fix: scope logging configuration to the haystack logger instead of root

haystack/logging.py

@@ -392,12 +392,15 @@ def configure_logging(use_json: bool | None = None) -> None:
     # Use OUR `ProcessorFormatter` to format all `logging` entries.
     handler.setFormatter(formatter)
 
-    root_logger = logging.getLogger()
+    # Scope the handler to the `haystack` logger hierarchy instead of the root logger so that we do not
+    # hijack the host application's logging configuration (e.g. a downstream `logging.basicConfig()` call).
+    # See https://github.com/deepset-ai/haystack/issues/8681.
+    haystack_logger = logging.getLogger("haystack")
     # avoid adding our handler twice
     old_handlers = [
         h
-        for h in root_logger.handlers
+        for h in haystack_logger.handlers
         if not (isinstance(h, logging.StreamHandler) and h.name == "HaystackLoggingHandler")
     ]
     new_handlers = [handler, *old_handlers]
-    root_logger.handlers = new_handlers
+    haystack_logger.handlers = new_handlers

releasenotes/notes/scope-logging-to-haystack-logger-8b544e319b11b368.yaml

@@ -0,0 +1,7 @@
+---
+fixes:
+  - |
+    `configure_logging()` no longer attaches its handler to (and replaces the handlers of) the root logger.
+    The Haystack logging handler is now scoped to the `haystack` logger hierarchy, so importing Haystack no
+    longer hijacks the host application's logging configuration. A downstream `logging.basicConfig()` call is
+    now respected instead of being silently overridden.