The ML Platform Behind Uber's AI
Michelangelo AI: 우버의 엔드‑투‑엔드 머신러닝 플랫폼
아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
메이커가 쓴 소개는 위의 한 줄이 전부입니다. 더 자세한 내용은 저장소 README에서 읽을 수 있습니다. 저장소 README ↗
README 전체를 불러오는 중입니다.
README 전문 보기 ↗Read-only control dashboard for multi-agent software projects. It shows the task board, pending approvals, session names and token costs for any project that keeps its agent state in markdown tables.
It makes no model calls. Opening it costs no tokens.
project repo (docs/agents/*.md) → GitHub API (read-only token, cached 60 s)
→ adapter (markdown tables → typed model, validated)
→ Board · Task · Approvals · Costs · Health
The markdown format is the contract. See docs/FORMAT.md. Projects are listed in pantheon.config.json; adding one is a config entry, not a code change (the sample project proves it).
npm install
npm run dev # http://localhost:3000
npm run test:run # parser, validation, naming, source tests
npm run typecheck
npm run lint
npm run build
Without GITHUB_READ_TOKEN it shows the bundled fixture snapshot under fixtures/, with a visible banner. With the token it reads the live repo. Copy .env.example to .env.local to set one.
pantheon.config.json: id, name, source (github repo and branch, or fixtures directory), the paths of the board, approvals and usage files, and the session and branch naming patterns.Import the repo as a Vercel project (framework: Next.js, no custom build command). Set GITHUB_READ_TOKEN in the environment (fine-grained, contents: read, only the tracked repos). Turn on Vercel Authentication so only the owner can open it.