AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
OpenFormosa is an open family of AI foundation models for Taiwan's language contexts. It helps AI speak Taiwan, hear Taiwan, and read Taiwan.
Developer toolsJavaScript · HTMLGitHub @OpenFormosaOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
No updates detected yet. GitHub releases will appear here.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
OpenFormosa website — openformosa.com. Jekyll + GitHub Pages. Taiwan-rooted open AI foundation models.
Translation pending
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
The source for openformosa.com — a Jekyll site for OpenFormosa, a Taiwan-rooted open AI initiative around model training, benchmark evaluation, a pretrain base model, ASR, TTS, OCR, and local culture.
Deployed to GitHub Pages from main via the pages.yml GitHub Actions workflow.Read the full README ↗The source for openformosa.com — a Jekyll site for OpenFormosa, a Taiwan-rooted open AI initiative around model training, benchmark evaluation, a pretrain base model, ASR, TTS, OCR, and local culture.
Deployed to GitHub Pages from main via the pages.yml GitHub Actions workflow.
data-i18n layer.The site uses UTF-8 source files, so export a UTF-8 locale before building or a bundled theme's SCSS will error on macOS's default US-ASCII locale.
bundle install
export LANG=en_US.UTF-8 LC_ALL=en_US.UTF-8
bundle exec jekyll serve
Then open http://127.0.0.1:4000/.
For a static preview that mirrors the deployed build:
export LANG=en_US.UTF-8 LC_ALL=en_US.UTF-8
bundle exec jekyll clean
bundle exec jekyll build
node scripts/validate-public-build.mjs _site
python3 -m http.server 4000 --bind 127.0.0.1 --directory _site
Deployment is automatic: every push to main runs pages.yml, which builds with a generated production config, validates _site, and deploys to GitHub Pages.
The production url / baseurl come from repository Actions variables (Settings → Secrets and variables → Actions → Variables):
SITE_URL to https://openformosa.com and leave SITE_BASEURL unset (GitHub rejects an empty variable value, so for a root domain the variable is simply not created and the workflow defaults baseurl to "").https://<owner>.github.io/<repo>): set SITE_URL to https://<owner>.github.io and SITE_BASEURL to /<repo>.The committed CNAME file (openformosa.com) is copied into _site on every build so the custom domain persists across deploys.
To enable it on a fresh repository:
SITE_URL repository variable as above.openformosa.com.export LANG=en_US.UTF-8 LC_ALL=en_US.UTF-8
cp _config.production.example.yml _config.production.yml # already targets openformosa.com
bundle exec jekyll clean
bundle exec jekyll build --config _config.yml,_config.production.yml
node scripts/validate-public-build.mjs _site --require-absolute --expected-base="https://openformosa.com"
Before announcing changes, check sitemap.xml, robots.txt, feed.xml, and the social preview image. See PUBLICATION_CHECKLIST.md.
Blog posts live in _posts/*.md and are rendered by Jekyll/kramdown with GitHub-flavored Markdown. Posts also load MathJax, so authors can use inline math like $E=mc^2$ and display math blocks like:
$$
\operatorname{WER} = \frac{S + D + I}{N}
$$
Use fenced code blocks, tables, headings, lists, links, and ordinary Markdown normally. For bilingual post bodies, wrap each language in <div class="post-lang-zh" markdown="1">…</div> and <div class="post-lang-en" markdown="1">…</div>; the active language is toggled by the site's language switcher.
This site documents model training, evaluation, and release evidence. It does not collect community datasets or accept raw data uploads. See CONTRIBUTING.md and SECURITY.md.