Adapters

One adapter per source. A module with three names: NAME (the source it writes), sniff(path) -> bool ("is this file mine?", cheap, never raises) and run(input_path, since) -> iterator[dict] where each dict has the keys at, end, tz, source, kind, tier, payload (see SPEC §2) — everything except the chain fields, which the logbook adds on append. Give every line a stable payload.raw_id; logbook add skips lines whose (source, raw_id) are already in the log, so re-adding an export appends nothing.

Adapters built by the project ship inside the package under logbook/adapters/ (ADR 0012); the first is dawarich. Third-party adapters are separate packages registering the logbook.adapters entry point. logbook add <file-or-folder> tries every registered adapter's sniff and runs the first match.

Rules: pure by default (no network), never write to the source, never invent tier lower than the source deserves, fixture-tested with synthetic data only.

Planned first (ADR 0008): google-takeout, apple-health, whatsapp, then email.

Wanted — pick one and open an issue with the adapter template: immich, dawarich, strava, garmin, apple-fitness, eight-sleep, oura, withings, telegram, imessage, signal, gmail, ics-calendar, google-calendar, flighty, flightradar24, notion, obsidian, spotify, apple-music, letterboxd, goodreads, kindle-highlights, youtube-history, chrome-history, twitter-archive, instagram-export, bank-csv (generic, with per-bank mapping files), revolut, paypal, amazon-orders, uber, airbnb, vivino, steam, boat-passage logs.