Files
deepseek-harness/docs/subsystems/web-server.md
T
imccyu 73931626be feat(webserver): add resource routes and in-process readResource (#6901)
* feat(webserver): add resource routes and in-process readResource

A webserver route or the fallback seat may now answer through a resource,
a function from a Fetch Request to a Response, besides the existing
node:http handler. register accepts either route form and registerFallback
accepts a handler or { resource }; existing handler callers are unchanged.
HTTP GET and HEAD requests reach a resource; webServer.readResource(request,
peer) follows the same match order and reaches only resources, so handler
routes stay unreachable to in-process readers.

client-modules registers /plugins as the fetchBundle resource, and
frontend-static claims the fallback seat with a resource that renders the
index for a supplied Peer, or runs Connection's existing authorizeIndex for
an HTTP request and renders for the browser Peer.

* fix(webserver): flush resource headers and keep one drain listener per response

Resource responses send their status and headers before the first body
chunk, register one drain/close listener pair per response so the gzip
middleware's zlib stream does not accumulate listeners across backpressure
waits, and readResource drops the body of a HEAD read as HTTP does. HTTP
requests and readResource share one responder lookup.

* chore(scripts): prune the retired client-modules unknown-cast baseline entry

* feat(frontend-static): let browsers revalidate dist files except the index

Every dist file other than the rendered index now carries
cache-control: no-cache and a weak ETag from a SHA-1 of its bytes; a
request whose If-None-Match equals that ETag receives an empty 304. The
index keeps sending no cache headers. In-process reads through
webServer.readResource revalidate the same way.

* fix(webserver): stop reading Host for resource URLs and copy no held body

A resource request's URL is the request target resolved against the same
placeholder origin the route match uses, so a missing or unparsable Host no
longer turns a dist or /plugins request into 400. The webserver ends a
resource response itself and treats any earlier close as an abort, which
also covers the Web Worker tunnel marking an aborted response ended first.
client-modules bundle bodies and frontend-static file bodies are the held
bytes themselves instead of copies, and frontend-static no longer exports
its internal serveStatic helper.

* docs(webserver): document resource route types and the handler-or-resource fallback

Paste WebResource and WebResourceRoute beside WebRoute on the web-server
subsystem page and link the catalog signatures to it, replace the remaining
'fallback handler' wording now that the seat takes a handler or a resource,
and note why the drain listener removal is a no-op under gzip.
2026-10-22 13:41:18 +08:00

