Support Confluence data center + allow for specifying labels to ignore (#624)

backend/danswer/configs/app_configs.py

@@ -109,9 +109,11 @@ POSTGRES_DB = os.environ.get("POSTGRES_DB") or "postgres"
 #####
 GOOGLE_DRIVE_INCLUDE_SHARED = False
 GOOGLE_DRIVE_FOLLOW_SHORTCUTS = False
+
 FILE_CONNECTOR_TMP_STORAGE_PATH = os.environ.get(
     "FILE_CONNECTOR_TMP_STORAGE_PATH", "/home/file_connector_storage"
 )
+
 # TODO these should be available for frontend configuration, via advanced options expandable
 WEB_CONNECTOR_IGNORED_CLASSES = os.environ.get(
     "WEB_CONNECTOR_IGNORED_CLASSES", "sidebar,footer"
@@ -128,6 +130,14 @@ NOTION_CONNECTOR_ENABLE_RECURSIVE_PAGE_LOOKUP = (
     == "true"
 )
 
+CONFLUENCE_CONNECTOR_LABELS_TO_SKIP = [
+    ignored_tag
+    for ignored_tag in os.environ.get("CONFLUENCE_CONNECTOR_LABELS_TO_SKIP", "").split(
+        ","
+    )
+    if ignored_tag
+]
+
 #####
 # Query Configs
 #####

backend/danswer/connectors/confluence/connector.py

@@ -9,6 +9,7 @@ from urllib.parse import urlparse
 from atlassian import Confluence  # type:ignore
 from requests import HTTPError
 
+from danswer.configs.app_configs import CONFLUENCE_CONNECTOR_LABELS_TO_SKIP
 from danswer.configs.app_configs import CONTINUE_ON_CONNECTOR_FAILURE
 from danswer.configs.app_configs import INDEX_BATCH_SIZE
 from danswer.configs.constants import DocumentSource
@@ -30,17 +31,12 @@ logger = setup_logger()
 # 3. Segment into Sections for more accurate linking, can split by headers but make sure no text/ordering is lost
 
 
-def extract_confluence_keys_from_url(wiki_url: str) -> tuple[str, str]:
+def _extract_confluence_keys_from_cloud_url(wiki_url: str) -> tuple[str, str]:
     """Sample
     https://danswer.atlassian.net/wiki/spaces/1234abcd/overview
-    wiki_base is danswer.atlassian.net/wiki
+    wiki_base is https://danswer.atlassian.net/wiki
     space is 1234abcd
     """
-    if ".atlassian.net/wiki/spaces/" not in wiki_url:
-        raise ValueError(
-            "Not a valid Confluence Wiki Link, unable to extract wiki base and space names"
-        )
-
     parsed_url = urlparse(wiki_url)
     wiki_base = (
         parsed_url.scheme
@@ -52,6 +48,42 @@ def extract_confluence_keys_from_url(wiki_url: str) -> tuple[str, str]:
     return wiki_base, space
 
 
+def _extract_confluence_keys_from_datacenter_url(wiki_url: str) -> tuple[str, str]:
+    """Sample
+    https://danswer.ai/confluence/display/1234abcd/overview
+    wiki_base is https://danswer.ai/confluence
+    space is 1234abcd
+    """
+    # /display/ is always right before the space and at the end of the base url
+    DISPLAY = "/display/"
+
+    parsed_url = urlparse(wiki_url)
+    wiki_base = (
+        parsed_url.scheme
+        + "://"
+        + parsed_url.netloc
+        + parsed_url.path.split(DISPLAY)[0]
+    )
+    space = DISPLAY.join(parsed_url.path.split(DISPLAY)[1:]).split("/")[0]
+    return wiki_base, space
+
+
+def extract_confluence_keys_from_url(wiki_url: str) -> tuple[str, str, bool]:
+    is_confluence_cloud = ".atlassian.net/wiki/spaces/" in wiki_url
+
+    try:
+        if is_confluence_cloud:
+            wiki_base, space = _extract_confluence_keys_from_cloud_url(wiki_url)
+        else:
+            wiki_base, space = _extract_confluence_keys_from_datacenter_url(wiki_url)
+    except Exception as e:
+        error_msg = f"Not a valid Confluence Wiki Link, unable to extract wiki base and space names. Exception: {e}"
+        logger.error(error_msg)
+        raise ValueError(error_msg)
+
+    return wiki_base, space, is_confluence_cloud
+
+
 def _comment_dfs(
     comments_str: str,
     comment_pages: Collection[dict[str, Any]],
@@ -79,10 +111,17 @@ class ConfluenceConnector(LoadConnector, PollConnector):
         wiki_page_url: str,
         batch_size: int = INDEX_BATCH_SIZE,
         continue_on_failure: bool = CONTINUE_ON_CONNECTOR_FAILURE,
+        # if a page has one of the labels specified in this list, we will just
+        # skip it. This is generally used to avoid indexing extra sensitive
+        # pages.
+        labels_to_skip: list[str] = CONFLUENCE_CONNECTOR_LABELS_TO_SKIP,
     ) -> None:
         self.batch_size = batch_size
         self.continue_on_failure = continue_on_failure
-        self.wiki_base, self.space = extract_confluence_keys_from_url(wiki_page_url)
+        self.labels_to_skip = set(labels_to_skip)
+        self.wiki_base, self.space, self.is_cloud = extract_confluence_keys_from_url(
+            wiki_page_url
+        )
         self.confluence_client: Confluence | None = None
 
     def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
@@ -90,9 +129,10 @@ class ConfluenceConnector(LoadConnector, PollConnector):
         access_token = credentials["confluence_access_token"]
         self.confluence_client = Confluence(
             url=self.wiki_base,
-            username=username,
+            # passing in username causes issues for Confluence data center
+            username=username if self.is_cloud else None,
             password=access_token,
-            cloud=True,
+            cloud=self.is_cloud,
         )
         return None
 
