92eba07da7
Closes #59. Part of spec #55, and the ticket that fixes the reported bug #47. Architecture: `docs/adr/0007-backend-hosts-cover-bytes.md`. Does not close #47 or #55. ## What changed A Reader bookmarks a Series nobody holds yet — the exact case in #47 — and within seconds the list shows its artwork instead of a broken image. The first Bookmark to create a Series fires `Store.OnSeriesCreated` after commit, and the new `latest.Acquirer` turns that into **one** series-page fetch that yields both the Latest Chapter and the cover URL. The bytes go through the gated cover fetcher from #57 and are stored content-addressed through #56, so the wire carries an absolute URL on this deployment's own origin — never a third-party address, and never one that 404s. ### Store - Migration `0009_series_cover_address.sql` adds `series.cover_address`. The two facts are now split: `series.cover` is the third-party source address the bytes came from (the acquisition path's dedupe key), `series.cover_address` is the SHA-256 they are stored under. An empty `cover_address` is precisely what "no Cover yet" means, which is the distinction both the API and the UI depend on. - `SetSeriesCover` writes the address only after the bytes are on disk, so the wire can never name an object that is not there. - `CoverWireURL` builds `PUBLIC_BASE_URL + /covers/<sha256>` for every scanned row, and returns `""` for a blank address. - The cover columns are gone from `Upsert`'s `INSERT` and its `DO UPDATE`. A client-supplied cover cannot reach the shared Series row on any path, not just the creation path. - `Open` now rejects a base URL that is not an absolute `http(s)` origin: `PUBLIC_BASE_URL=bookmarks.example.com` would otherwise start cleanly and emit addresses no browser can load. ### Acquisition - `internal/latest/acquire.go`: one fetch, gated by the poller's own `fetchableSeriesURL` (a `series_url` arrives in a client-supplied PUT body, so without the gate a token-holder chooses what the server fetches from its own network position). - Asynchronous and log-and-drop. The Bookmark, its progress and its Latest Chapter are already committed; a Site that is down or a cover that cannot be produced disturbs none of them. - Bounded by a two-slot semaphore. A bulk sync creating N Series would otherwise fire N simultaneous requests from one IP — the traffic shape the poller's stagger exists to avoid. - Cancelled at shutdown (shares the poller's context) and stamps `latest_checked_at`, so the poller does not refetch the same page a tick later. - Browser-backed Sites (kagane, novelfull) are deliberately skipped: their pages only yield a Cloudflare challenge to the TLS client, so the request would be spent for nothing. They arrive in #62. ### Wire and route - `GET /covers/{address}` serves the bytes publicly and uncredentialed with `Cache-Control: public, max-age=604800, immutable`. The address is gated by a `^[0-9a-f]{64}$` pattern and cross-checked against a pure function of itself before any filesystem read, so no request shaped like a traversal reaches disk. - `PUT /bookmarks/{key}` still accepts a `cover` field and discards it, permanently. Rejecting it would break every installed userscript the moment this deploys, and ADR-0004's compatibility argument depends on those scripts continuing to work. The decode site says so in place of a TODO nobody intends to keep. - `store.CoverContentType` canonicalises comix's non-standard `image/jpg` to `image/jpeg`, so one image cannot land under two spellings. This one was found by the live smoke test, not by reading. ### Config `PUBLIC_BASE_URL` is new and required (cover URLs must go out absolute — the userscript renders them on third-party origins, where a relative path resolves against the Site). Documented in `.env.example`, `docker-compose.yml` (`:?` so compose fails too), `DEPLOY.md` and `backend/AGENTS.md`. ## Acceptance criteria All twelve of #59's criteria are met; the checklist on the issue is ticked with the evidence. ## Verification - `go test ./...` green (Docker-backed Postgres suite). - Live smoke against a real backend + Postgres: bookmarking `comix:n8we-dungeons-and-crayons` produced `"cover": "http://127.0.0.1:8099/covers/8ce74d80…"` and `"latest_chapter": "Chapter 81"` within seconds of the PUT; `curl` on that address returned `200`, `Content-Type: image/jpeg`, `Cache-Control: public, max-age=604800, immutable`, and a 280x420 JPEG. That run is what surfaced the `image/jpg` content type. - Mutation-checked the asynchrony test: removing the `go` from `Acquire` turns `TestAcquireDoesNotBlockTheWrite` red. ## Reviewed Both axes of `/code-review` were run against this diff before commit. Their findings that were actionable here are folded in: the concurrency bound, the shutdown tie, the `PUBLIC_BASE_URL` validation, the missing `latest_checked_at` stamp, and a test that could not fail. ## Known sequencing A kagane/novelfull Series created between this deploy and #62 has no cover source at all: the acquisition skips those Sites and `Upsert` no longer persists the userscript-scraped address. This is #59's stated boundary rather than a defect, but it is a user-visible gap on two Sites and should order #62 accordingly. Reviewed-on: #68 Co-authored-by: Sulthan Zaki <sultankiki05@gmail.com> Co-committed-by: Sulthan Zaki <sultankiki05@gmail.com>
13 KiB
13 KiB
Guidance for OpenCode (and Claude Code) working under backend/. See root AGENTS.md for the project-wide architecture diagram, hard constraints, and design system.
