524d9978f1
* fix(migration): resumable underlag import without inline extraction, same-origin MCP storage URLs The Fortnox underlag import ran every file's AI extraction inline inside one request and hit the hosted 300 s function limit after ~17 of 113 files (twice on 2026-08-21); the UI showed the generic "underlagen kunde inte importeras" although the files it did reach were linked. The import now works in time-budgeted slices with a stable cursor (the UI loops until the server reports the end and shows "x av y") and opts out of extraction (extractionOwner 'none', stamped skipped:opted_out): every file is linked to its posted verifikat on arrival, so the booking is already known. MCP signed Storage URLs (upload_url, signed_url, download_url) are served through a same-origin proxy, /api/storage/[...path], because Claude Desktop's sandbox only reaches the MCP host and blocked the PUT to <project>.supabase.co. The signed token stays the only credential; the proxy forwards only signed documents-bucket paths to our own Storage host and is a no-op rewrite when NEXT_PUBLIC_APP_URL is unset. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013YoZ8iboyTj221axW6Gdtm * fix(mcp): keep the storage-proxy note out of the size-capped tool descriptions The per-tool 280-char cap and the tools/list payload ceiling both tripped on the two sentences added to gnubok_create_document_upload and gnubok_get_document_content; the why now lives in a code comment. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013YoZ8iboyTj221axW6Gdtm * fix(review): id cursor, stall = error, capped upload body, encoded dot segments Review follow-ups on #1783: - the import cursor is the last handled provider attachment id, not an index, so a file Fortnox adds or removes mid-sweep shifts nothing - a partial answer whose cursor does not advance (or the round guard) is reported as ARCIM_DOCUMENT_IMPORT_STALLED instead of "complete"; the slices already landed stay reported and the retry button resumes - the storage proxy reads the PUT body as a capped stream instead of buffering an unbounded payload before measuring it - object paths are rejected when any segment decodes to "." or ".." (or holds a separator), and the URL fetch() would actually request is re-checked against the allowlist after normalisation - download_url description no longer claims a direct Storage URL Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013YoZ8iboyTj221axW6Gdtm --------- Co-authored-by: Jakob Wennberg <311770904+jakobwennberg-oss@users.noreply.github.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
178 lines
6.1 KiB
TypeScript
178 lines
6.1 KiB
TypeScript
/**
|
|
* Same-origin proxy for Supabase Storage signed URLs.
|
|
*
|
|
* Agent sandboxes (Claude Desktop's code execution, some MCP clients) only
|
|
* allow network egress to the host the MCP server lives on. Our signed
|
|
* Storage URLs point at <project>.supabase.co, so a model-free upload or a
|
|
* document download from such a sandbox was refused before it left the box
|
|
* (user report 2026-08-21). These helpers rewrite a signed URL onto this
|
|
* app's own origin (`/api/storage/...`) and resolve it back to the upstream
|
|
* Storage URL inside the proxy route.
|
|
*
|
|
* The signed token travels unchanged and is the only credential on both
|
|
* sides: the proxy grants nothing the public Storage host did not already
|
|
* grant, and it refuses every path that is not a signed object path on the
|
|
* documents bucket. Without NEXT_PUBLIC_APP_URL (a self-host that never set
|
|
* it) the rewrite is a no-op rather than a broken localhost link.
|
|
*/
|
|
|
|
export const STORAGE_PROXY_ROUTE = '/api/storage'
|
|
|
|
/** Upstream prefix under the Storage API that every signed object URL shares. */
|
|
const UPSTREAM_OBJECT_PREFIX = '/storage/v1/object/'
|
|
|
|
/**
|
|
* Only token-authenticated object paths on the documents bucket: signed
|
|
* downloads (`sign/documents/...`) and signed uploads
|
|
* (`upload/sign/documents/...`). Anything else (public objects, bucket
|
|
* admin, other buckets) is not this proxy's business.
|
|
*/
|
|
const ALLOWED_OBJECT_PATH_RE = /^(sign|upload\/sign)\/documents\/[^/]+/
|
|
|
|
export type StorageProxyResolution =
|
|
| { ok: true; url: string }
|
|
| { ok: false; reason: 'unsupported_path' | 'missing_token' | 'storage_unconfigured' }
|
|
|
|
/**
|
|
* True when every segment of the (still percent-encoded) object path is a
|
|
* plain name: no empty segment, no `.`/`..` in raw or percent-encoded form
|
|
* (`%2e%2e`, `.%2e`, `%2e.`: the WHATWG URL parser normalises those away,
|
|
* which would let `sign/documents/%2e%2e/other/x` leave the documents
|
|
* bucket), no backslash (a path separator for https URLs), no malformed
|
|
* escapes.
|
|
*/
|
|
function hasOnlyPlainSegments(objectPath: string): boolean {
|
|
for (const segment of objectPath.split('/')) {
|
|
if (segment === '') return false
|
|
let decoded: string
|
|
try {
|
|
decoded = decodeURIComponent(segment)
|
|
} catch {
|
|
return false
|
|
}
|
|
if (decoded === '.' || decoded === '..') return false
|
|
if (decoded.includes('\\') || decoded.includes('/')) return false
|
|
}
|
|
return true
|
|
}
|
|
|
|
function trimTrailingSlash(value: string): string {
|
|
return value.replace(/\/+$/, '')
|
|
}
|
|
|
|
function upstreamStorageOrigin(): string | null {
|
|
const raw = process.env.NEXT_PUBLIC_SUPABASE_URL?.trim()
|
|
if (!raw) return null
|
|
try {
|
|
return new URL(raw).origin
|
|
} catch {
|
|
return null
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Rewrite a Supabase signed Storage URL onto this app's origin. Returns the
|
|
* input unchanged when it is not a signed documents-bucket URL on our Storage
|
|
* host, or when the app has no public URL configured.
|
|
*/
|
|
export function toSameOriginStorageUrl(
|
|
signedUrl: string,
|
|
appBaseUrl: string | undefined = process.env.NEXT_PUBLIC_APP_URL,
|
|
): string {
|
|
const base = appBaseUrl?.trim()
|
|
const upstream = upstreamStorageOrigin()
|
|
if (!base || !upstream) return signedUrl
|
|
|
|
let url: URL
|
|
try {
|
|
url = new URL(signedUrl)
|
|
} catch {
|
|
return signedUrl
|
|
}
|
|
if (url.origin !== upstream) return signedUrl
|
|
if (!url.pathname.startsWith(UPSTREAM_OBJECT_PREFIX)) return signedUrl
|
|
|
|
// Keep the pathname exactly as Storage encoded it: object keys may hold
|
|
// spaces and non-ASCII, and the token was signed over the real key.
|
|
const objectPath = url.pathname.slice(UPSTREAM_OBJECT_PREFIX.length)
|
|
if (!ALLOWED_OBJECT_PATH_RE.test(objectPath)) return signedUrl
|
|
if (!url.searchParams.has('token')) return signedUrl
|
|
|
|
return `${trimTrailingSlash(base)}${STORAGE_PROXY_ROUTE}/${objectPath}${url.search}`
|
|
}
|
|
|
|
/**
|
|
* Resolve the proxied path (everything after `/api/storage/`, still
|
|
* percent-encoded as received) plus its query back to the upstream Storage
|
|
* URL. Fails closed on anything outside the signed documents-bucket paths.
|
|
*/
|
|
export function resolveUpstreamStorageUrl(
|
|
objectPath: string,
|
|
search: URLSearchParams,
|
|
): StorageProxyResolution {
|
|
const upstream = upstreamStorageOrigin()
|
|
if (!upstream) return { ok: false, reason: 'storage_unconfigured' }
|
|
if (!ALLOWED_OBJECT_PATH_RE.test(objectPath) || !hasOnlyPlainSegments(objectPath)) {
|
|
return { ok: false, reason: 'unsupported_path' }
|
|
}
|
|
if (!search.get('token')) return { ok: false, reason: 'missing_token' }
|
|
|
|
const query = search.toString()
|
|
const url = `${upstream}${UPSTREAM_OBJECT_PREFIX}${objectPath}${query ? `?${query}` : ''}`
|
|
// Belt and braces: whatever fetch() will actually request, after URL
|
|
// normalisation, must still sit inside the allowlist.
|
|
let normalisedPath: string
|
|
try {
|
|
normalisedPath = new URL(url).pathname
|
|
} catch {
|
|
return { ok: false, reason: 'unsupported_path' }
|
|
}
|
|
if (
|
|
!normalisedPath.startsWith(UPSTREAM_OBJECT_PREFIX) ||
|
|
!ALLOWED_OBJECT_PATH_RE.test(normalisedPath.slice(UPSTREAM_OBJECT_PREFIX.length))
|
|
) {
|
|
return { ok: false, reason: 'unsupported_path' }
|
|
}
|
|
|
|
return { ok: true, url }
|
|
}
|
|
|
|
/**
|
|
* Read a request body into memory, aborting as soon as it exceeds `maxBytes`
|
|
* instead of buffering an unbounded payload first and measuring afterwards
|
|
* (the proxy is unauthenticated until Storage checks the token, so a
|
|
* self-host without a platform body limit must not be forced to hold an
|
|
* arbitrary upload in RAM). Returns null when the cap is exceeded.
|
|
*/
|
|
export async function readBodyWithCap(
|
|
body: ReadableStream<Uint8Array> | null,
|
|
maxBytes: number,
|
|
): Promise<ArrayBuffer | null> {
|
|
if (!body) return new ArrayBuffer(0)
|
|
const reader = body.getReader()
|
|
const chunks: Uint8Array[] = []
|
|
let total = 0
|
|
try {
|
|
for (;;) {
|
|
const { done, value } = await reader.read()
|
|
if (done) break
|
|
if (!value) continue
|
|
total += value.byteLength
|
|
if (total > maxBytes) {
|
|
await reader.cancel()
|
|
return null
|
|
}
|
|
chunks.push(value)
|
|
}
|
|
} finally {
|
|
reader.releaseLock()
|
|
}
|
|
const out = new Uint8Array(total)
|
|
let offset = 0
|
|
for (const chunk of chunks) {
|
|
out.set(chunk, offset)
|
|
offset += chunk.byteLength
|
|
}
|
|
return out.buffer
|
|
}
|