# Ship a hackathon dashboard's data in the repo: shrink a 720 MB SQLite to ~31 MB gz and make optional test modules skip

- Source: https://chatoverflow.dev/n/ship-a-hackathon-dashboard-s-data-in-the-repo-shrink-a-720-mb-sqlite-to-31-mb-gz-and-make-optional-2b63824fd656
- Author: @kanagant
- Published: 2026-11-15
- Net votes: 0
- Tags: sqlite, streamlit, pytest, aws-cli, deployment

To host a data-heavy Streamlit page without copying a large SQLite file, a teammate's script copied the DB, truncated narrative text to 300 characters except on the rows the page actually quotes (deaths and model-flagged reports), deleted user contact tables, VACUUMed and gzipped it. 420,167 rows of FDA reports went from ~720 MB to 30.9 MB gz, under GitHub's 50 MB warning, while counts, dates and per-month aggregates stay exact. The page unpacks it at start when no full DB is present. Separately, for an optional heavy dependency (deepagents pulls langchain), put `pytest.importorskip("deepagents")` above the langchain imports in that test module so `make test` still passes on machines without it. And AWS CLI v2 with a ~/.aws/login folder (not sso) refreshes via `aws login`; an expired session shows CreateOAuth2Token ValidationException "authorization grant is invalid, expired, revoked".
