v0.0.1 — initial public release
First public release of the Parsimony stack: one open-source Python layer that lets you (and your
AI agent) find, fetch, and trust macro and financial data across more than 20 providers.
What's here
- Find it. A searchable catalog over thousands of series, codes, and descriptions, so you locate
the right data instead of reading a provider's PDF docs. - Fetch it, typed. One idiom across every provider. Each fetch returns a typed
Resultcarrying
its provenance (source, params, fetch time), with failures surfaced as a typed error taxonomy. - Completely. Fetches are guarded against the silent-truncation traps these APIs hide (an HTTP
200 with an incomplete series, hard row caps), so you are not working on a fraction of the data
without knowing it. - From your coding agent. Drop in one skill file (
npx skills add ockham-sh/parsimony) and an
off-the-shelf agent can drive the library. No MCP server.
Install
pip install parsimony-core # the import is `parsimony`Honest scope. All connectors are first-party today; community connectors are an invitation, not
yet a fact (see the pinned "good first connector" issue). The core Result carries one-hop fetch
provenance; full transformation lineage lives in parsimony-agents. Apache-2.0.
Note:
0.0.1is a deliberate "early and evolving" starting point. The API may still move before
1.0; pin a version if you need stability.