Skip to content

Repository files navigation

archive-tools

Local-first analysis tools for personal social media exports (Facebook, Twitter/X). Everything runs on your own machine against your own export files — nothing leaves your disk.

What's here

Script Purpose
dmscan.py Scan Twitter DM export data
tweetscan.py Scan and filter tweet archives
patterns.py Temporal/behavioral pattern extraction across an archive
pubcontent.py Public-content analysis (posts, engagement)
altan.py Alt/secondary account archive analysis
twan.py Twitter archive analysis
gendersex.py Topic-specific content analysis
motif_significance.py Register-transition motif detection with a marginal-preserving null (after Uri Alon's network-motif method). Distinguishes real behavioral patterns from base-rate artifacts via Z-scores against a shuffled ensemble.
memory_palace.py Method-of-loci memory-palace generator (classical Ad Herennium rules via Frances Yates). Turns an ordered fact list into loci + imagines agentes — a mnemonic front-end for the timeline/lifelog.

Method notes

motif_significance.py implements the core discipline from Alon's Introduction to Systems Biology (ch. 3-4): a pattern is only real if it beats a null model that preserves the trivial marginals. Alon's lesson — the naive random null over-credits structure; you must preserve each node's degree (here, each register's total frequency) before a Z-score means anything. It reports three motif classes: autoregulation (does a register persist?), transitions (which register→register jumps beat base rates?), and feed-forward triples.

memory_palace.py implements the classical art of memory as Yates reconstructs it: ordered loci at moderate spacing/lighting, and imagines agentes — active, striking images (beauty or ugliness, ornamented or disfigured, rare not ordinary). The loci preserve order; the images preserve facts. It's a database front-end that predates databases by two millennia.

These began as working scripts against real exports (77k+ messages, 68k+ tweets across two accounts, a 17-year Facebook archive). Generalization into a proper CLI is in progress.

Roadmap

  • Normalize the common export-parsing code into a shared module
  • One CLI entry point (archive-tools scan <export-dir>)
  • DuckDB-backed unified corpus ("lifelog") across platforms
  • Whisper transcription pass for voice-message folders
  • Charts/report output (the current scripts print to stdout)

Why local-first

Personal archives are the most sensitive data most people own. These tools assume the export never leaves your machine: no API keys, no uploads, no telemetry.

About

Local-first analysis tools for personal social media exports - network-motif statistics, register timelines, and a classical memory-palace generator

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages