Supercharge your AI agents with data from the web and beyond. Building the library for superintelligence. 🔥
TypeScript aiai-agentsai-crawlerai-scrapingai-searchcrawlerdata-extractionhtml-to-markdownllmmarkdownscraperscrapingweb-crawlerweb-dataweb-data-extractionweb-scraperweb-scrapingweb-searchwebscraping
fix(api): build exchange proxy paths from validated segments (#5331)
David Tippett910fd466aa55
diff --git a/apps/api/src/__tests__/routes/exchange-connectors.routes.test.ts b/apps/api/src/__tests__/routes/exchange-connectors.routes.test.tsindex 37dc5a19b..ca66dda56 100644--- a/apps/api/src/__tests__/routes/exchange-connectors.routes.test.ts+++ b/apps/api/src/__tests__/routes/exchange-connectors.routes.test.ts@@ -241,6 +241,34 @@ describe("Alexandria connector routes on the Exchange proxy", () => {);});+ it("re-encodes path segments before forwarding", async () => {+ keyResolvesAs({});+ await request(app)+ .patch("/exchange/team-skills/item%201")+ .set("authorization", `Bearer ${KEY}`)+ .send({ name: "Digest" });+ expect(mocks.upstreamFetch).toHaveBeenCalledTimes(1);+ expect(upstream().url).toBe(+ "https://exchange.test/v1/team-skills/item%201",+ );+ });++ it.each([+ "/exchange/supply/..%2Fretrieve",+ "/exchange/supply/%2e%2e%2Fretrieve",+ "/exchange/supply/x%5C..%5Cretrieve",+ "/exchange/supply/a//b",+ "/exchange/supply/%E0%A4%A",+ ])("refuses %s before reaching Exchange", async path => {+ keyResolvesAs({ exchangeRetrieve: true });+ const response = await request(app)+ .post(path)+ .set("authorization", `Bearer ${KEY}`)+ .send({});+ expect(response.status).toBe(400);+ expect(mocks.upstreamFetch).not.toHaveBeenCalled();+ });+it("never forwards a caller's own x-exchange-connectors header", async () => {keyResolvesAs({});// A route open without the flag: the header must still not reach Exchange.diff --git a/apps/api/src/routes/exchange.ts b/apps/api/src/routes/exchange.tsindex 8615a3268..7352b41c4 100644--- a/apps/api/src/routes/exchange.ts+++ b/apps/api/src/routes/exchange.ts@@ -25,6 +25,23 @@ const INGEST_TIMEOUT_MS = 50_000;const FORWARDED_REQUEST_HEADERS = ["accept", "x-request-id"];const FORWARDED_RESPONSE_HEADERS = ["content-type", "x-request-id"];+const FORWARDED_PATH_ROOTS = new Set([+ "analytics",+ "applications",+ "claims",+ "connectors",+ "discover",+ "enrichment",+ "ingest",+ "platform",+ "provider-terms",+ "publisher",+ "rates",+ "records",+ "skills",+ "supply",+ "team-skills",+]);function dispatcherFor(timeout: number) {return new Agent({@@ -43,6 +60,32 @@ function upstreamBase(): string | null {return config.FIRE_EXCHANGE_URL.replace(/\/+$/, "");}+// Rebuilds the upstream path from validated, re-encoded segments so it always+// names the route that authorized the request.+function upstreamPath(req: Request): string | null {+ const segments = req.path.split("/").slice(1);+ if (segments.length > 1 && segments.at(-1) === "") segments.pop();++ const decoded: string[] = [];+ for (const segment of segments) {+ let value: string;+ try {+ value = decodeURIComponent(segment);+ } catch {+ return null;+ }+ if (!value || value === "." || value === ".." || /[/\\]/.test(value)) {+ return null;+ }+ decoded.push(value);+ }+ if (!FORWARDED_PATH_ROOTS.has(decoded[0])) return null;++ const queryStart = req.originalUrl.indexOf("?");+ const query = queryStart === -1 ? "" : req.originalUrl.slice(queryStart);+ return "/v1/" + decoded.map(encodeURIComponent).join("/") + query;+}+function exchangeProxy(timeout: number,options: {@@ -67,6 +110,11 @@ function exchangeProxy(return exchangeError(res, 503, "This endpoint is not available.");}+ const path = upstreamPath(req);+ if (path === null) {+ return exchangeError(res, 400, "Invalid request path.");+ }+if (requiresRetrieveFlag && !authedReq.acuc?.flags?.exchangeRetrieve) {return exchangeError(res,@@ -77,7 +125,6 @@ function exchangeProxy(const hasBody = req.method !== "GET";const billingCountry = await billingCountryForTeam(authedReq.auth.team_id);- const path = req.originalUrl.replace(/^\/exchange/, "/v1");try {const upstream = await fetch(base + path, {