Back to browse
GitHub Repository

Local content-addressed archive with observation history. Stores bytes by SHA-256, preserves locator state as contiguous spans, compresses with zstd and corpus-trained dictionaries. SQLite-backed.

7 starsPython

Farchive – SQLite-backed history-preserving compressed archive

by ekns·Apr 13, 2026·6 points·0 comments

AI Analysis

●●●BangerBig BrainNiche Gem

Span-based history tracking with corpus-trained zstd dictionaries beats standard dedup.

Strengths
  • Locator spans preserve A→B→A as three records, not one collapsed entry
  • Corpus-trained zstd dictionaries achieve 5-10x compression on similar documents
  • Single SQLite file with SQL queries — no server, no daemon, directly queryable
Weaknesses
  • Niche use case — most users won't need history-preserving archives
  • Python-only implementation limits adoption in other ecosystems
Target Audience

Data engineers, web scrapers, ML teams managing dataset versions

Similar To

git-annex · IPFS · DVC

Similar Projects