348 lines
19 KiB
Markdown

# HTTP Server
English | [中文](web-server.zh.md)
[dsh-host-webserver](../../packages/host/webserver) is the browser HTTP carrier for the GUI host: one `node:http` server — `node:https` when `tls` is configured — providing `ctx.webServer`, a named-route registry, optional gzip response compression, index.html transform callbacks, and one fallback, a handler or a resource, that a plugin may claim. It is not part of the agent loop and not a capability seam; its only harness concept is the Peer an index render speaks for, and another plugin registers every feature route, including the `/api` bridge, plugin bundles, and the HMR event stream ([layering reference](../../packages/boot/app-boot/README.md)). It serves browsers only: Electron loads the built files over `file://` and sends fetch requests through an IPC bridge instead of this server.
Source: [`packages/host/webserver/src/index.ts`](../../packages/host/webserver/src/index.ts)
## Routes
```ts type-equiv
/** Route match kind: 'exact' matches the pathname verbatim; 'prefix' p matches p and p/<anything>. */
type WebRouteKind = 'exact' | 'prefix'
```
```ts type-equiv
/** One named route registration. */
interface WebRoute {
kind: WebRouteKind
/** Absolute pathname, no trailing slash. */
path: string
/** Owns the full response lifecycle (may hold the response open, e.g. SSE). */
handler: (req: IncomingMessage, res: ServerResponse) => void | Promise<void>
}
```
```ts type-equiv
/**
* Answers one GET or HEAD request with a Fetch response. An HTTP request
* arrives without a Peer and carries the client's headers; an in-process
* {@link WebServer.readResource} read passes the Peer its caller admitted.
*/
type WebResource = (request: Request, peer?: PeerScope) => Promise<Response>
```
```ts type-equiv
/** One named route answered by a resource instead of a handler. */
interface WebResourceRoute {
kind: WebRouteKind
/** Absolute pathname, no trailing slash. */
path: string
/** Answers HTTP GET and HEAD requests and in-process reads. */
resource: WebResource
}
```
Match order is fixed: exact table first, then longest matching prefix, then the registered fallback. Registration order carries no request-facing semantics — named routes are composed to be disjoint, and the fallback seat answers anything no named route claims; one owner only, a second registration throws. The shipped Web composition claims the seat with [`dsh-host-frontend-static`](../../packages/host/frontend-static/src/index.ts), the SPA dist server with locked semantics: Connection authenticates the dist root and configured index before their HTML is read; non-index assets remain public; non-GET/HEAD is 405, traversal outside the dist root is 403, existing files are served directly, absent or non-file targets are empty 404 responses, and unknown extensions ship as octet-stream.
A route or the fallback may give a resource, a function from a Fetch `Request` to a `Response`, instead of a handler. HTTP `GET` and `HEAD` requests reach it without a Peer, and `readResource(request, peer)` reads it in process for the Peer its caller admitted; a read follows the same match order but reaches only resources, so a route that returns owner-private data or performs an action stays a handler even when it answers `GET`.
<a id="config"></a>
## Config
```ts type-equiv
/** Web server listen, TLS, and response-compression config. */
interface Config {
/**
* Listen address: a concrete IPv4 or IPv6 literal of one local interface,
* for example the container's own Pod address from `hostname -i`. A loopback
* literal (any address in 127/8, `::1`, or a mapped form of either) keeps the
* server on this machine; any other literal serves the network that address
* belongs to, over plain HTTP unless `tls` is set. The unspecified address —
* IPv4 any, IPv6 any, and the IPv4-mapped forms of IPv4 any — is rejected at
* load: it would expose the port on every interface at once.
*/
host: string
/** Listen port; zero requests an OS-assigned port. */
port: number
/**
* Serve HTTPS with this certificate and key instead of plain HTTP. Both files
* are read once, before the listener binds: an unreadable or empty file,
* invalid PEM, or a key that does not match the certificate rejects
* initialization rather than falling back to HTTP. The material is never
* re-read, so replacing a certificate takes a reload. Omitted or null listens
* over plain HTTP.
*/
tls?: TlsConfig
/** Response compression for socket-backed HTTP requests. @default 'none' */
compression?: 'none' | 'gzip'
/** Gzip DEFLATE level from 0 through 9. @default 1 */
compressionLevel?: number
