mem9-ai/drive9-fe

최근 기록된 소스 활동: 2026년 10월 27일 10:07

SkillsMP 카탈로그 업데이트: 2026년 11월 12일 12:52

수집된 skills: 1

GitHub 스타: 0
GitHub 포크: 4

AI 어시스턴트로 설치

이 프롬프트를 복사해 사용 중인 AI 어시스턴트에 입력하세요.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/mem9-ai/drive9-fe.

저장소 README

README 원본 파일: README.md

README 원본 보기

README 전체 펼치기

# Drive9 website The existing website is plain HTML in `site/`. Its homepage is `site/index.html`; there is no React router for the marketing pages. ## Studio build `vendor/studio/` contains the Console's built static files. The source repository, full commit, Node/npm versions, build command and SHA-256 for every file are in `vendor/studio-build.json`. These files are reviewed and committed together. Netlify verifies the files, copies them under `site/studio/` and generates routing; it does not fetch `tidb.ai`, install its dependencies or need a private deploy key. There is no Git submodule. Automatic cross-repository publication is handled by the source repository workflow; Netlify only receives this public website checkout. To prepare a website checkout with Node 24.22.1: ```sh node --test scripts/*.test.mjs tests/*.test.mjs STUDIO_API_ORIGIN=https://api.example.invalid node scripts/prepare-studio.mjs ``` The example origin is a placeholder, not a working backend. Set the real public HTTPS API origin privately in the deployment environment. Do not include a path, trailing slash, credentials, or query string. Generated `site/studio/` and `site/_redirects` are ignored by Git. The homepage, downloads and documentation remain in `site/`; only the Console's static output is copied under `/studio/`. ### Automatic publication gate The source repository's **Publish Studio Console** workflow uses a dedicated GitHub App installed only on this repository, with Contents and Pull requests write plus Checks read. It updates the `automation/studio-console` branch, waits for the required check, and merges its artifact PR. The **Studio artifact** check in this repository independently validates checksums, exact source/build metadata, allowed public files, common credential markers, and website routes. Generated publication PRs may change only `vendor/studio/` and `vendor/studio-build.json`. The check has no deployment credentials and runs for every PR, including fork PRs. It uses `pull_request_target` and explicitly checks out the immutable trusted default-branch revision. It reads candidate Git blobs at the event's exact head SHA as data: it never checks out or executes candidate scripts/workflows. All bot-authored/updated PRs and the managed publication branch are restricted to artifact paths, including changes/deletions of validators and workflows. Human code PRs still run ordinary **Website tests** separately. The trusted workflow alone has Checks write, solely to report **Studio artifact** against the candidate head (the target event's own SHA is the default branch). The publishing App has Checks read and no Workflows permission; it cannot write this result or alter workflows. See GitHub's [trusted-event guidance](https://docs.github.com/en/actions/reference/security/securely-using-pull_request_target). Before enabling the publisher, an administrator must require **Studio artifact** on main, from the GitHub Actions integration, require an up-to-date branch, and require a PR with zero mandatory human approvals. Enforce the rule for administrators too and add no App bypass. This lets a successful generated PR merge without a human click while pending/failed checks block it. These are GitHub settings; this workflow file alone does not enable protection. Confirm the live settings and an App-authored PR before declaring automatic publication operational. No private-source credential is available here. Provenance validation checks the declared source repository, full SHA, pinned toolchain/build command and all artifact hashes; it does not independently rebuild private source. Marker checks cannot prove a bundle secret-free. Review publication tooling and any deliberately public frontend literals accordingly. ### Updating Studio manually Use an authorized local checkout of `mem9-ai/tidb.ai`, Node 24.22.1 and npm 11.6.2. Choose a reviewed full source commit. Export it into an empty directory so local untracked files, `.env` files and provider configuration cannot enter the build. From this website repository, reproduce the recorded version as follows: ```sh studio_revision=$(node -p "require('./vendor/studio-build.json').commit") studio_source=$(mktemp -d) git -C /path/to/tidb.ai archive "$studio_revision" | tar -x -C "$studio_source" ( cd "$studio_source" env -i HOME="$HOME" PATH="$PATH" npm ci --no-audit --no-fund env -i HOME="$HOME" PATH="$PATH" npm run build:console -- --base=/studio/ ) node scripts/studio-artifact.mjs "$studio_source/dist/console" "$studio_revision" node --test scripts/*.test.mjs tests/*.test.mjs ``` For an update, set `studio_revision` to the newly reviewed full commit instead. The import command accepts only Console HTML, its favicon, JS/CSS/font assets, config and the reviewed files in `skills/tidb-agent-workspace/` and its legacy `skills/drive9-workspace/` alias (`SKILL.md`, `private-csv.md`, `workspace-api.md`, `install.py`, `report.py` and `private_csv.py`, at most 64 KiB each); it rejects source maps, unexpected files and symlinks. It replaces the local demo config with the same-origin `/studio` config and removes stale assets. Review the complete artifact and manifest diff, including a secret/internal-data scan, before committing. File hashes detect drift, not whether a bundle is safe. To check reproducibility, rebuild from a second clean export and compare the file hashes after the same config normalization. Only static files are published; the source manifest lives outside `site/`. Remove the temporary export when finished. ## Netlify Keep the Netlify project's **base directory set to `site`**. The checked-in `site/netlify.toml` runs tests and preparation via `../scripts/` and publishes `.