> ## Documentation Index
> Fetch the complete documentation index at: https://docs.lostmindai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# LostMind AI - PropTech Variance Commentary Tool - Architecture

> architecture documentation for LostMind AI - PropTech Variance Commentary Tool

# Architecture Overview

## System Overview

The PropTech Variance Commentary Tool ingests property finance workbooks, normalises data into a relational model, and generates AI-authored variance narratives for month-end reporting.

## Core Modules

* **Data Ingestion & Sanitisation** — Converts Knight Frank-style XLSM exports into a privacy-safe dataset
* **SQL Modelling Layer** — Builds structured tables for ledgers, payables, and property metadata
* **Variance Engine** — Calculates period deltas and surfaces exceptions that need commentary
* **AI Commentary Service** — Prompts foundation models to draft auditor-friendly explanations
* **Publishing Pipeline** — Outputs summaries back into Excel, PDF, or downstream dashboards

## Integrations

* **Unified AI Client** connects to approved models (Gemma, Gemini, Claude) through the LostMind AI ai-compute service
* **TurboRepo Services** supply shared authentication, logging, and observability utilities
* **Security Scans** enforce data-governance policies before reports leave the environment

## Security Considerations

* Mandatory sanitisation step strips vendor names and sensitive PII before processing
* Configurable AI routing keeps regulated data on compliant infrastructure
* Audit logs capture every generated commentary block for finance review

*Summarised from PropTech variance automation design notes.*