/** Minimum known response length eligible for gzip; unknown-length streams are eligible. @default 1024 */
compressionThresholdBytes?: number
}
```
```ts type-equiv
/**
* TLS material for the HTTPS listener: one certificate chain file and the
* private key file it pairs with. Both hold PEM text and both resolve against
* the process working directory.
*/
interface TlsConfig {
/** Certificate chain file, leaf certificate first, PEM, no passphrase. */
certFile: string
/** Private key file for the chain's leaf certificate; unencrypted PEM. */
keyFile: string
}
```
No composition can bind every interface at once: the unspecified address is rejected at load. The exported `isWildcardHost` and `isLoopbackHost` classify a bind address from its parsed value, and an IPv6 `%zone` is kept for `listen` while classification reads the address alone. The carrier owns no authentication or Origin policy, and it serves plain HTTP unless `tls` supplies a certificate and key. `compression` defaults to `none`; the shipped Web bundle selects gzip level 1 with a 1024-byte threshold. The shipped `dsh web` command selects loopback by default and accepts one concrete `--host` address, warning at startup when a non-loopback bind serves plain HTTP; its Connection plugin supplies Host/Origin checks plus browser-session authentication for every Host API route and stream, and the bind address plus `--trusted-host` govern which browser Host is accepted; `--public-url` changes only the advertised URL. Other compositions own their bind and route-authentication policy. The dist location is an assembly fact of the frontend plugin that claims the seat.
## The service
`WebServer` (`ctx.webServer`) listens immediately on activation; a listen failure (EADDRINUSE…) rejects initialization, and the boot process reports the failed fiber. The `tls` pair is read once before the listener binds, and an unreadable, empty, or mismatched pair rejects initialization the same way instead of falling back to HTTP. `register(route)` adds one named route and returns its disposer; a duplicate `(kind, path)` throws because route patterns are a composition-level contract and a collision is a misconfiguration. Gzip wraps eligible socket-backed responses inside the server, so handlers keep direct `ServerResponse` ownership, and a resource hands the service a `Response` to write back. Existing content encodings, `Cache-Control: no-transform`, ranges, SSE, ZIP, and the packaged `.gz` Worker image remain identity responses. `collectIndexInjections(peer)` gathers structured `IndexInjection` rows over one `webserver/index-inject` emit that passes the Peer the page speaks for, and `renderIndex(html, peer)` renders them into successful root and configured index responses before applying the raw `tapIndex(transform)` escape-hatch transforms in registration order; [dsh-client-modules](../../packages/client/modules) answers the event with the boot manifest rows. `port` reads the listening port, including the port assigned by the OS when `config.port` is 0, and `protocol` reports the scheme the listener serves (`http:` or `https:`).
A request whose handling throws (a malformed %-escape hitting `decodeURIComponent`, a client dropping mid-body) is logged as a warning and answered 400 — or the socket destroyed when headers are already out — never a process exit. Disposal pairs `close()` with `closeAllConnections()` because a handler may hold its response open (SSE) and such connections never end on their own; without the force-close, teardown would hang. A TLS socket counts as a connection only after its handshake, so the server tracks raw TLS sockets beside upgraded ones and destroys both at disposal. The package never prints: the URL line belongs to the shell. Per-package operational detail, including the dev-mode bundle watch pipeline, stays in the [README](../../packages/host/webserver/README.md).
<!-- BEGIN GENERATED cordis-surface (gen-cordis-catalog.ts) — do not edit between markers -->
<a id="cordis-surface"></a>
## Cordis API
Generated from source by `scripts/gen-cordis-catalog.ts` (verified fresh by `pnpm run verify-cordis-catalog` in doc-sync; regenerate with `pnpm run gen-cordis-catalog`) — the language sides differ only in locale-specific paired document paths. Signature blocks use a `ts cordis-catalog` fence and keep the original source JSDoc; dispatch modes are defined in the [primer](../cordis-primer.md#dispatch-modes), and the framework-inherited `ctx` API lives in [cordis-api/inherited.md](../cordis-api/inherited.md).
<a id="ctxconnection--hostconnectionhandle"></a>
### `ctx.connection` — `HostConnectionHandle`
Host `ctx.connection` members consumed by transport-independent adapters.
