AI Updates

English | 繁體中文

AI Updates tracks the official changelogs for Claude Code, Codex, Antigravity, Usage, and GitHub CLI. It preserves immutable source records, supports human-reviewed plain-language updates in five languages (zh-TW, zh-CN, en, ja, and ko), and publishes feeds for a static website and compatible applications.

Website

https://aqua5230.github.io/ai-updates/

Sources

Tool	Official source
Claude Code	Anthropic Claude Code changelog
Codex	OpenAI Codex releases
Antigravity	Antigravity CLI changelog (falls back to local scripts/sync_agy.py if public source is unavailable)
Usage	Usage changelog
GitHub CLI	GitHub CLI releases

How It Works

data/raw/ stores daily imports of official changelog entries. Records are written per version and are never overwritten. data/curated/ holds human-reviewed, plain-language versions of those entries in five languages.

scripts/build.py combines both layers into generated artifacts:

• ai_updates.json: compatibility feed for the legacy application, containing up to three latest curated versions per tool.
• daily.json: daily feed containing up to three latest versions per tool; it uses curated content when available and otherwise retains the official source text with curated: false.
• docs/data.json: version index for the static website; complete raw and curated history for each tool is stored in docs/history/<tool_id>.json so the homepage does not need to load the full history at once.

Repository Layout

data/
  raw/<tool_id>/<version>.json       # Immutable official source records
  curated/<tool_id>/<version>.json   # Human-reviewed five-language records
docs/
  index.html                         # GitHub Pages site
  data.json                          # Generated website version index
  history/<tool_id>.json             # Generated full history per tool
  v/                                 # Generated per-version static pages
  feed.xml                           # Generated RSS feed
  sitemap.xml                        # Generated sitemap
  robots.txt                         # Generated robots.txt
  llms.txt                           # Generated LLM reference file
scripts/
  fetch.py                           # Fetch official changelogs
  build.py                           # Build generated feeds
  sync_agy.py                        # Import local Antigravity CLI changelog output
.github/workflows/
  daily.yml                          # Daily fetch and review-issue workflow
  sync-usage.yml                     # Compatibility-feed sync workflow
ai_updates.json                      # Generated compatibility feed
daily.json                           # Generated daily feed
tests/                               # Test suite

Local Development

Requires Python 3.13. The repository's runtime scripts use only the Python standard library.

python3 scripts/fetch.py
python3 scripts/build.py
uv run ruff check scripts tests
uv run pytest -q

The test suite builds into a temporary directory and never writes to the working tree, so the two commands above are safe to run in any order.

If the public Antigravity source is unavailable, import changelog output from an installed CLI instead:

python3 scripts/sync_agy.py

This command runs agy changelog; existing version records are not overwritten.

Curation Workflow

1. The daily workflow opens an issue for each newly imported version.
2. Create data/curated/<tool>/<version>.json with the reviewed five-language plain-language items.
3. Run python3 scripts/build.py and pytest.
4. Open a pull request and merge it after review.

Curated content must accurately reflect the official entries. The original field retains the original English text.

Automation

• daily.yml runs daily, fetches official changelogs, opens review issues for newly discovered versions, builds generated feeds, and commits changed source and feed files.
• sync-usage.yml runs when ai_updates.json changes on main and synchronizes that file to aqua5230/usage.

Related Project

aqua5230/usage is the primary consumer of the compatibility feed.

Maintainer Notes

sync-usage.yml requires a fine-grained personal access token named USAGE_SYNC_TOKEN.

1. In GitHub, open Settings → Developer settings → Personal access tokens → Fine-grained tokens.
2. Set the resource owner to aqua5230, choose Only select repositories, and select aqua5230/usage.
3. Grant only Contents: Read and write repository permission.
4. Add the token as the USAGE_SYNC_TOKEN Actions secret in this repository under Settings → Secrets and variables → Actions.

Do not store the token in repository files or workflows. Replace it when it expires.