fix: strip only trailing extension in LlamaCloud local file name

get_local_file_name used str.replace(file_ext, "") to drop the extension,
which removes every occurrence of the extension substring. For files whose
stem also ends with the extension (e.g. report.pdf.pdf) the stem was
corrupted (report instead of report.pdf). Use os.path.splitext root instead.

Author: JSap0914

.changeset/llamacloud-local-file-name-extension.md

@@ -0,0 +1,10 @@
+---
+"@create-llama/llama-index-server": patch
+---
+
+fix: only strip the trailing extension when building the local LlamaCloud file name
+
+`get_local_file_name` used `str.replace(file_ext, "")`, which removes every
+occurrence of the extension string. For files whose stem also ends with the
+extension (e.g. `report.pdf.pdf`) this corrupted the stem (`report` instead of
+`report.pdf`). It now uses the root returned by `os.path.splitext`.

python/llama-index-server/llama_index/server/utils/llamacloud.py

@@ -22,8 +22,10 @@ def get_local_file_name(
     if llamacloud_file_name is None or pipeline_id is None:
         raise ValueError("Couldn't find llamacloud_file_name and pipeline_id")
     # Construct the local file name
-    file_ext = os.path.splitext(llamacloud_file_name)[1]
-    file_name = llamacloud_file_name.replace(file_ext, "")
+    file_root, file_ext = os.path.splitext(llamacloud_file_name)
+    # Only strip the trailing extension; using str.replace would remove every
+    # occurrence of the extension (e.g. "report.pdf.pdf" -> "report").
+    file_name = file_root
     sanitized_file_name = re.sub(r"[^A-Za-z0-9_\-]", "_", file_name)
     return f"{sanitized_file_name}_{pipeline_id}{file_ext}"
 

python/llama-index-server/tests/utils/test_llamacloud.py

@@ -0,0 +1,37 @@
+from llama_index.server.utils.llamacloud import (
+    get_local_file_name,
+    is_llamacloud_file,
+)
+
+
+def test_get_local_file_name_single_extension() -> None:
+    assert (
+        get_local_file_name(llamacloud_file_name="report.pdf", pipeline_id="123")
+        == "report_123.pdf"
+    )
+
+
+def test_get_local_file_name_keeps_repeated_extension_in_stem() -> None:
+    # The stem also ends with the extension string. Only the trailing
+    # extension must be stripped, otherwise the stem is corrupted.
+    assert (
+        get_local_file_name(llamacloud_file_name="report.pdf.pdf", pipeline_id="123")
+        == "report_pdf_123.pdf"
+    )
+
+
+def test_get_local_file_name_extension_substring_in_stem() -> None:
+    assert (
+        get_local_file_name(llamacloud_file_name="backup.tar.tar", pipeline_id="abc")
+        == "backup_tar_abc.tar"
+    )
+
+
+def test_get_local_file_name_from_metadata() -> None:
+    metadata = {"file_name": "my.report.pdf.pdf", "pipeline_id": "p1"}
+    assert get_local_file_name(node_metadata=metadata) == "my_report_pdf_p1.pdf"
+
+
+def test_is_llamacloud_file() -> None:
+    assert is_llamacloud_file({"pipeline_id": "p1"}) is True
+    assert is_llamacloud_file({"file_name": "a.pdf"}) is False