The clients page was slow on panels with many clients for two independent reasons: the server rebuilt the whole picture on every request, and the browser rebuilt the whole table on every poll. Server side, ListPaged loaded every client row, every client_inbounds link and every client_traffics row into Go memory, then filtered, sorted and paginated in a loop -- on a request the page repeats every five seconds. Every predicate now runs in SQL and only the requested page's ids are hydrated, so the cost tracks the page size rather than the client count. Measured on SQLite with a realistic status mix: the default view at 100k clients goes from 1,072ms to 64ms. Behaviour is preserved deliberately in the subtle places -- the cross-panel global-traffic overlay is folded into the same used-bytes expression the predicates and sort use, LIKE wildcards are escaped so a search for "a_b" stays literal, and the two different tiebreak rules the in-memory comparator had are reproduced per sort key. The summary's per-bucket email lists are capped at 200 with exact counters beside them. They only back hover popovers, but shipping every match made the response grow with the panel: at 100k clients it carried ~42k emails, and the page revalidated all of them through a strict Zod parse every five seconds. The popover now shows a "+N" chip for the remainder. Browser side, the page fired three sequential list requests per load and threw the first two away: the query went out before the persisted sort was applied, and again before the configured page size was known -- 0 meaning "one long page" is indistinguishable from "not loaded yet". The page size is now derived rather than mirrored through an effect, and the previous visit's value is remembered so the single request goes out at mount instead of queueing behind /setting/defaultSettings. Then the per-poll work. Reading isFetching made it a tracked property, so the refetch interval notified twice per cycle and re-rendered the page even when structural sharing left the data identical. Xray reports a traffic row per client whether or not it moved bytes, so the speed map was mostly zeros and was replaced wholesale every push; zero rows are now dropped and an unchanged result returns the previous object, which lets React bail out instead of re-rendering. The five Tooltip-wrapped buttons and the inbound chips per row do not depend on traffic at all and are now memoised, keyed on the email because a push replaces the row object of every client whose counters moved. antd's hashed:false drops 3,311 :where(.css-<hash>) wrappers and 29% of the generated stylesheet, and a pinned cssVar key stops each of the eleven page-level ConfigProviders minting its own token scope. Two callers that only need the mutations, GroupsPage and ClientBulkAddModal, no longer start the list query -- the groups page had been polling the full paged list every five seconds for data it never renders.
3x-ui frontend
React 19 + Ant Design 6 + TypeScript + Vite 8. Three SPA bundles —
index.html (admin panel SPA, all /panel/* routes), login.html
(login + 2FA), and subpage.html (public subscription viewer). All
three are built into ../internal/web/dist/ and embedded into the Go binary
via embed.FS.
State is split between local useState, TanStack Query for server
state, and useTheme / useWebSocket contexts. Form validation,
API parsing, and the xray config model all run through a single
shared Zod schema tree (see Schemas).
Dev
npm install
npm run dev
Vite serves on http://localhost:5173/. API calls and /panel/*
routes proxy to the Go panel at http://localhost:2053/, so start
the Go panel first (go run main.go) and then Vite. The proxy
auto-rewrites /panel, /panel/settings, /panel/inbounds,
/panel/xray to the matching Vite-served HTML, so the sidebar's
production-style links work without round-tripping through Go.
Scripts
| Command | What |
|---|---|
npm run dev |
Vite dev server with API + WS proxy to Go |
npm run build |
Regenerates OpenAPI + Zod, then builds into ../internal/web/dist/ |
npm run preview |
Serve the built bundle locally |
npm run typecheck |
tsc --noEmit (strict, no emit) |
npm run lint |
ESLint flat config (@typescript-eslint + react-hooks) |
npm run test |
Vitest single run (schema fixtures, link parsers, …) |
npm run test:watch |
Vitest watch mode |
npm run storybook |
Storybook dev server on :6006 (component workbench + autodocs) |
npm run build-storybook |
Static Storybook build — CI compile-checks every story |
npm run gen:api |
Build public/openapi.json from pages/api-docs/endpoints.ts |
npm run gen:zod |
Run the Go-side openapigen tool → src/generated/{zod,types}.ts |
CI runs typecheck, lint, test, build, and build-storybook on
every PR (see ../.github/workflows/ci.yml).
One-off: scan for deprecated APIs
Run this command to sweep the codebase for usages of APIs marked
with the JSDoc @deprecated tag (AntD prop renames, Zod renames,
removed Web APIs, etc.):
npx eslint --config eslint.deprecated.config.js src
It's a type-aware ESLint run against eslint.deprecated.config.js
and is not wired into npm run lint because typed linting triples
the wall-clock time.
Production build
npm run build
Outputs to ../internal/web/dist/ (HTML at the root, hashed JS/CSS under
assets/). manualChunks splits AntD, icons, codemirror, and
react-query into separate vendor bundles to keep the per-page
initial JS small. The Go binary embeds this directory at compile
time and internal/web/controller/dist.go serves the per-page HTML.
