xLLM Official Engineering Insights

Static GitHub Pages report for the official jd-opensource/xllm repository.

Live Site

View the report: https://pjgao.github.io/xllm-commit-analysis/

The site shows:

• cumulative repository contribution metrics
• cumulative hot module distribution
• cumulative contributor ranking
• daily activity, module, contributor, risk, and test-focus analysis
• commit-level records and CSV export

Local Usage

npm run build:data
npm run serve

Local preview:

http://localhost:4175

Automation

GitHub Actions runs daily and on manual dispatch:

1. fetches jd-opensource/xllm
2. runs scripts/build-official-data.mjs
3. generates AI summaries for new commits through GitHub Models
4. updates site/data/official-xllm.json
5. deploys site/ to GitHub Pages

AI summary behavior:

• Browser-side code never calls AI APIs.
• GitHub Actions uses GITHUB_TOKEN with models: read.
• Existing commit and daily summaries are reused from site/data/official-xllm.json.
• Only new commits without aiSummary are sent to the model.
• Daily summaries are generated into dailySummaries[YYYY-MM-DD] and regenerated only when that day's commit set changes.
• AI_SUMMARY_MAX_PER_RUN defaults to 20 to control commit-summary usage.
• AI_DAILY_SUMMARY_MAX_PER_RUN defaults to 7 to control daily-summary usage.
• AI_MODEL defaults to openai/gpt-4.1-mini; configure repository variables to switch models.

For local data builds without AI:

npm run build:data

For local AI summary generation, provide a GitHub token with GitHub Models access:

$env:ENABLE_AI_SUMMARY = "true"
$env:GITHUB_TOKEN = "<token>"
$env:AI_SUMMARY_MAX_PER_RUN = "5"
$env:AI_DAILY_SUMMARY_MAX_PER_RUN = "2"
npm run build:data