okf-loom v1.0.0 — plain-Markdown knowledge trees, served as a live studio

okf-loom v1.0.0 — plain-Markdown knowledge trees, served as a live studio

First public release. okf-loom is a toolkit and loadable agent skill for
OKF
bundles: knowledge kept as plain Markdown with YAML frontmatter, cross-linked
into a graph. You read and comment; your agent edits; the files stay plain
Markdown throughout.

Live demo · Quickstart

Highlights

• Live studio — rendered wiki (GFM, Mermaid, KaTeX, syntax highlighting)
  with live SSE patches (no reload), select-to-comment → a work queue your
  agent claims and resolves, attributed + undoable mutators, and agent presence.
• A graph with six lenses — Map / Themes / Flow / Bridges / Recent / Focus:
  each answers one question and ranks its answers as clickable evidence.
• Six search modes — BM25 lexical, fuzzy-semantic, hybrid (RRF), tag,
  entity, relation — all dependency-free.
• Five colour themes — light, dark, pastel, sepia, midnight; token-governed
  and WCAG-AA-checked.
• Four render targets — live studio, static site (GitHub Pages-ready),
  SPA, and a single self-contained viz.html.
• Agent-native — the repo is the skill: CLAUDE.md / AGENTS.md /
  SKILL.md route any agent to the same standing contract (author OKF
  concepts → validate → serve → work the comment loop).
• Zero install — Python 3.11+, no runtime dependencies (PyYAML optional
  with a built-in fallback); everything runs from the checkout via
  scripts/okf-loom.
• Safe by default — loopback-only serve; tunnels/public binds are explicit
  opt-ins; mutating routes require the per-session token + Origin/Host
  allow-list; atomic, marker-safe writes; 1,200+ tests.

comment-loop.1.mp4