Skip to content

CLAUDE.md

Latest commit

 

History

History
30 lines (25 loc) · 2.71 KB

CLAUDE.md

File metadata and controls

30 lines (25 loc) · 2.71 KB
Raw

Disguise Release Notes Aggregator

Project Overview

Static site that aggregates disguise designer release notes (r14–r32) into a single searchable table. Uses AI (via GitHub Models) to extract and categorise changes from HTML release pages. AI-extracted content may be inaccurate — always refer to the official release notes.

Architecture

  • scrape.mjs — Node CLI script. Fetches release pages from help.disguise.one, converts HTML to markdown via turndown, sends to AI for structured extraction, outputs data/releases.json.
  • index.html — Single static HTML file with inline CSS/JS. Loads data/releases.json via fetch. Provides search, sort, and filter UI.
  • data/releases.json — Generated tree-structured file committed to repo. Schema: [{ version, build, starter_build, released, url, changes: [{ dsof, category, description }] }].
  • data/.ai-cache.json — AI response cache keyed by content hash. Committed to repo.
  • validate.mjs — Node CLI script, no dependencies. Validates data/releases.json before the daily PR is allowed to auto-merge. Structural checks apply to every entry; stricter quality checks apply only to entries that are new or changed versus the base ref, because pre-r30 entries contain known junk (empty build, empty released, comma-joined dsof) that is grandfathered in.

Commands

  • node scrape.mjs — Scrape all release pages and extract with AI (requires AI_TOKEN env var)
  • node scrape.mjs --version r32 — Scrape a specific major version
  • node scrape.mjs --version r32.2 --force — Re-extract a version, ignoring cache
  • node scrape.mjs --cache-only — Rebuild releases.json from AI cache (no fetching or API calls)
  • node scrape.mjs --fix-urls — Patch cached entries with correct anchor URLs
  • npx serve . — Serve locally for testing (fetch requires HTTP, not file://)
  • node validate.mjs — Validate data/releases.json against origin/master
  • node validate.mjs --no-base — Structure/quality checks only, no comparison (note: fails on grandfathered legacy entries)

CI

  • .github/workflows/update-releases.yml — Daily scrape. Always opens/updates the rolling auto/update-releases PR, then auto-merges it only if validation passes, the diff is purely additive, and nothing outside data/ changed. A failed validation leaves the PR open and fails the run.
  • .github/workflows/validate.yml — Runs validate.mjs on human PRs. Its job is named data-validation to match the commit status the daily workflow posts, so one required check covers both paths.

Conventions

  • Use conventional commits (e.g., feat:, fix:, chore:)
  • No framework for the frontend — plain HTML/CSS/JS
  • turndown is the only npm dependency (HTML to markdown conversion)