Dynamic Showcase Environments: Auto-Provisioning the Agentic Experience

Quick Links

1. 1.Monolithic Pipelines vs. Dynamic Orchestration
2. 2.The Auto-Provisioning Engine
3. 3.TECHNICAL IMPLEMENTATION: DYNAMIC ROUTE INJECTION
4. 4.Architectural Elegance

Monolithic Pipelines vs. Dynamic Orchestration

Demonstrating Agentic AI demands more than static SPAs (Single Page Applications). It requires spinning up live, interactive computational environments. Previously, deploying these environments meant relying on monolithic CI/CD pipelines—manually cloning repos, configuring environment variables, and waiting for Vercel/Netlify build hooks to resolve. This approach was highly unscalable for sales teams needing rapid, on-the-fly deployments.

To solve this, we overhauled our Showcase Management architecture, moving from static build pipelines to a fully dynamic Auto-Provisioning engine.

The Auto-Provisioning Engine

The Auto-Provisioning engine leverages Next.js App Router's dynamic routing combined with localized localStorage and IndexedDB abstractions to instantly mock database persistence for isolated Hubs.

TECHNICAL IMPLEMENTATION: DYNAMIC ROUTE INJECTION

HOW DO WE ISOLATE TENANTS IN A SINGLE RUNTIME? We use Next.js dynamic catch-all segments (e.g., app/gts/hub/[id]/page.tsx) to act as a runtime dependency injector. The [id] param references a Preset blueprint, which dictates exactly which React components, APIs, and AI personas to mount for that specific user.

When a representative triggers a showcase generation, the orchestration pipeline executes in three asynchronous phases:

1. 1.Preset Resolution (The Blueprint): The system queries the gts_custom_presets datastore to retrieve the configuration JSON. This blueprint maps the exact contentIds (Webpages, Document Vaults) and personaId (the specific AI system prompt) required for the deployment.
2. 2.Dynamic Generation (The Instantiation): A unique, isolated Hub entity is instantiated. Rather than deploying a new server, the Next.js runtime dynamically injects the Preset configuration into a pre-compiled [id]/page.tsx template. This achieves multi-tenant isolation within a single running instance.
3. 3.Telemetry Anchoring: Crucially, the system attaches a unique telemetry beacon to the Hub. The vanity token (e.g., tok_enterprise_demo) is intercepted by lib/telemetry.ts upon initial page load. The standard anonymous visitor ID (anon_...) is forcibly overwritten with the exact token identity, ensuring every click, scroll, and read-depth metric is perfectly synced to the [PROSPECT LEDGER].

Architectural Elegance

The elegance of this system lies in its complete abstraction of backend complexity. Representatives generate complex, customized multi-agent environments with a single click, entirely bypassing the need for DevOps intervention. The resulting environment is mathematically isolated, cryptographically secure, and perfectly wired into our deep-telemetry pipelines.