Overview

Initial release of 3D Model Assist AI, a Python-based workflow script for WhereScape 3D that uses large language models (LLMs) to automatically classify source system columns into Data Vault 2.0 attribute types. The script integrates with the WhereScape 3D CLI to export model data, invoke an LLM for classification, and re-import the enriched model with attribute types and documentation annotations applied.

New Features

• Multi-provider LLM support via LangChain, including OpenAI, Anthropic, Azure OpenAI, Google Generative AI, Groq, xAI, Mistral, Ollama, and AWS Bedrock.
• Data Vault 2.0 classification engine with a structured rule hierarchy covering business keys, foreign key relationships, multi-active data, satellite volatility tiers, audit columns, and reference data.
• User modelling rules — a Tkinter UI lets users enter, load, and save custom prompts that act as hard overrides on the LLM's classification logic.
• Confidence scoring — each column classification is assigned a high, medium, or low confidence level, with low-confidence results flagged in the log for manual review.
• Documentation injection — classification rationale and confidence levels are written directly into the models documentation subjects, visible within WhereScape 3D.
• Retry logic for transient network errors using exponential backoff (up to 3 attempts).
• Token usage logging across all supported providers.

Requirements