docling-haystack 0.2.0__tar.gz → 0.3.0__tar.gz

docling_haystack-0.3.0/CHANGELOG.md

@@ -0,0 +1,13 @@
+# Changelog
+
+## [integrations/docling-v0.2.0] - 2026-04-08
+
+### 🚀 Features
+
+- Add Docling document converter (#3066)
+
+### 🚜 Refactor
+
+- *(docling)* Add meta parameter to run(); introduce sources; deprecate paths (#3103)
+
+<!-- generated by git-cliff -->

{docling_haystack-0.2.0 → docling_haystack-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-haystack
-Version: 0.2.0
+Version: 0.3.0
 Summary: Haystack integration for docling
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/docling#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues

{docling_haystack-0.2.0 → docling_haystack-0.3.0}/src/haystack_integrations/components/converters/docling/converter.py

@@ -1,14 +1,15 @@
 """Docling Haystack converter module."""
 
 import json
-import os
-import tempfile
+import mimetypes
 import warnings
 from abc import ABC, abstractmethod
 from enum import Enum
+from io import BytesIO
 from pathlib import Path
 from typing import Any
 
+from docling_core.types.io import DocumentStream
 from haystack import Document, component
 from haystack.components.converters.utils import normalize_metadata
 from haystack.dataclasses import ByteStream
@@ -18,6 +19,29 @@ from docling.datamodel.document import DoclingDocument
 from docling.document_converter import DocumentConverter
 
 
+def _bytestream_to_document_stream(source: ByteStream) -> DocumentStream:
+    """
+    Build a `DocumentStream` from a Haystack `ByteStream`.
+
+    Resolves the stream name by checking common metadata keys (`file_path`, `file_name`, `name`) and falling back to
+    MIME-type extension guessing so that docling can reliably detect the input format.
+    """
+    meta = source.meta or {}
+    raw_name = meta.get("file_path") or meta.get("file_name") or meta.get("name")
+
+    if raw_name:
+        name = Path(raw_name).name
+    else:
+        name = "document"
+
+    if not Path(name).suffix and source.mime_type:
+        ext = mimetypes.guess_extension(source.mime_type)
+        if ext:
+            name = f"{name}{ext}"
+
+    return DocumentStream(name=name, stream=BytesIO(source.data))
+
+
 class ExportType(str, Enum):
     """Enumeration of available export types."""
 
@@ -141,14 +165,8 @@ class DoclingConverter:
         documents: list[Document] = []
         for source, source_meta in zip(sources, meta_list, strict=True):
             if isinstance(source, ByteStream):
-                # docling requires a file path; write ByteStream data to a temp file
-                with tempfile.NamedTemporaryFile(delete=False) as tmp:
-                    tmp.write(source.data)
-                    tmp_path = Path(tmp.name)
-                try:
-                    dl_doc = self._converter_instance.convert(source=tmp_path, **self.convert_kwargs).document
-                finally:
-                    os.unlink(tmp_path)
+                doc_stream = _bytestream_to_document_stream(source)
+                dl_doc = self._converter_instance.convert(source=doc_stream, **self.convert_kwargs).document
                 # merge ByteStream meta (e.g. file_path, mime_type) with user-supplied meta
                 merged_meta = {**(source.meta or {}), **source_meta}
             else:

{docling_haystack-0.2.0 → docling_haystack-0.3.0}/tests/test_converter.py

@@ -1,14 +1,18 @@
 import json
+import mimetypes
 import warnings
+from io import BytesIO
 from types import SimpleNamespace
 from typing import Any
-from unittest.mock import MagicMock, patch
+from unittest.mock import MagicMock
 
 import pytest
+from docling_core.types.io import DocumentStream
 from haystack.core.serialization import component_from_dict, component_to_dict
 from haystack.dataclasses import ByteStream
 
 from haystack_integrations.components.converters.docling import DoclingConverter, ExportType
+from haystack_integrations.components.converters.docling.converter import _bytestream_to_document_stream
 
 
 def test_run_doc_chunks_minimal() -> None:
@@ -356,13 +360,84 @@ def test_run_with_bytestream_source() -> None:
 
     bytestream = ByteStream(data=b"%PDF-1.4 fake pdf content", meta={"file_path": "uploaded.pdf"})
 
-    with patch("os.unlink"):
-        result = converter.run(sources=[bytestream])
+    result = converter.run(sources=[bytestream])
 
     documents = result["documents"]
     assert len(documents) == 1
     # ByteStream meta is merged into the output document
     assert documents[0].meta["file_path"] == "uploaded.pdf"
-    # docling was called with a temp file path, not the ByteStream directly
+    # docling was called with a DocumentStream, not a temp file path
     call_args = converter_mock.convert.call_args
-    assert call_args.kwargs["source"] != bytestream
+    passed_source = call_args.kwargs["source"]
+    assert isinstance(passed_source, DocumentStream)
+    assert passed_source.name == "uploaded.pdf"
+    assert isinstance(passed_source.stream, BytesIO)
+
+
+class TestBytestreamToDocumentStream:
+    def test_uses_file_path(self) -> None:
+        bs = ByteStream(data=b"data", meta={"file_path": "report.pdf"})
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "report.pdf"
+        assert ds.stream.read() == b"data"
+
+    def test_strips_directory_from_file_path(self) -> None:
+        bs = ByteStream(data=b"data", meta={"file_path": "/some/deep/path/report.pdf"})
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "report.pdf"
+
+    def test_uses_file_name_key(self) -> None:
+        bs = ByteStream(data=b"data", meta={"file_name": "slide-deck.pptx"})
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "slide-deck.pptx"
+
+    def test_uses_name_key(self) -> None:
+        bs = ByteStream(data=b"data", meta={"name": "notes.docx"})
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "notes.docx"
+
+    def test_file_path_takes_priority_over_file_name(self) -> None:
+        bs = ByteStream(data=b"data", meta={"file_path": "real.pdf", "file_name": "other.pdf"})
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "real.pdf"
+
+    def test_file_name_takes_priority_over_name(self) -> None:
+        bs = ByteStream(data=b"data", meta={"file_name": "chosen.pdf", "name": "ignored.pdf"})
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "chosen.pdf"
+
+    def test_guesses_extension_from_mime_type(self) -> None:
+        mime = "application/pdf"
+        expected_ext = mimetypes.guess_extension(mime)
+        bs = ByteStream(data=b"data", meta={"file_path": "report"}, mime_type=mime)
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == f"report{expected_ext}"
+
+    def test_keeps_extension_when_present(self) -> None:
+        # mime_type should not override an already-present extension
+        bs = ByteStream(data=b"data", meta={"file_path": "report.pdf"}, mime_type="text/plain")
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "report.pdf"
+
+    def test_no_meta_no_mime_type(self) -> None:
+        bs = ByteStream(data=b"data")
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "document"
+
+    def test_no_meta_with_mime_type(self) -> None:
+        mime = "application/pdf"
+        expected_ext = mimetypes.guess_extension(mime)
+        bs = ByteStream(data=b"data", mime_type=mime)
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == f"document{expected_ext}"
+
+    def test_empty_meta_no_mime_type(self) -> None:
+        bs = ByteStream(data=b"data", meta={})
+        ds = _bytestream_to_document_stream(bs)
+        assert ds.name == "document"
+
+    def test_returns_document_stream_with_bytesio(self) -> None:
+        bs = ByteStream(data=b"hello", meta={"file_path": "f.pdf"})
+        ds = _bytestream_to_document_stream(bs)
+        assert isinstance(ds, DocumentStream)
+        assert isinstance(ds.stream, BytesIO)