fix: keep the server_url path when splitting PDFs

CHANGELOG.md

@@ -1,3 +1,8 @@
+## 0.46.3
+
+### Fixes
+* Keep the deployment path in split-PDF requests. With a `server_url` that includes a path, such as `https://host/deployment`, splitting a PDF sent its collection request to `https://host/general/docs` instead of `https://host/deployment/general/docs`, which returned 404 before any chunks were processed. The collection URL is now derived from the partition request by replacing only its trailing `/general/v0/general` route, for sync and async calls and for a `server_url` passed to the operation.
+
 ## 0.46.2
 
 ### Fixes

_test_unstructured_client/unit/test_request_utils.py

@@ -4,7 +4,6 @@
 
 from unstructured_client._hooks.custom.request_utils import (
     create_pdf_chunk_request_params,
-    get_base_url,
     get_multipart_stream_fields,
 )
 
@@ -75,14 +74,3 @@ def test_create_pdf_chunk_request_params(input_form_data, page_number, expected_
     assert form_data == expected_form_data
 
 
-@pytest.mark.parametrize(
-    ("url", "expected_base_url"),
-    [
-        ("https://api.unstructuredapp.io/general/v0/general", "https://api.unstructuredapp.io"),
-        ("https://api.unstructuredapp.io/general/v0/general?some_param=23", "https://api.unstructuredapp.io"),
-        ("http://localhost:3000/general/v0/general", "http://localhost:3000"),
-    ],
-)
-def test_get_base_url(url: str, expected_base_url: str):
-    assert get_base_url(url) == expected_base_url
-

_test_unstructured_client/unit/test_split_collection_urls.py

@@ -0,0 +1,123 @@
+import asyncio
+import io
+
+import httpx
+import pytest
+from pypdf import PdfWriter
+
+from unstructured_client import UnstructuredClient
+from unstructured_client._hooks.custom.split_pdf_hook import _get_collection_url
+from unstructured_client.models import operations, shared
+from unstructured_client.utils import BackoffStrategy, RetryConfig
+
+
+PREFIXES = [
+    "",
+    "/deployment",
+    "/one/two",
+    "/general/v0/general/deployment",
+    "/encoded%2Fprefix",
+]
+
+
+@pytest.fixture
+def pdf_bytes():
+    document = PdfWriter()
+    for _ in range(4):
+        document.add_blank_page(width=612, height=792)
+    buffer = io.BytesIO()
+    document.write(buffer)
+    return buffer.getvalue()
+
+
+@pytest.mark.parametrize("prefix", PREFIXES)
+def test_collection_url_preserves_encoded_prefix_and_replaces_terminal_route(prefix):
+    url = httpx.URL(
+        f"https://local.invalid{prefix}/general/v0/general?strategy=fast#fragment"
+    )
+    assert (
+        str(_get_collection_url(url)) == f"https://local.invalid{prefix}/general/docs"
+    )
+
+
+def make_sdk(monkeypatch, prefix, override):
+    sent = []
+
+    def respond(request):
+        sent.append((request.method, str(request.url)))
+        return httpx.Response(200, json=[{"text": "hello"}], request=request)
+
+    transport = httpx.MockTransport(respond)
+    original_async_client = httpx.AsyncClient
+    sync_client = httpx.Client(transport=transport)
+    async_client = original_async_client(transport=transport)
+    sdk = UnstructuredClient(
+        server_url="http://unused.invalid"
+        if override
+        else f"http://local.invalid{prefix}",
+        client=sync_client,
+        async_client=async_client,
+    )
+    # Split chunks create their own async HTTP client; keep every request offline.
+    monkeypatch.setattr(
+        httpx,
+        "AsyncClient",
+        lambda **kwargs: original_async_client(transport=transport, **kwargs),
+    )
+    return sdk, sent, sync_client, async_client
+
+
+def request(pdf_bytes, split):
+    return operations.PartitionRequest(
+        partition_parameters=shared.PartitionParameters(
+            files=shared.Files(content=pdf_bytes, file_name="test.pdf"),
+            split_pdf_page=split,
+        )
+    )
+
+
+def assert_routes(sent, prefix, split):
+    partition = f"http://local.invalid{prefix}/general/v0/general"
+    if split:
+        assert sent[0] == ("GET", f"http://local.invalid{prefix}/general/docs")
+        assert len(sent) == 3
+        assert sent[1:] == [("POST", partition)] * 2
+    else:
+        assert sent == [("POST", partition)]
+
+
+@pytest.mark.parametrize("prefix", PREFIXES)
+@pytest.mark.parametrize("override", [False, True])
+@pytest.mark.parametrize("split", [False, True])
+def test_public_sync_sdk_preserves_deployment_routes(
+    monkeypatch, pdf_bytes, prefix, override, split
+):
+    sdk, sent, sync_client, async_client = make_sdk(monkeypatch, prefix, override)
+    result = sdk.general.partition(
+        request=request(pdf_bytes, split),
+        server_url=f"http://local.invalid{prefix}" if override else None,
+        retries=RetryConfig("none", BackoffStrategy(1, 1, 1, 1), False),
+    )
+    assert result.status_code == 200
+    assert_routes(sent, prefix, split)
+    sync_client.close()
+    asyncio.run(async_client.aclose())
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("prefix", PREFIXES)
+@pytest.mark.parametrize("override", [False, True])
+@pytest.mark.parametrize("split", [False, True])
+async def test_public_async_sdk_preserves_deployment_routes(
+    monkeypatch, pdf_bytes, prefix, override, split
+):
+    sdk, sent, sync_client, async_client = make_sdk(monkeypatch, prefix, override)
+    result = await sdk.general.partition_async(
+        request=request(pdf_bytes, split),
+        server_url=f"http://local.invalid{prefix}" if override else None,
+        retries=RetryConfig("none", BackoffStrategy(1, 1, 1, 1), False),
+    )
+    assert result.status_code == 200
+    assert_routes(sent, prefix, split)
+    sync_client.close()
+    await async_client.aclose()

