First public release

What is it?

sms-toolkit turns Android SMS/call history into a clean, browsable archive.
SMS Backup & Restore makes full backups on every scheduled run, so a year of
daily backups is dozens of multi-GB XML files that are ~99% duplicates.
sms-toolkit streams those files, drops the duplicates by MD5 hash, and keeps
one SQLite database — importing only what's genuinely new, every time.

Highlights

• Streaming de-dup import — sms and calls import XML backups in
  append mode by default; already-imported messages are skipped, so
  re-importing the same folder twice adds nothing.
• Chat-style viewer — serve starts a messaging-app-like web UI:
  conversation list, sent/received bubbles, day dividers, inline MMS photos
  with lightbox, plus an archive list view with search and pagination.
• Normalization & contacts — canonical phone/date backfill and a
  contacts table built from addresses (including multi-participant MMS),
  both previewable before writing (--report, --dry-run).
• IFTTT import, attachment export, year export — Excel-based IFTTT
  ingestion, MMS parts out to files with CSVs, and one-year carves into
  standalone databases.
• Screenshot OCR — extract conversations from message-app screenshots
  (four strategies) and fuzzy-match them against the database.
• Diagnostics — diag checks integrity, row counts, tag breakdown,
  date range, and MMS inventory with a 0/1 exit code.

Install

Requires Python 3.10+.

git clone
cd sms-toolkit
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e ".[excel,viewer,ocr]"

Extras are independent: excel (IFTTT import), viewer (Flask UI),
ocr (screenshot extraction). The base install covers import, normalize,
contacts, exports, and diagnostics.

Verification

184/184 tests pass; ruff, ruff format, and mypy-strict green. The suite
includes differential tests that run the refactored code against the
original pre-refactor scripts on identical inputs, and the refactored
package preserves the originals' MD5 dedup hashes byte-for-byte (they are
primary keys in existing databases).