Agent Trainer Package Artifacts
Relevant source files
The following files were used as context for generating this wiki page:
Agent Trainer Package Artifacts consist of hardened, white-label AI assistant scaffolds designed for domain-specific intelligence grounding. These artifacts facilitate the transition from raw proprietary corpora to commercially usable deployment layers. The package focuses on preserving a user's specific context, vocabulary, and operating environment within an inspectable framework.
Sources: showcase/agent-trainer-pricing.html:237-239, showcase/registry.json:96-102, showcase.html:138-140
Package Architecture and Components
The Agent Trainer Package, specifically the "Productized Floor" version, includes a self-contained ZIP file containing the infrastructure necessary for a bespoke AI deployment. It integrates a specific corpus model with technical scaffolds for data handling and user interface.
Sources: showcase/agent-trainer-pricing.html:436-440, showcase/agent-trainer-sprint.html:401-404
Core Components
The system distributes functionality across several technical layers:
- Corpus Layer: Implements a Four-Lane Corpus Model and retrieval scripts to ground the AI in proprietary data.
- Infrastructure: Uses a Supabase schema with Row-Level Security (RLS) and retrieval RPCs.
- Interface Scaffolds: Provides React component scaffolds and a browser-based Setup Wizard.
- Execution Environments: Supports cross-platform entrypoints including Shell, Windows, Docker, and Browser.
Sources: showcase/agent-trainer-pricing.html:247-259, showcase/agent-trainer-pricing.html:436-440, showcase/ui-component-sheet.html:56-60
System Workflow
The following diagram illustrates the lifecycle of an Agent Trainer artifact from onboarding to delivery.
The diagram represents the "Revenue Pipeline Flow" and the subsequent deployment steps required for a functional artifact.
Sources: showcase/agent-trainer-sprint.html:79-119, showcase/agent-trainer-pricing.html:455-458
Revenue Pipeline and Transaction Logic
The package utilizes a specific transactional flow to gate access to deliverable assets. Technical resolution of data handoff between these steps is critical for system stability.
Transactional Flow Steps
| Step | Action | Dependency |
|---|---|---|
| Package Gen | Collects onboarding fields (Email, Company) | User Input |
| Payment | Processes transaction via Stripe | Stripe API |
| Key Issued | Automatically generates unique access key | Webhook |
| Key Entry | Client enters key in portal | Issued Key |
| Download | Delivers the deliverable asset (ZIP) | Validated Key |
Sources: showcase/agent-trainer-sprint.html:81-118, showcase/agent-trainer-pricing.html:362-368
Key Data Structures and Configuration
Artifacts are customized using specific configuration tokens and environment variables.
- White-Label Tokens:
KITNAME,KITDOMAIN, andKITPRIMARYCOLORcontrol branding. - Session Management: Data must persist from onboarding forms through to the generation handler.
- RBAC: Role-Based Access Control separates Admin Dashboard environments from Client-facing trainers.
Sources: showcase/agent-trainer-sprint.html:335-342, showcase/agent-trainer-pricing.html:451-453
Synthesis and Rendering Engine
Artifacts are produced by a Generative Render Engine that synthesizes source material according to specific styles and format directives.
Synthesis Styles
The engine applies different lenses to the source material during the generation of the package artifacts:
- Faithful: Accurate and complete reproduction.
- Technical: Precise and evidence-based synthesis.
- Founder Voice: Bold and direct communication style.
- PLK Resonant: Infusion of Personal Language Kit (PLK) vocabulary (e.g., luminosity, kinetic energy).
Sources: showcase/render-engine.html:68-79, showcase/render-engine.html:150-160
Artifact Generation Process
This diagram describes how the Render Engine interacts with the Runtime to create artifacts.
Sources: showcase/render-engine.html:58-66, showcase/render-engine.html:309-325
Development and Bug Tracking
Current development focuses on resolving state handoff failures and visual instability within the package builder.
Active Sprint Issues
- DEF-02: Persistent Email Prompt. The system fails to pass buyer email data to the generation step due to missing React Context Provider boundaries or lazy Firestore listeners.
- DEF-01: Selection Highlight Instability. Positional "jumping" occurs due to redundant
setStatecalls or race conditions between local UI and async Firestore state. - AT-02: Admin/Client Separation. Requires implementation of role-based access control to isolate the Admin Dashboard Trainer.
Sources: showcase/agent-trainer-sprint.html:128-142, showcase/agent-trainer-sprint.html:184-198, showcase/agent-trainer-sprint.html:324-332
Summary
Agent Trainer Package Artifacts provide a structured methodology for deploying bespoke AI systems grounded in proprietary knowledge. The system combines a robust "Productized Floor" code scaffold with a specific revenue pipeline and a multi-style synthesis engine. Development prioritizes data persistence and clear environment separation to ensure the "Revenue gate" requirements are met before production billing is enabled.
Sources: showcase/agent-trainer-sprint.html:56-61, showcase/agent-trainer-pricing.html:241-245