` (the current `site` directory). Publish paths are relative to the base, so setting publish to `site` would incorrectly select `site/site`. Only access to this website repository is needed. The custom `build.ignore` command checks `site/`, `vendor/`, `scripts/` and `tests/` from the repository root. Netlify's default comparison only covers its `site/` base, so it can skip a Console-only artifact update outside that directory. Exit 0 skips unchanged inputs; exit 1 builds. Missing/invalid prior commits, unavailable shallow history and Git errors conservatively build. Documentation-only changes can skip. Keep this input list aligned with future build dependencies. `tests/netlify-ignore.test.mjs` runs the configured command from `site/` against isolated Git history, including artifact-only updates and deletions. See [Netlify's ignore-build contract](https://docs.netlify.com/build/configure-builds/ignore-builds/). The `tests/netlify-build.test.mjs` integration test executes the configured command from `site` in a temporary checkout and checks the published output and missing origin failure. It lives outside the build's `scripts/*.test.mjs` glob to avoid recursively running itself. Set `STUDIO_API_ORIGIN` in the Netlify project's environment variables, with the Builds scope and the deploy contexts you intend to publish (Production for the live site). A missing/invalid origin or modified/incomplete artifact fails the build. Rebuild after changing the variable. No GitHub App, private submodule key or new Netlify deployment token is required by this build. Deploy the API independently using the application's deployment requirements. This repository does not run the API, provision storage or configure OAuth. The generated [rewrite rules](https://docs.netlify.com/manage/routing/redirects/rewrites-proxies/) keep the Console's current public API contract: | Browser path | Destination | | --- | --- | | `/studio`, `/studio/` | Console entry, using Netlify's directory handling | | `/studio/assets/*` | Built static files | | `/studio/*` | Console entry, only when no file matches | | `/studio/config.json` | Studio configuration, `apiOrigin: ""`, `apiPathPrefix: "/studio"` | | `/studio/api/*` | API origin, preserving `/api/*` | | `/studio/openapi.json` | API schema | All Studio routes stay under `/studio`; website root API/config paths are untouched. There is no global SPA fallback: existing pages and unknown non-Studio paths retain their original behavior. API errors must remain API errors rather than HTML. Do not add a redirect that only adds/removes a trailing slash: Netlify [matches both forms](https://docs.netlify.com/manage/routing/redirects/redirect-options/#trailing-slash), which makes such a forced redirect loop back to itself. For TiDB Cloud SSO, register the callback `https://drive9.ai/studio/api/auth/tidbcloud/callback` and set the backend's post-login URL to `https://drive9.ai/studio/`. Set `WORKSPACE_PUBLIC_PATH=/studio` on the backend so session and login-flow cookies use the same prefix. The gateway removes `/studio` when forwarding to the backend; its internal routes stay unchanged. Keep OAuth secrets and owner allowlists only in the backend's private runtime configuration. Verify the backend's public origin, proxy trust and secure-cookie settings before enabling it for users. The pinned application includes the merged SSO and public-path support. Live OAuth configuration is separate from publishing these static files. After deploying, verify the homepage, a direct Studio refresh, JS/CSS loading, API error responses, login callback and logout on the actual Netlify domain. Local build/tests do not constitute live OAuth or Netlify deployment acceptance. ## Workspace discovery pages The Studio artifact may include only the approved `discover/` entry, get-started, sales-analysis, API and privacy pages, plus `robots.txt`, `sitemap.xml` and `llms.txt`. These files retain source checksums and credential-marker checks; arbitrary nested HTML, source maps and symlinks are still rejected. The importer and trusted Git-candidate validator share the same public-file allowlist. Validate a complete newly built artifact, including discovery pages, through both paths before reporting a Console update as published; passing tests against the old vendored artifact alone does not verify publication of new file types. Production preparation gives the existing public pages canonical URLs under `https://drive9.ai/studio/discover/`, adds them to the root sitemap, and links the Workspace guide from the homepage. Preview preparation keeps noindex and omits canonicals. Unknown discovery URLs return 404 before the application fallback. The application and `/studio/admin` retain noindex. Robots allow crawlers to see those noindex rules; backend authorization protects the data. Admin additionally requires TiDB Cloud authentication and a verified PingCAP email in the backend. Neither this website nor its build reads or configures backend identities/secrets.

저장소 README를 불러오는 중…

이 저장소의 skills

소프트웨어 개발자

직업 카테고리 1개 · 100% 분류됨

skill

직업 분류

설명

업데이트

drive9

소프트웨어 개발자

Persistent network filesystem for AI agents - store, mount, search, and share files across sessions with contexts and vault secrets.

원문 언어: 영어

2026-10-27

수집된 skill 1개 중 1개를 표시합니다.