```ts cordis-catalog
/**
* Compose exact Fetch routes and the shared-channel RPC interceptor.
* @param channel - shared channel mounted by Connection.
* @returns Fetch handler for trusted, authenticated requests.
*/
createSharedFetchHandler(channel: '/api'): ConnectionFetchHandler
/**
* Apply Connection's Host/Origin checks and browser authentication to
* another Web route.
* @param request - request headers from the HTTP or upgrade request.
* @returns rejection status, or undefined when the route may accept the request.
*/
requestRejection(request: ConnectionTrustRequest): ConnectionRequestRejection
/**
* Admit an HTTP request against its signed browser or API credential and live Peer.
* @param request - request headers from the HTTP or upgrade request.
* @returns the credential's Peer, or the rejection status.
*/
admit(request: ConnectionTrustRequest): PeerAdmission
/**
* Create an independent in-process Peer, disposed with the Context that read `ctx.connection`.
* @param options - fixed principal and optional interaction participation (default `answer`).
* @returns a new Peer distinct from every other Peer.
*/
createPeer(options: { readonly principal: Principal; readonly interactions?: 'answer' | 'none' }): PeerScope
/**
* Claim a physical socket for an API Peer; browser sockets share their owner Peer.
* @param request - authenticated upgrade request headers.
* @returns the Peer and idempotent release, or 409 for duplicate API ownership.
* The carrier releases on socket termination or failed upgrade; release removes
* API admission synchronously without cancelling independent HTTP calls.
*/
admitConnection(request: ConnectionTrustRequest): ConnectionAdmission
/**
* Authenticate one frontend index request, owning a token redirect or 401.
* The cookie this mints is `Secure` when the mounted Web carrier serves TLS,
* which only that listener's protocol decides.
* @param request - root or configured-index HTTP request.
* @param response - response owned when the result is false.
* @returns true only when the frontend may serve index.html.
*/
authorizeIndex(request: ConnectionIndexRequest, response: ConnectionIndexResponse): boolean
/**
* Add the fresh process token to an ordinary Web application URL.
* @param baseUrl - clean application URL whose authority and mount are preserved.
* @returns tokenized URL for initial login; a mount proxy strips its prefix before {@link authorizeIndex}.
*/
authenticatedUrl(baseUrl: string): string
```
Types: [Principal](typert.md)
Source: [`packages/client/connection/src/rpc.ts`](../../packages/client/connection/src/rpc.ts)
<a id="ctxwebapp--webapp"></a>
### `ctx.webApp` — `WebApp`
The Web GUI this process serves, provided by the `web-runtime` row.
```ts cordis-catalog
/**
* The URL this GUI advertises, also used for `DSH_WEB_URL` and the startup URL line.
* @returns the configured `publicUrl`, otherwise the listener origin.
*/
getAdvertisedUrl(): string
```
Source: [`packages/bundle/web-app/src/index.ts`](../../packages/bundle/web-app/src/index.ts)
<a id="ctxwebserver--webserver"></a>
### `ctx.webServer` — `WebServer`
The browser HTTP carrier service. Activation loads any configured TLS material, then listens immediately; a material or listen failure rejects initialization, and the boot process reports the failed fiber. Route registration order does not affect requests because configured named routes must be distinct, and a request no named route matches is answered 404 until the fallback owner registers.
```ts cordis-catalog
/**
* The root URL local clients use to reach the bound listener: the protocol,
* the {@link bindUrlHost} form of the bind address, and the listening port.
* @returns the origin, for example `http://127.0.0.1:3080`.
* @throws before `listen` binds a port, or when the bind address has no URL form.
*/
getOrigin(): string
/**
* Register a named route. Duplicate (kind, path) throws — route patterns are
* a composition-level contract, so a collision is a misconfiguration.
* @param route - kind, path, and the owning handler or resource.
* @returns the disposer removing the route.
*/
register(route: WebRoute | WebResourceRoute): () => void
/**
* Register an exact-path HTTP upgrade route. Duplicate paths throw because
* one socket can have only one protocol owner.
* @param route - pathname and handler owning negotiation plus socket use.
* @returns the disposer removing the route.
*/
registerUpgrade(route: WebUpgradeRoute): () => void
/**
* Claim the fallback seat: the handler or resource answering every request
* no named route matches (the SPA dist server in the shipped Web
* composition). One owner only — a second registration throws, because two
