STATUS: INTERNAL R&D RESTRICTED_ACCESS

3Dims AICatalog

AI-Orchestrated Data Structuring Engine · Vision-Based Extraction · Dynamic Schema

React 19 (Vite)Python (FastAPI, Asyncio)SQLitePydantic (Dynamic Model Generation)OpenAI Vision ModelsOpenAI Assistants API
[SYSTEM_OVERVIEW]

Internal Tool for Technical Product Catalog Reconstruction. An AI-orchestrated data structuring engine originally built for a 3D construction configurator. The system solves the 'Manufacturer Catalog' problem: transforming chaotic, inconsistent PDF layouts full of mixed tables and scattered specs into clean, structured business objects.

Unlike traditional OCR or fixed-schema parsers, it employs a 'Meta-Schema' architecture where the data structure is defined dynamically (manually or by AI) without code changes. The system then orchestrates a multi-stage pipeline—grouping pages by product, denoising images, and using Vision interaction to extract strict, type-safe data.

Core Loops
Schema Definition (AI-Assisted)Meta-Prompt Generation (Context Injection)Logical Page Grouping & ChunkingImage Denoising (Label Injection)Vision Extraction (Strict JSON)Human-in-the-Loop Mass Editing

/// DEPLOYMENT_SCOPE

  • ::Dynamic Schema Engine
  • ::Meta-Prompting Orchestrator
  • ::Vision-Based Structure Extraction
  • ::Hybrid Relational-Document Storage
  • ::Mass Data Validation Grid

Engineering Architecture

Meta-Schema Architecture

Product structure is not hardcoded in the application. Product parsing rules are stored as JSON ('product_schema'). Pydantic models are generated at runtime to validate extraction, allowing the system to switch from processing 'Ovens' to 'Cladding Panels' instantly without a deploy.

Meta-Prompting System

One AI agent generates the system prompt for another. The 'Architect' agent analyzes the active schema and compiles a strict, schema-compliant instruction set for the 'Worker' agent. This ensures the extraction model operates within clearly defined structural constraints.

Image Denoising & Label Injection

Images in catalogs introduce semantic noise ('...looks modern...'). The system extracts all images and replaces them with tokens ('<image_1>') in the text stream. This keeps the model focused on technical specifications while preserving image references.

Context Chunking Strategy [6-Pages]

Catalogs are processed in empirically determined 6-page chunks (sliding window). This balance prevents context overload in the LLM while maintaining enough 'product boundary' context to capture items spanning multiple pages.

AI Engine

Vision-Grounded Extraction

The model receives high-res page images, not just text. This allows it to correctly parse complex, broken, or multi-column tables where text-only extraction loses the row/column relationship.

Logical Page Grouping

Before extraction, a lightweight analysis pass groups pages into 'Product Units'. This ensures that a product spanning pages 4–5 is treated as a single extraction context, rather than splitting data across arbitrary page logic.

Strict JSON Mode Strategy

The system enforces a 'Strict Schema or Reject' policy. If the vision model returns a structure that doesn't validate against the dynamic Pydantic schema, the extraction result is rejected. No 'best guess' data enters the database.

Reliability Principles

  • Hallucination MinimizationExplicit no-guess rules in prompts and image denoising reduce speculative completions.
  • Schema SafetyDynamic schemas are validated against a strict meta-model to ensure structural integrity.
  • Token ControlFixed 6-page chunking stabilizes context size and prevents attention degradation.

Invariants (Strict Rules)

  • ::Schema-on-Write: No migrations for old data
  • ::Strict Validation: Malformed JSON is rejected; no speculative data is stored
  • ::Product-Based Context: Logic transcends page boundaries
  • ::Image Noise Removal: Illustrations replaced by token labels
  • ::Strict Typing: Field values must match Pydantic dynamic models

Data Entities

product_schema (dynamic JSON definition)products (hybrid: indexed fields + JSON characteristics)parsing_jobs (async pipeline state)assistant_threads (context isolation)

UX Philosophies

  • Excel-like Mass Editor for rapid validation
  • Split-Screen: PDF Viewer vs Extracted Data
  • Schema-Aware Dynamic Rendering
  • Dark Mode / Glassmorphism (Engineering Aesthetic)

Future Signals

  • Distributed Queueing for large catalogs
  • Export API for external retailers
  • Advanced Table Recognition (specialized model)
  • Self-Correction Loops (Agent Critic)