Weft is a federation of sibling developer-tools projects, each authoritative for one domain, each useful standalone, each composable pairwise, and enrich-only (never load-bearing) when composed.
This folder (~/weft) is the authoritative source of truth for federation-wide interoperability — the roster, the doctrine, the shared identity model, the integration matrix, the asterisk register, the glossary, and the cross-product contract index. It is documentation only: no code, no runtime, no store. There is nothing called "Weft" to install or run.
Canon is split by domain (resolved 2026-06-06 — conflict-register.md §B-6): doctrine.md owns the roster/membership/composition-law, SHIPPING.md owns release posture/gate, contracts-index.md owns the contract index. This README, federation-map.md, and compatibility.md are derived views that point to the domain owner, not competing canon.
As of 2026-06-05 this hub was rebuilt to be authoritative. It previously drifted badly (wrong roster, stale versions, wrong identity model). The fix is structural — see The authority model below — and every conflict found is recorded in conflict-register.md.
| Member | Repo | Lang | Domain authority | Briefing |
|---|---|---|---|---|
| Loomweave | ~/loomweave |
Rust | Code structure + identity authority (SEI) | members/loomweave.md |
| Filigree | ~/filigree |
Python | Work state / issue lifecycle | members/filigree.md |
| Wardline | ~/wardline |
Python | Trust-boundary analysis | members/wardline.md |
| Legis | ~/legis |
Python | Git/CI governance & attestations | members/legis.md |
| Charter | ~/charter |
Python | Requirements, traceability, verification | members/charter.md |
Not members, but part of the story:
- Shuttle — a roadmap thought-bubble for a future change-execution authority. No repo; lowest priority; displaceable by any better idea.
- Lacuna — the deliberately-flawed demonstration specimen the suite is run against.
- roadmap-ideas.md — a bench of speculative sub-app candidates (incl. Shuttle). Ideas only; none are members until they pass the go/no-go test.
Excluded entirely: wardline-watcher, wardline.old, weft.old (the defunct heavier builds).
The hub drifted because it restated project-internal facts that move (versions, tool counts, rule lists, routes). The rule that prevents recurrence:
- weft owns the interop layer and is authoritative for it: the roster, the doctrine, the SEI identity model, the asterisk register, the URI/addressing status, the integration matrix, the glossary, and the contract index.
- weft points to each project for that project's own surface facts. Any version/status shown here is marked "snapshot — not authoritative; see the repo."
- each project points back to weft for federation patterns, keeping authority only over its own domain surface.
Audit invariant: weft contains no restated project-internal fact without a pointer; every federation-pattern doc in every member repo resolves to weft.
- New here? doctrine.md (the federation law) → federation-map.md (how they compose) → the member briefing you care about.
- Building a cross-tool integration? contracts-index.md (points to the authoritative schema in the owning repo) + sei-standard.md (identity is the spine).
- Building your own tool to drop into the federation? federation-sdk.md — the member-builder's interface contract: the conformance spine, every interface surface, and the oracle gate.
- Pre-release conflict flush? conflict-register.md — walk Class B before any point release.
- Shared vocabulary? glossary.md. Documented axiom violations? asterisk-register.md.
- How it ships / launch & go-to-market? SHIPPING.md (PM-decided 2026-06-05) + compatibility.md (integration-liveness now; version matrix at 1.0).
- How this doc set is organized & governed? MANIFEST.md.
Each member is authoritative for one domain, solo-useful, meaningfully composable pairwise, and enrich-only — never load-bearing — when composed. (doctrine.md §5.)