- Backend (
backend/): stdlibnet/http(handful routes, no framework) + Postgres overjackc/pgx/v5(pure Go,CGO_ENABLED=0-> static binary -> distroless/scratch image). Reverse proxy terminates TLS; Go service listens plain:8080. Single binary, split into packages underbackend/internal/:store(Bookmark type, Postgres persistence, migration runner),latest(background poller, site parsers, TLS fetcher),session(cookie signing, login rate limiter),httpmw(Auth/Gzip/CORS middleware),api(JSON bookmark handlers),userscript(userscript-serving handler),web(browser UI handler +templates/+static/,go:embed-ed).backend/main.gois the composition root — the only place that wires packages together intonewRouter. Root-level*_test.gohold integration tests that exercise the full router; unit tests for a package live beside it underinternal/. - Schema is migration-owned.
internal/store/migrations/*.sqlisgo:embed-ed and applied on every start bystore.migrate: one numbered file per change, one transaction each, versions recorded inschema_migrations. Files are append-only — editing an applied one changes nothing on a database that already ran it. No column probing, no data-fixup migrations: both were SQLite-era machinery and are gone. - Tests need Docker.
internal/pgteststarts onepostgres:17-alpinecontainer per test binary (TestMain->pgtest.Main) and hands each test its own database (pgtest.URL(t)). A package whose tests touch the store must have thatTestMain. - Reader-owned store, four tables.
readersis keyed by Discord user ID and carries the SHA-256 of the Reader's userscript credential plus atoken_epoch(issue #24). Credentials are derived, never stored:token.Token(TOKEN_KEY, discord_id, epoch)(HMAC,internal/token), and only its SHA-256 sits inreaders.token_sha256, so install URLs can be rebuilt after any restart while a database leak yields nothing but hashes. The seed creates the owner row at startup; its epoch-0 hash is refreshed on every start only while the row has never been rotated, so a restart can never resurrect a rotated-away credential. Every other row is created by that Reader's own first login (Store.EnsureReader, idempotent ondiscord_id, and it never rewrites an existing row's hash). Rotation isStore.RotateToken(epoch bump + hash rewrite in one transaction), driven by the web UI.serieskeyed(site, series_id)(asura|demonic|comix|kagane|novelfull|lightnovelworld) owns the shared facts — title, cover, canonical URL,kind(manga|novel), Latest Chapter,latest_checked_at— andbookmarksholds only what differs between readers: progress, favourite, lifecycle bucket,updated_at. A bookmark is keyed(reader_id, site, series_id)— no surrogate id; the wirekeyis derived assite:series_idon read — and every store read/write is scoped to the reader it names. Auth resolves the acting Reader from the presented credential (httpmw.Auth) and nothing else — there is no unauthenticated-by-Reader route and no global token; the reader id travels in the request context. Sync last-write-wins; the wire format stays flat (ADR-0004).Store.Upsertdecomposes one flat body across two tables and enforces the ownership rule: clienttitle/series_url/coverare written only when the series row is new (ADR-0003). - Endpoints:
GET /bookmarks,PUT /bookmarks/{key}(upsert; seeupdated_atrule below),DELETE /bookmarks/{key},GET /healthz(no auth). - Web UI: same binary serve the browser UI on a second
hostname —
