cubacityhistory-pp-cli

A read-only Go CLI over a local SQLite mirror of the Cuba City community-history corpus that volunteer admin "Historical Jottings" has built since 2010:

• 516 blog posts from cubacityhistory.blogspot.com
• 63 Facebook posts from facebook.com/cubacityhistory
• 2,469 images and 345 reader comments preserved offline

It does not hit Facebook or Blogger directly — refreshing the underlying mirror is a separate workflow that lives in the companion research-wiki repo (thetimechain/random → wiki/topics/cuba-city-wi-history/raw/). This CLI is the query layer over what's already on disk.

Why this exists

Local-history corpora maintained by single volunteer admins are some of the most fragile bodies of work on the internet: one moved-house, one platform shutdown, one ill-timed disk failure, and 15 years of community memory evaporates. This CLI is one piece of a two-piece preservation pact:

1. The archive (in the wiki repo) — markdown + images + raw HTML snapshots on a local drive, mirrored to GitHub, with Wayback redundancy.
2. The CLI (this repo) — a stable, agent-readable query surface that doesn't depend on the source platforms being alive.

Install

go install github.com/thetimechain/cubacityhistory-pp-cli/cmd/cubacityhistory-pp-cli@latest

Or build from source:

git clone https://github.com/thetimechain/cubacityhistory-pp-cli
cd cubacityhistory-pp-cli
go install ./cmd/cubacityhistory-pp-cli

Quick Start

# One-time: parse the local archive into the SQLite store
cubacityhistory-pp-cli sync --archive ~/github/random/wiki/topics/cuba-city-wi-history/raw

# Verify health
cubacityhistory-pp-cli doctor

# Search across post bodies + titles
cubacityhistory-pp-cli posts search "tornado" --limit 5

# Search across comments
cubacityhistory-pp-cli comments search "school" --limit 3

# Top commenters — the community's living memory holders
cubacityhistory-pp-cli authors --limit 10

# List posts from a specific year
cubacityhistory-pp-cli posts list --year 2024 --source facebook

# Pull one post with its full comment tree
cubacityhistory-pp-cli posts get pfbid031F6NZhBBuzk3NtWATY3UMLQLdMMN59YSC8RPQZTZAhQ4Rhgm9aqjhBWv3UgDj3dCl

Commands

Command	What it does
sync --archive <dir>	Parse the local archive into SQLite. Idempotent.
doctor	Verify the store, report counts.
posts list [--year Y] [--source S] [--with-images-only]	List posts, newest first.
posts get <id-or-permalink>	Fetch one post + comments.
posts search <query> [--snippet]	FTS5 search across bodies + titles.
comments list [--post ID] [--author NAME]	List comments.
comments search <query>	FTS5 across comment bodies.
authors	Top commenters by count.

Global flags

• --json — machine-readable output
• --csv — CSV output (where applicable)
• --compact — JSON output, high-gravity fields only
• --select date,title,id — pick specific fields (JSON only)
• --limit N — cap results (0 = unlimited; default 50)
• --db <path> — override the SQLite store location

Environment:

• CUBACITY_DB — alternate SQLite store path (default: $HOME/.cubacityhistory/store.db)

Architecture

Pure Go. Uses modernc.org/sqlite (no cgo, no compiler needed). SQLite's built-in FTS5 powers full-text search. Schema in internal/store/store.go.

The sync command parses markdown files with YAML frontmatter (the wiki's existing format) and populates four tables: posts, posts_fts (virtual FTS5), comments, comments_fts, images. Re-syncing is idempotent — truncates and re-inserts.

Companion archive

This CLI's data source is wiki/topics/cuba-city-wi-history/raw/ in thetimechain/random. The two together form the complete preservation copy.

Known quirks

• Blogspot comments use a ### Comment N — Author (date) header format rather than FB's ### Author — _time_, so the authors command shows Comment 1/Comment 2 as top names for the blogspot corpus. The real author lives in the comment body. Fixing this in a future revision.
• Image counts in doctor reflect what was visible on disk at sync time for FB only; blogspot images live under a different slug pattern and are not yet cross-referenced. The images themselves are preserved in the archive — this is a query-layer gap, not a preservation gap.

License

MIT.