v0.1.0

v0.1.0 — 2026-07-13

First public release.

crible is a self-hosted fundamental stock screener — zero API keys, zero
subscription, forever. That zero-key contract is CI-enforced: the whole test
suite runs with an empty environment, and every core flow works without any
account or key.

• Universe: ~161k equities / 117 countries from FinanceDatabase, searchable
  by symbol or name.
• Data (all keyless): Yahoo via yfinance (rolling, rate-budgeted crawl,
  Europe-first), audited EU figures from filings.xbrl.org (ESEF xBRL-JSON,
  matched through the GLEIF ISIN→LEI file) — audited values outrank scraped
  ones at reconciliation.
• Scores: Piotroski F, Altman Z, Beneish M and 150+ ratios, every number
  traceable to its inputs; composite quality/value/momentum percentile ranks
  with peer groups, never imputed.
• Screening: one filter DSL compiled to parametrized DuckDB SQL, shared by
  the CLI, the HTTP API and the React SPA — full-universe screens in
  milliseconds.
• In-browser demo: the GitHub Pages demo runs the real screener entirely
  client-side — the DSL compiled in TypeScript (golden-locked to the Python
  compiler) executing on DuckDB-WASM over Parquet published by a nightly
  open-data refresh with a last-good guarantee.
• Deploy: two-container Docker Compose (ingest + api), multi-arch
  images (amd64/arm64) on GHCR. Designed for private-LAN self-hosting.

Rolling open dataset (nightly)

Nightly keyless open-data refresh. Bootstrap a self-hosted crible from it with crible bootstrap — no crawl needed.