@@ -185,6 +225,17 @@ class ConfluenceConnector(LoadConnector, PollConnector):
             )
             return ""
 
+    def _fetch_labels(self, confluence_client: Confluence, page_id: str) -> list[str]:
+        try:
+            labels_response = confluence_client.get_page_labels(page_id)
+            return [label["name"] for label in labels_response["results"]]
+        except Exception as e:
+            if not self.continue_on_failure:
+                raise e
+
+            logger.exception("Ran into exception when fetching labels from Confluence")
+            return []
+
     def _get_doc_batch(
         self, start_ind: int, time_filter: Callable[[datetime], bool] | None = None
     ) -> tuple[list[Document], int]:
@@ -200,6 +251,19 @@ class ConfluenceConnector(LoadConnector, PollConnector):
             last_modified = datetime.fromisoformat(last_modified_str)
 
             if time_filter is None or time_filter(last_modified):
+                page_id = page["id"]
+
+                # check disallowed labels
+                if self.labels_to_skip:
+                    page_labels = self._fetch_labels(self.confluence_client, page_id)
+                    label_intersection = self.labels_to_skip.intersection(page_labels)
+                    if label_intersection:
+                        logger.info(
+                            f"Page with ID '{page_id}' has a label which has been "
+                            f"designated as disallowed: {label_intersection}. Skipping."
+                        )
+                        continue
+
                 page_html = (
                     page["body"]
                     .get("storage", page["body"].get("view", {}))
@@ -212,7 +276,7 @@ class ConfluenceConnector(LoadConnector, PollConnector):
                 page_text = (
                     page.get("title", "") + "\n" + parse_html_page_basic(page_html)
                 )
-                comments_text = self._fetch_comments(self.confluence_client, page["id"])
+                comments_text = self._fetch_comments(self.confluence_client, page_id)
                 page_text += comments_text
 
                 doc_batch.append(

deployment/docker_compose/docker-compose.dev.yml

@@ -74,8 +74,10 @@ services:
       - API_TYPE_OPENAI=${API_TYPE_OPENAI:-}
       - API_VERSION_OPENAI=${API_VERSION_OPENAI:-}
       - AZURE_DEPLOYMENT_ID=${AZURE_DEPLOYMENT_ID:-}
-      - CONTINUE_ON_CONNECTOR_FAILURE=${CONTINUE_ON_CONNECTOR_FAILURE:-}
       - NUM_INDEXING_WORKERS=${NUM_INDEXING_WORKERS:-}
+      # Connector Configs
+      - CONTINUE_ON_CONNECTOR_FAILURE=${CONTINUE_ON_CONNECTOR_FAILURE:-}
+      - CONFLUENCE_CONNECTOR_LABELS_TO_SKIP=${CONFLUENCE_CONNECTOR_LABELS_TO_SKIP:-}
       # Danswer SlackBot Configs
       - DANSWER_BOT_SLACK_APP_TOKEN=${DANSWER_BOT_SLACK_APP_TOKEN:-}
       - DANSWER_BOT_SLACK_BOT_TOKEN=${DANSWER_BOT_SLACK_BOT_TOKEN:-}

web/src/app/admin/connectors/confluence/page.tsx

@@ -20,17 +20,37 @@ import { ConnectorsTable } from "@/components/admin/connectors/table/ConnectorsT
 import { usePopup } from "@/components/admin/connectors/Popup";
 import { usePublicCredentials } from "@/lib/hooks";
 
-// Copied from the `extract_confluence_keys_from_url` function
-const extractSpaceFromUrl = (wikiUrl: string): string | null => {
-  if (!wikiUrl.includes(".atlassian.net/wiki/spaces/")) {
-    return null;
-  }
-
+const extractSpaceFromCloudUrl = (wikiUrl: string): string => {
   const parsedUrl = new URL(wikiUrl);
   const space = parsedUrl.pathname.split("/")[3];
   return space;
 };
 
+const extractSpaceFromDataCenterUrl = (wikiUrl: string): string => {
+  const DISPLAY = "/display/";
+
+  const parsedUrl = new URL(wikiUrl);
+  const spaceSpecificSection = parsedUrl.pathname
+    .split(DISPLAY)
+    .slice(1)
+    .join(DISPLAY);
+  const space = spaceSpecificSection.split("/")[0];
+  return space;
+};
+
+// Copied from the `extract_confluence_keys_from_url` function
+const extractSpaceFromUrl = (wikiUrl: string): string | null => {
+  try {
+    if (wikiUrl.includes(".atlassian.net/wiki/spaces/")) {
+      return extractSpaceFromCloudUrl(wikiUrl);
+    }
+    return extractSpaceFromDataCenterUrl(wikiUrl);
+  } catch (e) {
+    console.log("Failed to extract space from url", e);
+    return null;
+  }
+};
+
 const Main = () => {
   const { popup, setPopup } = usePopup();