This version of the module is deprecated: Merged into conglinyizhi/moonbit-llm-faceoff: the page, its server and the static report now ship in that single module

moonbit-llm-faceoff-web

Static page for the faceoff benchmark results (Rabbita SSG + precss)

moonbit

rabbita

llm

benchmark

report

Download zip

Author

conglinyizhi

Version

0.0.1

License

MIT

Repository

github.com/conglinyizhi/moonbit-llm-faceoff

Last updated

7 days ago

Downloads

Dependencies

#moonbit-llm-faceoff-web

The web half of faceoff: a page that runs a case set against several OpenAI-compatible models, shows the answers side by side, and lets you annotate them — plus the static report generator for the same data.

A separate module from conglinyizhi/moonbit-llm-faceoff on purpose: the page consumes the benchmark CLI's exported data.json instead of calling into the library, which keeps the statistics in one implementation.

#Packages

shared — parsing, metrics, diff and rendering, used by both the page and the report. This is the part worth depending on: moon addconglinyizhi/moonbit-llm-faceoff-web, then import conglinyizhi/moonbit-llm-faceoff-web/shared.
cmd/server — the local server the page talks to (/api/meta, runs, case sets, presets, annotations).
cmd/ssg — the static report generator.
cmd/app — the page itself (Rabbita).

#Running it

Cloning the repository is still the way: the page build also needs scripts/build-web.mbtx and the Makefile, which sit outside this module.

make dev         # dev server: rebuilds on change
make serve       # build once, then serve
make serve-demo  # mock endpoint + demo data + page, for a quick look

The main repository's README has the full picture, including how to point the page at a real gateway.

Site source Report issue Packages Build queue Skills Statistics