_test_unstructured_client/unit/test_split_pdf_hook.py

@@ -725,14 +725,12 @@ def test_before_request_raises_pdf_validation_error_when_pdf_check_fails():
 
     with patch("unstructured_client._hooks.custom.request_utils.get_multipart_stream_fields") as mock_get_fields, \
          patch("unstructured_client._hooks.custom.pdf_utils.read_pdf") as mock_read_pdf, \
-         patch("unstructured_client._hooks.custom.pdf_utils.check_pdf") as mock_check_pdf, \
-         patch("unstructured_client._hooks.custom.request_utils.get_base_url") as mock_get_base_url:
+         patch("unstructured_client._hooks.custom.pdf_utils.check_pdf") as mock_check_pdf:
 
         # Set up the mocks
         mock_get_fields.return_value = mock_form_data
         mock_read_pdf.return_value = mock_pdf_reader
         mock_check_pdf.side_effect = pdf_utils.PDFValidationError(error_message)
-        mock_get_base_url.return_value = "http://localhost:8888"
 
         # Call the method under test and verify it raises PDFValidationError
         with pytest.raises(pdf_utils.PDFValidationError) as exc_info:
@@ -857,14 +855,12 @@ def _make_hook_with_split_request(
     with patch("unstructured_client._hooks.custom.request_utils.get_multipart_stream_fields") as mock_get_fields, \
          patch("unstructured_client._hooks.custom.pdf_utils.read_pdf") as mock_read_pdf, \
          patch("unstructured_client._hooks.custom.pdf_utils.check_pdf") as mock_check_pdf, \
-         patch("unstructured_client._hooks.custom.request_utils.get_base_url") as mock_get_base_url, \
          patch.object(hook, "_trim_large_pages", side_effect=lambda pdf, fd: pdf), \
          patch.object(hook, "_get_pdf_chunk_paths", return_value=[]), \
          patch.object(hook, "_get_pdf_chunks_in_memory", return_value=pdf_chunks or []):
         mock_get_fields.return_value = form_data
         mock_read_pdf.return_value = mock_pdf_reader
         mock_check_pdf.return_value = mock_pdf_reader
-        mock_get_base_url.return_value = "http://localhost:8888"
 
         result = hook.before_request(hook_ctx, request)
 
