v0.1.0

Weavatrix Refactor 0.1.0

The first published release of the refactoring layer. weavatrix-refactor is the
only Weavatrix component that writes code, and it sits in the middle of the stack:
weavatrix-online ⊃ weavatrix-refactor ⊃ weavatrix. Install it and you get the
full read-only core plus the tools that turn its proven analysis into applied
edits.

What it adds on top of the core

The read-only core (weavatrix 0.3.12) provides the analysis surface; this
layer's own read-only producers compute and prove edit plans against it, and its
apply tools apply them:

• Plan producers (read-only): rename_symbol (LSP, SQL, or graph-resolved
  by language), rename_related_symbols, change_signature, edit_symbol,
  bulk_replace, and organize_imports emit an applyable
  weavatrix.edit-plan.v1; move_file, move_symbol, and delete_readiness
  emit a review plan / dry-run / verdict you act on directly, not an apply
  envelope. Each edit carries its provenance tier, and only proven tiers are
  applyable.
• Appliers: apply_edit_plan and rollback_last_apply. Both are closed by
  default and open only with WEAVATRIX_ALLOW_SOURCE_EDITS=1, a single-use
  confirm token bound to the previewed plan, and a re-checked file hash. Every
  apply is reversible.

Guarantees carried up from the core

The core stays provably read-only; all writing lives here, behind the write gate.
A plan this layer computed and hash-verified against the core applies here only
if the working tree still matches the hashes it was proven against — otherwise it
fails closed with no partial writes, and rollback converges rather than wedging.

Provenance

Published from tag v0.1.0 with npm provenance (SLSA) and mirrored to the
official MCP Registry as io.github.sergii-ziborov/weavatrix-refactor.