Layout
frontend/
├── index.html, login.html, subpage.html # 3 Vite entries
├── tsconfig.json
├── eslint.config.js
├── eslint.deprecated.config.js # On-demand type-aware lint config that flags
│ # usages of APIs marked with JSDoc @deprecated
├── vitest.config.ts
├── vite.config.js
├── .storybook/ # Storybook config (main.ts, preview.tsx)
├── scripts/
│ └── build-openapi.mjs # endpoints.ts → openapi.json
└── src/
├── entries/ # Per-page bootstrap (createRoot + render)
├── main.tsx # Shared root for the admin SPA (index.html)
├── routes.tsx # react-router routes mounted under /panel/
├── pages/ # One folder per route, page component + helpers
│ ├── index/, login/, inbounds/, clients/, xray/, nodes/,
│ ├── settings/, api-docs/, sub/
├── layouts/ # AdminLayout (sidebar + header + outlet)
├── components/ # Cross-page React components (+ co-located *.stories.tsx)
├── hooks/ # useClients, useTheme, useWebSocket, …
├── api/ # fetch client + CSRF handling, TanStack Query bridge,
│ # WebSocket client + queryClient.ts
├── i18n/ # react-i18next init (locales in internal/web/translation/)
├── lib/xray/ # Pure functions: link generation, defaults,
│ # form ⇄ wire adapters, protocol capabilities
├── schemas/ # Zod source-of-truth (see "Schemas" below)
├── generated/ # Code-generated zod + ts types from Go
│ # (DO NOT hand-edit — regenerated by gen:zod)
├── models/ # Thin legacy types still in transit
│ # (DBInbound, Status, AllSetting)
├── styles/ # Shared CSS modules
├── test/ # Vitest specs + golden fixtures
│ ├── *.test.ts
│ ├── __snapshots__/
│ └── golden/fixtures/ # Per-(protocol × network × security) JSON
└── utils/ # HttpUtil, ClipboardManager, SizeFormatter, …
Schemas
src/schemas/ is the single source of truth for the xray
configuration model. Every API response is parsed through it,
every form field is validated against it, and TypeScript types
are inferred via z.infer<typeof X> — never hand-written.
schemas/
├── primitives/ # Atomic reusable schemas (port, protocol, sniffing, …)
├── api/ # Backend response shapes (e.g. SlimInboundSchema)
├── forms/ # User-facing form shapes (narrower than api/)
├── protocols/
│ ├── inbound/ # Per-protocol settings (vmess, vless, trojan, …)
│ ├── outbound/
│ ├── stream/ # Network transports (tcp, ws, grpc, xhttp, kcp, …)
│ └── security/ # TLS, Reality, none
├── client.ts, dns.ts, routing.ts, setting.ts, status.ts, xray.ts
└── _envelope.ts # Generic `Msg<T>` envelope wrapper
Patterns:
- Discriminated unions for polymorphic data — inbound
settingsisz.discriminatedUnion('protocol', […]), same for stream and security. - Three validation layers, non-overlapping:
- API boundary:
parseMsg(msg, schema, ctx)inside TanStack QueryqueryFn— warn-only in prod, throws in dev - Form input:
antdRule(schema.shape.field)on every<Form.Item>— blocks submit + per-field inline error - Wire request:
Schema.parse(payload)insidemutationFn— throws, because a malformed payload here is always a developer bug
- API boundary:
- No
.loose()or[key: string]: anyin production schemas.@typescript-eslint/no-explicit-any: erroris enforced.
Form pattern (Pattern A)
All non-trivial modals use this single pattern:
const [form] = Form.useForm<InboundFormValues>();
const onFinish = async () => {
const values = await form.validateFields();
await createInbound.mutateAsync(values);
};
<Form form={form} onFinish={onFinish}>
<Form.Item
name="port"
label="Port"
rules={[antdRule(InboundFormSchema.shape.port, t)]}
>
<InputNumber min={1} max={65535} />
</Form.Item>
</Form>
No safeParse-on-submit handlers, no useRef<any> for form
references, no inline z.string().min(1) in rules. Conditional
fields use <Form.Item dependencies={...} shouldUpdate> with the
nested protocol schema.
Testing
Vitest runs everything under src/test/. Schemas have golden
fixture suites — one JSON per (protocol × network × security)
combination round-tripped through schema.parse → link generator
→ snapshot. Regenerate snapshots after intentional changes:
npx vitest run -u
Fixtures live in src/test/golden/fixtures/ and are auto-discovered
via import.meta.glob.
Storybook
Reusable components in src/components/ are developed and documented in
Storybook (@storybook/react-vite). It is a component workbench, not part
of the shipped panel — nothing here is embedded into the Go binary. The built
Storybook is published with the docs site at
docs.sanaei.dev/storybook by
.github/workflows/docs-deploy.yml.
npm run storybook # dev server on http://localhost:6006
npm run build-storybook # static build; CI runs this to compile-check every story
Addons: @storybook/addon-docs renders an autodocs page per component,
@storybook/addon-a11y flags accessibility issues in the canvas, and
@storybook/addon-vitest runs every story as a headless-browser test under
npm run test (Playwright/Chromium — run npx playwright install chromium once
locally). The .storybook/preview.tsx decorator wraps every story in the AntD
ConfigProvider and adds a light/dark theme toggle to the toolbar.
Conventions for a story:
- Co-locate it with its component as
<Component>.stories.tsx. - Set
tags: ['autodocs']so it gets a generated docs page. - Document props via story metadata, not JSDoc (the repo bans
//comments): a component summary inparameters.docs.description.componentand per-prop text inargTypes[prop].description.satisfies Meta<typeof Component>keeps the metadata type-checked.
Adding a new page
Most new routes go inside the admin SPA (index.html) via
routes.tsx — no new HTML or Vite entry needed.
- Add the page component under
src/pages/<page>/. - Register it in
src/routes.tsxunder the/panel/...tree. - If you need a brand-new top-level bundle (login-style standalone
page), add the HTML at
frontend/<page>.html, an entry atsrc/entries/<page>.tsx, and register it inrollupOptions.inputinvite.config.js. Then add the Go controller call toserveDistPage(c, "<page>.html").