MLS Listing Sync Analyzer

MLS Listing Sync Analyzer audits a portfolio of property listings and tells you which ones are underperforming because of how they are written rather than what they are selling. Each listing gets a completeness score from 0 to 100 and a list of the specific issues holding it back: too few photos, a description that is too short to rank or persuade, required fields left empty, and a price that deviates from comparable properties by more than your threshold. The output is a prioritised fix list, so an agent with two hundred listings knows which twelve to fix this afternoon. Every score, count and percentage is computed deterministically. A language model writes the readable summary and nothing else, with a template fallback when no model is available — so a completeness score means the same thing every time you run it. It is read-only by construction. It never creates, updates or deletes a listing, and there is no code path that could. Listing free text is treated as data rather than instruction, and seller and customer contact details are masked. Each step is callable alone. `score_completeness` returns the scores and missing fields with no narration. `detect_optimizations` returns the issues, the recommendations and the price deviations. `run_full` does everything and writes the summary. The hosted version works on listings you pass in, so you can audit an export with no setup at all. The repository also ships a read-only MLS and portal client if you want to self-host it against your own portal API. Built for real estate agents, brokerages and listing managers who suspect some of their inventory is invisible for fixable reasons.