* fallbacks cannot compose.
* @param fallback - a handler owning the full response lifecycle of
* unmatched requests, or `{ resource }` answering them.
* @returns the disposer releasing the seat.
*/
registerFallback(fallback: WebRoute['handler'] | Pick<WebResourceRoute, 'resource'>): () => void
/**
* Read the response an HTTP request for the same URL would receive, as the
* resource computes it for `peer`, without an HTTP exchange. The request
* path picks the responder by the HTTP match order — exact, longest prefix,
* then the fallback — and only a resource answers. This method checks no
* cookie, Host, or Origin: the caller admits the reader before reading.
* @param request - the read; its URL path selects the responder.
* @param peer - the Peer the read speaks for, passed to the resource.
* @returns the resource's response, without a body for HEAD as over HTTP;
* 405 with `Allow: GET, HEAD` for any other method, checked before matching,
* where over HTTP a handler route or an unmatched path answers such a
* method itself; 404 when the path matches a handler route, or reaches the
* fallback seat while it is unclaimed or holds a handler. A resource that
* throws rejects the read, where an HTTP request would receive 400.
*/
async readResource(request: Request, peer: PeerScope): Promise<Response>
/**
* Register a raw-HTML index transform, the escape hatch for markup no
* {@link IndexInjection} row expresses: {@link renderIndex} applies taps in
* registration order after rendering the structured rows.
* @param transform - pure html-to-html function.
* @returns the disposer removing the transform.
*/
tapIndex(transform: (html: string) => string): () => void
/**
* Run an index.html body through the registered taps in registration order
* — called by the fallback owner on every index response it renders.
* @param html - the raw index.html body.
* @returns the transformed body.
*/
applyIndexTaps(html: string): string
/**
* Gather the structured injection table for one Peer: one
* `webserver/index-inject` emit, every subscriber pushes its current rows.
* Fresh per call, so subscribers read live state (module graph, theme
* preference, the Peer's access) at emit time.
* @param peer - the Peer the page speaks for.
* @returns rows in subscriber activation order.
*/
collectIndexInjections(peer: PeerScope): IndexInjection[]
/**
* Render one index.html body for one Peer: the structured injection table
* first, then the raw `tapIndex` transforms over the result.
* @param html - the raw index.html body.
* @param peer - the Peer the page speaks for.
* @returns the transformed body.
*/
renderIndex(html: string, peer: PeerScope): string
```
Source: [`packages/host/webserver/src/index.ts`](../../packages/host/webserver/src/index.ts)
<a id="connection-events"></a>
### `connection/*` events
<a id="connectionrequest--waterfall"></a>
#### `connection/request` — waterfall
Admit or wrap an authenticated shared API request, including body transfer. Existing requests continue when a listener refuses subsequent requests.
```ts cordis-catalog
/**
* Admit or wrap an authenticated shared API request, including body transfer.
* Existing requests continue when a listener refuses subsequent requests.
* @param request - Authenticated incoming HTTP request.
* @param response - Response owned until the delegated bridge settles.
* @param next - Delegate to the next listener or the shared API bridge.
* @mode waterfall
*/
'connection/request'(request: IncomingMessage, response: ServerResponse, next: () => Promise<void>): Promise<void>
```
Source: [`packages/client/connection/src/index.ts`](../../packages/client/connection/src/index.ts)
<a id="webserver-events"></a>
### `webserver/*` events
<a id="webserverindex-inject--emit"></a>
#### `webserver/index-inject` — emit
Collect the structured index injection table for the page `peer` speaks for. Emitted on every index render and every worker boot-payload request; listeners push their current rows, so a row's data is read fresh at emit time.
```ts cordis-catalog
/**
* Collect the structured index injection table for the page `peer` speaks
* for. Emitted on every index render and every worker boot-payload
* request; listeners push their current rows, so a row's data is read
* fresh at emit time.
* @param table - Mutable row table; listeners append in activation order.
* @param peer - The Peer the rendered page speaks for.
* @mode emit
*/
'webserver/index-inject'(table: IndexInjection[], peer: PeerScope): void
```
Source: [`packages/host/webserver/src/index.ts`](../../packages/host/webserver/src/index.ts)
<!-- END GENERATED cordis-surface -->