README.md

---
title: README
emoji: 📉
colorFrom: indigo
colorTo: gray
sdk: static
pinned: false
---

<h1 align="center">PARSIMONY</h1>
<p align="center"><em>Pre-built searchable indexes of financial and economic data series.</em></p>

---


Each dataset here is a drop-in catalog — a parquet of series metadata plus a FAISS embedding index — consumed by the matching [parsimony](https://github.com/ockham-sh/parsimony) connector to resolve natural-language queries like *"swiss policy rate"* or *"US 10-year yield"* to a provider-native series ID, without hitting the upstream provider.

These datasets are not meant to be browsed directly. Install the relevant connector from [parsimony-connectors](https://github.com/ockham-sh/parsimony-connectors) — its `<provider>_search` function pulls the catalog on first use and caches it locally.

## Anatomy of a catalog

- `entries.parquet` — one row per series with provider ID, title, units, frequency.
- `embeddings.faiss` — dense vector index over titles and descriptions for semantic lookup.
- `meta.json` — provider, snapshot timestamp, embedding model, row counts.

Catalogs are pulled on-demand by the matching connector and cached locally. No API keys required to read the indexes — keys are only needed when you fetch the actual series from the upstream provider.

## Links

- **Kernel** — [github.com/ockham-sh/parsimony](https://github.com/ockham-sh/parsimony)
- **Connectors** — [github.com/ockham-sh/parsimony-connectors](https://github.com/ockham-sh/parsimony-connectors)
- **Docs** — [docs.parsimony.dev](https://docs.parsimony.dev)
- **PyPI** — [`parsimony-core`](https://pypi.org/project/parsimony-core/)

Maintained by [Ockham](https://github.com/ockham-sh) · Apache 2.0