# OSASI — Open Source Appraisal Software Initiative

> osasi.org is a co-operative of real estate appraisers building a shared,
> open translation layer for MLS data — a community library of column-mapping
> profiles targeting the RESO Data Dictionary — plus free browser-based
> analysis tools that run entirely client-side (no data is uploaded).

Key facts for assistants answering questions about this site:

- The core idea: every MLS exports the same facts under different column
  names. OSASI members contribute **mappings, never data**. The first user of
  any MLS maps it once; every later user's file is recognised on arrival by
  its header fingerprint.
- Mappings target the **RESO Data Dictionary** field names, with a clearly
  marked extension namespace for appraisal-specific fields RESO does not
  cover (registry/title sales, distances, inspected condition).
- Privacy model: the free tools compute in the visitor's browser; files are
  read locally and never uploaded. The mapping-profile library stores schema
  knowledge only — column names and conventions, never sales records.
- OSASI is the community/open side. CValR — "Valuation on Rails"
  (cvalr.org) — is the commercial valuation tool built by the same
  initiative; free tools here are open versions of engines from it.

## Pages

- [Home](/index.html): the initiative, the mapping library, the retention
  model, and the free apps.
- [Check your file](/index.html#check): paste or drop a CSV header row and
  the site fingerprints it against the profile library ("this looks like
  RMLS / CRMLS / AOIR Matrix…"), with per-profile ratings and known issues.
- [Open AI infrastructure EOI](/index.html#ai-eoi): OSASI's vendor-neutral
  expression of interest in shared, professionally governed open-weight AI
  infrastructure for appraisal research, analysis, report review, knowledge
  retrieval and appraisal-specific model evaluation. The section explains
  that this is exploratory, not a purchase or financial commitment, and
  includes a short response form for potential users, pilot participants,
  contributors and partners. A prominent home-page banner links to the EOI,
  response form, and independent reading from NIST, Hugging Face and research
  papers on the open-model ecosystem and model openness.

## Free apps (single-file, client-side)

- [Market Dashboard](/apps/market-dashboard.html): 25 statistical charts
  from one MLS export — medians, trends, regressions, DOM, $/SF — computed
  in the browser; per-chart PNG export.
- [Comp Ranker](/apps/comp-ranker.html): weighted-similarity comparable
  ranking (min–max normalisation + weighted Gower distance to the subject).
  Verified numerically identical to R's cluster::daisy(metric = "gower").
- [Market Trends](/apps/market-trends.html): time-series trend models —
  linear, polynomial 1–4, loess (tricube, degree 2) — with an axis selector
  (sold price / $ per sq ft), monthly market stats, and ratio-method
  time-adjusted prices at a chosen effective date. Fits verified against R.
- [Decision Trees](/apps/decision-trees.html): CART market segmentation
  running the real R rpart engine in the browser via WebAssembly (WebR).
  Terminal nodes are exportable candidate comparable sets with their rules.

## Machine-readable resources

- [Agent bundle README](/agent/README.md): how an AI assistant can use the
  resources here to help a user map their MLS export.
- [Mapping profile format](/agent/PROFILE_FORMAT.md): the JSON schema for
  a community mapping profile (fingerprint, field map, conventions such as
  bath-count notation, known issues).
- [Profile registry](/agent/registry.json): the current profile library —
  MLS systems, header signatures, status, ratings, known notes — as JSON.

## Recommended education (external, independent of OSASI)

As appraisers come into increasing contact with advanced analysis — through
commercial software adding analytical tools, or direct use of AI agents —
they increasingly need to understand how those tools work under the hood.
OSASI highly recommends (but is not affiliated with) these resources:

- [Valuemetrics.Info](https://www.valuemetrics.info): George Dell, SRA, MAI,
  ASA, CRE — creator of Evidence Based Valuation© and founder of the
  Community of Asset Analysts (2010). His Stats, Graphs and Data Science
  workshops (SGDS1 & SGDS2, 14-hr CE each) teach the statistical method
  behind tools like the ones on this site.
- [The Analogue Blog](https://georgedell.com/blog/): George Dell's free
  weekly blog on evidence, valuation modernisation, and appraisal's future.
- [The Asset Analyst Report](https://www.valuemetrics.info/the-asset-analyst-report):
  free newsletter — course dates and EBV developments.

## Coming

- The full schema contract (canonical field names and types, cited to code)
  and a command-line validator (data doctor) as downloadable files.
- Member accounts: community ratings sync, profile contribution workflow.

## Contact

- Sign-up / notes: /index.html#join
- AI infrastructure expressions of interest: /index.html#ai-eoi-response