fix: preserve Confluence context paths (#5415)

* fix: preserve confluence context paths

* lint and minor changes

---------

Co-authored-by: Timothy Carambat <rambat1010@gmail.com>

collector/__tests__/utils/extensions/Confluence/ConfluenceLoader.test.js

@@ -0,0 +1,125 @@
+/* eslint-env jest, node */
+process.env.STORAGE_DIR = "test-storage";
+
+const { resolveConfluenceBaseUrl } = require("../../../../utils/extensions/Confluence");
+const {
+  ConfluencePagesLoader,
+} = require("../../../../utils/extensions/Confluence/ConfluenceLoader");
+
+describe("resolveConfluenceBaseUrl", () => {
+  test("cloud: strips path and returns origin only", () => {
+    expect(
+      resolveConfluenceBaseUrl("https://example.atlassian.net/wiki/spaces/SP", true)
+    ).toBe("https://example.atlassian.net");
+  });
+
+  test("self-hosted: preserves context path, strips trailing slash", () => {
+    expect(
+      resolveConfluenceBaseUrl("https://my.domain.com/confluence/", false)
+    ).toBe("https://my.domain.com/confluence");
+  });
+
+  test("self-hosted: returns origin when no context path", () => {
+    expect(
+      resolveConfluenceBaseUrl("https://my.domain.com/", false)
+    ).toBe("https://my.domain.com");
+  });
+});
+
+describe("ConfluencePagesLoader", () => {
+  afterEach(() => {
+    jest.restoreAllMocks();
+  });
+
+  describe("cloud mode", () => {
+    test("API requests include /wiki prefix", async () => {
+      const fetchMock = jest.spyOn(global, "fetch").mockResolvedValue({
+        ok: true,
+        json: jest.fn().mockResolvedValue({ size: 0, results: [] }),
+      });
+      const loader = new ConfluencePagesLoader({
+        baseUrl: resolveConfluenceBaseUrl("https://example.atlassian.net/wiki/spaces/SP", true),
+        spaceKey: "SP",
+        username: "user",
+        accessToken: "token",
+        cloud: true,
+      });
+
+      await loader.fetchAllPagesInSpace();
+
+      expect(fetchMock).toHaveBeenCalledWith(
+        "https://example.atlassian.net/wiki/rest/api/content?spaceKey=SP&limit=25&start=0&expand=body.storage,version",
+        expect.any(Object)
+      );
+    });
+
+    test("page URLs include /wiki prefix", () => {
+      const loader = new ConfluencePagesLoader({
+        baseUrl: resolveConfluenceBaseUrl("https://example.atlassian.net/wiki", true),
+        spaceKey: "SP",
+        username: "user",
+        accessToken: "token",
+        cloud: true,
+      });
+
+      const document = loader.createDocumentFromPage({
+        id: "123",
+        status: "current",
+        title: "Cloud page",
+        type: "page",
+        body: { storage: { value: "<p>Hello</p>" } },
+        version: { number: 1, by: { displayName: "User" }, when: "2026-01-01T00:00:00.000Z" },
+      });
+
+      expect(document.metadata.url).toBe(
+        "https://example.atlassian.net/wiki/spaces/SP/pages/123"
+      );
+    });
+  });
+
+  describe("self-hosted mode", () => {
+    test("API requests use context path without /wiki", async () => {
+      const fetchMock = jest.spyOn(global, "fetch").mockResolvedValue({
+        ok: true,
+        json: jest.fn().mockResolvedValue({ size: 0, results: [] }),
+      });
+      const loader = new ConfluencePagesLoader({
+        baseUrl: resolveConfluenceBaseUrl("https://my.domain.com/confluence/", false),
+        spaceKey: "SP",
+        username: "user",
+        accessToken: "token",
+        cloud: false,
+      });
+
+      await loader.fetchAllPagesInSpace();
+
+      expect(fetchMock).toHaveBeenCalledWith(
+        "https://my.domain.com/confluence/rest/api/content?spaceKey=SP&limit=25&start=0&expand=body.storage,version",
+        expect.any(Object)
+      );
+    });
+
+    test("page URLs use context path without /wiki", () => {
+      const loader = new ConfluencePagesLoader({
+        baseUrl: resolveConfluenceBaseUrl("https://my.domain.com/confluence/", false),
+        spaceKey: "SP",
+        username: "user",
+        accessToken: "token",
+        cloud: false,
+      });
+
+      const document = loader.createDocumentFromPage({
+        id: "123",
+        status: "current",
+        title: "Self-hosted page",
+        type: "page",
+        body: { storage: { value: "<p>Hello</p>" } },
+        version: { number: 1, by: { displayName: "User" }, when: "2026-01-01T00:00:00.000Z" },
+      });
+
+      expect(document.metadata.url).toBe(
+        "https://my.domain.com/confluence/spaces/SP/pages/123"
+      );
+    });
+  });
+});

collector/utils/extensions/Confluence/index.js

@@ -46,10 +46,11 @@ async function loadConfluence(
     };
   }
 
-  const { origin, hostname } = new URL(baseUrl);
+  const normalizedBaseUrl = resolveConfluenceBaseUrl(baseUrl, cloud);
-  console.log(`-- Working Confluence ${origin} --`);
+  const { hostname } = new URL(normalizedBaseUrl);
+  console.log(`-- Working Confluence ${normalizedBaseUrl} --`);
   const loader = new ConfluencePagesLoader({
-    baseUrl: origin, // Use the origin to avoid issues with subdomains, ports, protocols, etc.
+    baseUrl: normalizedBaseUrl,
     spaceKey,
     username,
     accessToken,
@@ -98,13 +99,13 @@ async function loadConfluence(
       id: v4(),
       url: doc.metadata.url + ".page",
       title: doc.metadata.title || doc.metadata.source,
-      docAuthor: origin,
+      docAuthor: normalizedBaseUrl,
       description: doc.metadata.title,
-      docSource: `${origin} Confluence`,
+      docSource: `${normalizedBaseUrl} Confluence`,
       chunkSource: generateChunkSource(
         {
           doc,
-          baseUrl: origin,
+          baseUrl: normalizedBaseUrl,
           spaceKey,
           accessToken,
           username,
@@ -182,8 +183,9 @@ async function fetchConfluencePage({
   }
 
   console.log(`-- Working Confluence Page ${pageUrl} --`);
+  const normalizedBaseUrl = resolveConfluenceBaseUrl(baseUrl, cloud);
   const loader = new ConfluencePagesLoader({
-    baseUrl, // Should be the origin of the baseUrl
+    baseUrl: normalizedBaseUrl,
     spaceKey,
     username,
     accessToken,
@@ -243,6 +245,21 @@ function validBaseUrl(baseUrl) {
   }
 }
 
+/**
+ * Resolves the Confluence base URL, preserving context paths for self-hosted deployments.
+ * @param {string} baseUrl
+ * @param {boolean} cloud
+ * @returns {string}
+ */
+function resolveConfluenceBaseUrl(baseUrl, cloud = true) {
+  const url = new URL(baseUrl);
+  // Cloud URLs use just the origin; self-hosted may have a context path like /confluence
+  if (cloud) return url.origin;
+
+  const contextPath = url.pathname.replace(/\/+$/, "");
+  return `${url.origin}${contextPath}`;
+}
+
 /**
  * Generate the full chunkSource for a specific Confluence page so that we can resync it later.
  * This data is encrypted into a single `payload` query param so we can replay credentials later
@@ -271,4 +288,5 @@ function generateChunkSource(
 module.exports = {
   loadConfluence,
   fetchConfluencePage,
+  resolveConfluenceBaseUrl,
 };