task.

Workflow orchestration

n8n for early orchestration, moving high-volume critical paths into dedicated workers when justified.

Background processing

Job queues for publishing, analytics ingestion, notifications, retry handling, and long-running generation.

Media processing

FFmpeg workers and object storage for practical media transformation, with strict limits and lifecycle policies.

Observability

Structured logs, error monitoring, audit trails, model usage records, and cloud-cost alerts.

Security controls

Least privilege, secrets management, secure OAuth, account revocation, and human publishing approval.

Commercial controls

Per-plan usage limits, budget caps, provider fallback rules, and margin monitoring before scale.

Intended cloud workloads

Clear use cases for startup-credit programmes.

The following workloads map directly to the product roadmap and expected customer activity.

WorkloadPurposeExpected phase
Application APIsWorkspace, content, approval, subscription, and integration servicesMVP
Managed PostgreSQLMulti-tenant product data, audit logs, and usage recordsMVP
AI inferenceStrategy, content generation, transformation, and recommendationsMVP onward
Object storageMedia assets, generated documents, and user uploadsMVP onward
Background workersPublishing, retries, notifications, analytics, and batch processingPhase 2
Media workersFFmpeg-based audio/video processingPhase 3
Monitoring and securityLogs, alerts, metrics, secrets, and auditabilityAll phases