@@ -2169,7 +2165,6 @@ def _chunk_paths_side_effect(*args, **kwargs):
     with patch("unstructured_client._hooks.custom.request_utils.get_multipart_stream_fields") as mock_get_fields, \
          patch("unstructured_client._hooks.custom.pdf_utils.read_pdf") as mock_read_pdf, \
          patch("unstructured_client._hooks.custom.pdf_utils.check_pdf") as mock_check_pdf, \
-         patch("unstructured_client._hooks.custom.request_utils.get_base_url") as mock_get_base_url, \
          patch("unstructured_client._hooks.custom.split_pdf_hook.futures.ThreadPoolExecutor", return_value=executor), \
          patch("unstructured_client._hooks.custom.request_utils.create_pdf_chunk_request", side_effect=RuntimeError("chunk build failed")), \
          patch.object(hook, "_trim_large_pages", side_effect=lambda pdf, fd: pdf), \
@@ -2178,7 +2173,6 @@ def _chunk_paths_side_effect(*args, **kwargs):
         mock_get_fields.return_value = mock_form_data
         mock_read_pdf.return_value = mock_pdf_reader
         mock_check_pdf.return_value = mock_pdf_reader
-        mock_get_base_url.return_value = "http://localhost:8888"
 
         with pytest.raises(RuntimeError, match="chunk build failed"):
             hook.before_request(hook_ctx, request)

src/unstructured_client/_hooks/custom/request_utils.py

@@ -7,10 +7,8 @@
 import os
 import tempfile
 from typing import Tuple, Any, BinaryIO, Optional, TextIO
-from urllib.parse import urlparse
 
 import httpx
-from httpx import URL
 from httpx._multipart import DataField, FileField
 
 from unstructured_client._hooks.custom.common import UNSTRUCTURED_CLIENT_LOGGER_NAME
@@ -467,16 +465,3 @@ def create_elements_file_response(elements_file: str) -> httpx.Response:
     )
     setattr(response, "_content", content)
     return response
-
-
-def get_base_url(url: str | URL) -> str:
-    """Extracts the base URL from the given URL.
-
-    Args:
-        url: The URL.
-
-    Returns:
-        The base URL.
-    """
-    parsed_url = urlparse(str(url))
-    return f"{parsed_url.scheme}://{parsed_url.netloc}"

src/unstructured_client/_hooks/custom/split_pdf_hook.py

@@ -35,7 +35,6 @@
     PARTITION_FORM_SPLIT_PDF_PAGE_KEY,
     PARTITION_FORM_STARTING_PAGE_NUMBER_KEY,
 )
-from unstructured_client._hooks.custom.request_utils import get_base_url
 from unstructured_client._hooks.types import (
     AfterErrorContext,
     AfterErrorHook,
@@ -66,6 +65,17 @@
 SPLIT_PDF_HEADER_PREFIX = "X-Unstructured-Split-"
 
 
+def _get_collection_url(partition_url: httpx.URL) -> httpx.URL:
+    """Keep deployment prefixes when switching to the split collection route.
+
+    Use the encoded path and replace only the terminal partition route. The
+    collection request does not inherit partition query parameters or fragments.
+    """
+    path = partition_url.raw_path.split(b"?", 1)[0]
+    collection_path = path.removesuffix(b"/general/v0/general") + b"/general/docs"
+    return partition_url.copy_with(raw_path=collection_path, query=None, fragment=None)
+
+
 class ChunkExecutionError(Exception):
     def __init__(self, index: int, inner: BaseException):
         super().__init__(str(inner))
@@ -719,11 +729,11 @@ def _before_request_unlocked(
             the synthetic collection request; otherwise, the original request.
         """
 
-        # Actually the general.partition operation overwrites the default client's base url (as
-        # the platform operations do). Here we need to get the base url from the request object.
+        # general.partition can override the client URL, so derive the collection URL
+        # from the actual partition request, preserving its deployment prefix.
         if hook_ctx.operation_id != "partition":
             return request
-        partition_base_url = get_base_url(request.url)
+        collection_url = _get_collection_url(request.url)
 
         if self.client is None:
             logger.warning("HTTP client not accessible! Continuing without splitting.")
@@ -926,7 +936,7 @@ def _before_request_unlocked(
             dummy_request_extensions[OPERATION_ID_EXTENSION_KEY] = operation_id
             return httpx.Request(
                 "GET",
-                f"{partition_base_url}/general/docs",
+                collection_url,
                 headers={"operation_id": operation_id},
                 extensions=dummy_request_extensions,
             )

src/unstructured_client/_version.py

@@ -3,10 +3,10 @@
 import importlib.metadata
 
 __title__: str = "unstructured-client"
-__version__: str = "0.46.2"
+__version__: str = "0.46.3"
 __openapi_doc_version__: str = "1.2.31"
 __gen_version__: str = "2.680.0"
-__user_agent__: str = "speakeasy-sdk/python 0.46.2 2.680.0 1.2.31 unstructured-client"
+__user_agent__: str = "speakeasy-sdk/python 0.46.3 2.680.0 1.2.31 unstructured-client"
 
 try:
     if __package__ is not None: