The author proposes integrating the Luna Decisions API into an n8n-based real estate monitoring workflow to improve decision-making processes. Currently, the system uses a standard code-based approach to detect price changes and a text-based LLM to generate summaries for Telegram notifications.
The existing infrastructure relies on Google Sheets as both a data store and a dashboard. The current logic for tracking price fluctuations is limited by issues such as currency conversion ambiguity, lack of historical price data, and incorrect categorization of price drops due to imprecise threshold settings.
The Luna Decisions model, identified as openai/gpt-6-luna-decisions on OpenRouter, offers a structured approach with typed scores. The model features a context window of 1.1 million tokens, with a reported pricing of $0.10 per million input tokens and $0 for output tokens as of October 6, 2026.
The author suggests a phased testing approach rather than immediate full-scale implementation. The initial goal is to classify price changes into categories like 'normal change', 'needs verification', or 'insufficient data' to assist human review, rather than attempting to provide automated investment advice.
Future improvements to the workflow include implementing a 'not found in latest scan' status to distinguish between sold properties and listings that were merely removed or hidden. The author emphasizes the need to validate API endpoints, authorization, and data serialization before production deployment.
