Chapter 1: Getting Started

The Blueprint Engine (The 70/30 Rule)

Target Audience: New users trying to understand the mental model of the platform.

When you first start building a document workflow in Strucsta, you don't start with a blank canvas. Instead, you describe what you want to build in the Command Center, and our system generates a starting point. We call this the Blueprint Engine.

However, it is crucial to understand the philosophy behind how this engine works. We operate on what we call the 70/30 Rule.

The 70/30 Rule Explained

Generative AI is incredible at creating scaffolding and understanding intent, but it struggles with absolute precision-especially when it comes to strict logic and pixel-perfect design. If an AI generates 100% of a legal contract's layout and logic autonomously, you cannot guarantee the margins will be correct or that the compliance clauses will perfectly align every single time.

Strucsta is designed for deterministic, production-ready document generation. Therefore, the AI builds the first 70%, and you perfect the remaining 30%.

The 70%: What the AI Builds

When you generate a Blueprint, the AI analyzes your goal and automatically constructs the Directed Acyclic Graph (DAG) for your workflow. It handles the heavy lifting of the initial setup:

Ingest Scaffolding: It determines the likely data fields you will need. For example, if you ask for an "Invoice Generator," it will automatically create schema fields for clientName, lineItems, taxRate, and totalAmount.
Reasoning Prompts: It drafts the initial system instructions for the LLM step and generates the strict zodResponseFormat JSON schema required to extract data reliably.
Layout Boilerplate: It generates a foundational JSONC (JSON with comments) template for the pdfmaker Render step, mapping the AI's output variables to a basic visual layout.

The 30%: What You Refine

The Blueprint gets you up and running in seconds, but human review is strictly required to turn that draft into a bulletproof pipeline. You will need to dive into the generated steps to refine the details:

Refining Inputs: You may need to add hidden fields to your forms, adjust webhook payloads, or change data types to match your exact internal systems.
Tuning the Prompt: The AI's initial prompt will be generic. You must inject your specific business rules, tone guidelines, and edge-case instructions (e.g., "If the client resides in California, ensure the CA-specific liability clause is selected").
Achieving Pixel-Perfection: The AI will give you a structurally sound PDF, but you will need to adjust the JSONC template to match your brand. This means defining your exact corporate fonts, colors, margins, and conditional table rendering.

Why This Approach?

By strictly enforcing human-in-the-loop refinement, Strucsta bridges the gap between the flexibility of AI and the rigid requirements of professional documents. You get the speed of AI generation without sacrificing the strict structure, total control, and reliability required for business operations.