GET /(list, or login page when no session),GET /auth/discord+GET /auth/discord/callback(Discord OAuth, ADR-0002),POST /logout,GET /static/*, htmx fragment endpoints under/ui/*. Templates + assetsgo:embed-ed underbackend/internal/web/, sobackend/Dockerfilemust copy the wholeinternal/tree, not just*.go. Sessions are rows in thesessionstable: the cookie carries only an opaque id, looked up (and expiry- checked) on every request, and deleting the row revokes the session. Guild membership is registration (issue #27):discordCallbackgates on membership (andDISCORD_REQUIRED_ROLEwhen set) and then callsStore.EnsureReader, so a refusal creates nothing and a returning Reader reuses their row. The owner is the only Reader with administrative reach:POST /readers/{id}/revoke(404 for anyone else) drops that Reader's sessions, and thereaderspanel renders only on the owner's page. A Reader with no bookmarks at all seeslistView.Fresh, whose empty state offers both install links instead of describing a filter. UI mutations read-modify-write throughStore.Get+Store.Upsertsoupdated_atrule stays one place. Seedocs/superpowers/specs/2026-07-25-web-ui-design.md. Design-tool caveat: templates link/static/style.cssroot-absolutely (correct — served from/), but impeccable detector resolves stylesheet href withpath.resolve(fileDir, href), drops directory on leading/and silently skip file. Relative href don't help either: template's directory isn't its served path. Sodetect.mjs backend/internal/web/templatesreports false clean — always passbackend/internal/web/statictoo. One finding there,overused-fonton "Instrument Serif", deliberate identity choice, not debt. - Every action that moves series out of list is confirm-gated.
Archive, finish, remove each open own
.confirm-rowdisclosure (toggleConfirmRow(key, kind)infilter.js,kind∈archive|finish|remove); restore fire instantly since it's the reversal. Remove's row wear ember wash, two reversible ones wear.calmgrey.--emberstay reserved for new-chapter signal: busy bar and inline error use--mute. - Latest-chapter poller: ticker goroutine in same binary re-check
each bookmarked series' newest published chapter from backend's own
network access, so
latest_chapterstay fresh when user not browsing. Second, parallel signal — userscript keep ownmaybeCaptureLatestOnSeriesPage/backgroundRefreshLatestlogic unchanged. Two independent clocks: per-series cooldown (series.latest_checked_at, enforced byStore.DueForLatestCheck's WHERE clause) and wake interval. The poller walks Series, not Bookmarks — a series referenced by several bookmarks is fetched once per cycle, and the due queue ordersreader_count DESC, latest_checked_at ASC(ADR-0003). Series row stamped before fetch so broken series wait out full cooldown instead of retrying every tick; found chapter written straight to the series row viaStore.SetLatestChapter, so a bookmark'supdated_at— and the list order — is never touched. Fetches usebogdanfinn/tls-clientwith Chrome profile as defence in depth against fingerprint-based blocking; any failure log and skip. kagane and novelfull sit behind Cloudflare JavaScript challenges the TLS client can't clear, so they are browser-only: fetched over CDP viaBROWSER_WS_URL, and simply not polled when that's unset. Seedocs/superpowers/specs/2026-07-26-server-latest-chapter-polling-design.md. The poller's series write is a single-column UPDATE (Store.SetLatestChapter), not a read-modify-write of the whole bookmark: it cannot revert read progress or moveupdated_at, so the old stale-re-read race is gone with the Get+Upsert flow. - Covers are acquired at creation, then served from our own origin
(ADR-0007): the first Bookmark of a Series fires
Store.OnSeriesCreated, whichlatest.Acquirerturns into one series-page fetch yielding both the Latest Chapter and the cover URL; the bytes then go throughlatest.CoverBytesFetcherintoStore.SetSeriesCover. It runs in a goroutine — the Reader's PUT must neither block on a Site nor fail with one — and every failure is logged and dropped, leaving the Bookmark intact. The wire'scoveris the absolutePUBLIC_BASE_URL + /covers/{sha256}once bytes exist and""before, never an address that 404s.GET /covers/{addr}is public and uncredentialed: the userscript renders it on a Site's origin, where no cookie or token of ours travels. A client-sentcoveris decoded and discarded, permanently (ADR-0004 compatibility). updated_atdrives list order, so moves only on real reading progress: server apply its timestamp when row new orlast_chapter_numchanges, else keep stored value — favouriting series or recording newly published chapter must not reorder list.PUTtherefore returns row as stored, clients must adopt that response rather than own payload. Seeplans/2026-07-25-bookmark-list-favorites-design.md§4.- Lifecycle buckets:
statuson each bookmark isreading|archived|finished, orthogonal tofavorite. Archived and finished appear only in own tab — not in All, Updated, Favourites, or recent strip. Poller keeps checking archived series and skip finished ones.finishedsettable only from web UI;PUT /bookmarks/{key}reject it with 400. Empty incoming status means "keep stored one" — resolved on theVALUESside ofStore.Upsert, not conflict clause, sinceexcluded.*is post-evaluation row and default applied there would wipe bucket on every PUT from client that predates column. Seedocs/superpowers/specs/2026-07-27-status-buckets-design.md. - Config via env:
TOKEN_KEY(derives every Reader's userscript credential; required),OWNER_DISCORD_ID(seeds the owner Reader — the administrator and the owner of every pre-registration bookmark; required),ALLOWED_ORIGINS(comma list),DATABASE_URL(Postgres connection URL, required — no default),COVER_DIR(required filesystem volume for content-addressed Cover bytes),PUBLIC_BASE_URL(required origin this deployment answers on, trailing slash trimmed; every Cover URL on the wire is built from it, absolute because the userscript renders on a Site's origin — ADR-0007),PORT(default8080),DISCORD_CLIENT_ID/_CLIENT_SECRET/_GUILD_ID/_REDIRECT_URI(required; Discord OAuth for the browser UI),DISCORD_REQUIRED_ROLE(optional role gate, empty by default),DISCORD_API_BASE(defaulthttps://discord.com/api/v10),LATEST_CHAPTER_POLL_ENABLED/_COOLDOWN/_BROWSER_COOLDOWN/_INTERVAL/_BATCH/_STAGGER(background latest-chapter poller; defaults on,1hplain-TLS cooldown,6hbrowser cooldown,10m/14/20s; both cooldowns have a15mfloor).USERSCRIPT_PATHandNOVEL_USERSCRIPT_PATH(files served at/u/{token}/manga-bookmark.user.jsand/u/{token}/novel-bookmark.user.js, defaults/userscript/manga-bookmark.user.jsand/userscript/novel-bookmark.user.js, both supplied by bindmount; the__API_TOKEN__placeholder inside them is substituted with the requesting Reader's credential at serve time).BROWSER_WS_URL(CDP endpoint of the browser, which runs on a separate machine and is reached over the tailnet — ADR-0006,chrome/docker-compose.yml. Used by the poller for kagane and novelfull and by the web UI's kagane cover proxy; unset — the default — disables browser polling and serves 404 for covers not already stored, leaving those sites to the userscript alone. Must be a tailnet IP, never a hostname: Chrome's DevTools handler 500s/json/versionfor any Host that isn't an IP orlocalhost). - kagane covers are proxied, not hot-linked: kagane serves cover images
behind the same challenge as its pages and with
cross-origin-resource-policy: same-origin, so no<img>on the web UI's origin can load one — not even from a browser holding the clearance cookieog:imageto/img/kagane/{id}.internal/web/cover.goreads the persistentcoverstable first, then fetches a miss throughlatest.BrowserFetcher.Image. The templates render.CoverURL, never.Cover. The id is matched against a UUID regex before it reaches the browser: the stored value is client-supplied, so an unchecked one is an SSRF primitive pointed at the deployment's own network. - Web UI also owns: session-gated
GET /install/{manga,novel}-bookmark.user.js(renders the bindmounted script with the acting Reader's derived credential substituted in — the credential never appears in page markup, the address bar, or a redirect;?download=1addsContent-Disposition: attachmentfor mobile Violentmonkey, which ignores a.user.jsnavigation) andPOST /rotate-token(atomic epoch bump + hash rewrite; invalidates every installed copy, so the panel warns to reinstall on all devices). Owner-onlyPOST /readers/{id}/revoke(drops one Reader's session rows and re-renders thereaderspanel; 404 for any non-owner) is the only route that reaches across Readers.