Pass OCR options from conversion_options to docling-serve chunker

CHANGELOG.md

@@ -1,6 +1,12 @@
 # Changelog
 ## [Unreleased]
 
+### Added
+
+- **docling-serve Chunker OCR Options**: The docling-serve chunker now respects OCR settings from `conversion_options`
+  - Passes `do_ocr`, `force_ocr`, `ocr_engine`, and `ocr_lang` to the chunking API
+  - Allows disabling OCR via config when running docling-serve in read-only containers
+
 ### Fixed
 
 - **CI**: Cache HuggingFace tokenizer to prevent flaky test failures when HuggingFace has transient outages

docs/configuration/processing.md

@@ -230,6 +230,8 @@ providers:
 
 Conversion options work identically for both local and remote processing.
 
+**Note:** When using `chunker: docling-serve`, OCR options (`do_ocr`, `force_ocr`, `ocr_engine`, `ocr_lang`) from `conversion_options` are passed to the chunking API. This is useful when running docling-serve in a read-only container where OCR model downloads fail—set `do_ocr: false` to disable OCR entirely.
+
 ### Chunking Strategies
 
 **Hybrid chunking** (default):

haiku_rag_slim/haiku/rag/chunkers/docling_serve.py

@@ -61,9 +61,13 @@ class DoclingServeChunker(DocumentChunker):
         )
         self.chunker_type = config.processing.chunker_type
 
-    def _build_chunking_data(self) -> dict[str, str]:
+    def _build_chunking_data(self) -> dict[str, str | list[str]]:
         """Build form data for chunking request."""
-        return {
+        opts = self.config.processing.conversion_options
+        data: dict[str, str | list[str]] = {
+            "convert_do_ocr": str(opts.do_ocr).lower(),
+            "convert_force_ocr": str(opts.force_ocr).lower(),
+            "convert_ocr_engine": opts.ocr_engine,
             "chunking_max_tokens": str(self.config.processing.chunk_size),
             "chunking_tokenizer": self.config.processing.chunking_tokenizer,
             "chunking_merge_peers": str(
@@ -73,6 +77,9 @@ class DoclingServeChunker(DocumentChunker):
                 self.config.processing.chunking_use_markdown_tables
             ).lower(),
         }
+        if opts.ocr_lang:
+            data["convert_ocr_lang"] = opts.ocr_lang
+        return data
 
     async def _call_chunk_api(self, document: "DoclingDocument") -> list[dict]:
         """Call docling-serve chunking API and return raw chunk data.

tests/test_chunker.py

@@ -351,6 +351,10 @@ class TestDoclingServeChunker:
         config.processing.chunk_size = 512
         config.processing.chunking_merge_peers = False
         config.processing.chunking_use_markdown_tables = True
+        config.processing.conversion_options.do_ocr = False
+        config.processing.conversion_options.force_ocr = True
+        config.processing.conversion_options.ocr_engine = "tesseract"
+        config.processing.conversion_options.ocr_lang = ["en", "de"]
         chunker = DoclingServeChunker(config)
 
         result_data = {"chunks": [{"text": "Chunk 1", "chunk_index": 0}]}
@@ -370,6 +374,40 @@ class TestDoclingServeChunker:
         assert data["chunking_max_tokens"] == "512"
         assert data["chunking_merge_peers"] == "false"
         assert data["chunking_use_markdown_tables"] == "true"
+        # OCR options from conversion_options
+        assert data["convert_do_ocr"] == "false"
+        assert data["convert_force_ocr"] == "true"
+        assert data["convert_ocr_engine"] == "tesseract"
+        assert data["convert_ocr_lang"] == ["en", "de"]
+
+    @pytest.mark.asyncio
+    @patch("haiku.rag.providers.docling_serve.httpx.AsyncClient")
+    async def test_chunk_omits_empty_ocr_lang(self, mock_client_class, config):
+        """Test that ocr_lang is omitted when empty (default)."""
+        # Ensure ocr_lang is empty (default)
+        config.processing.conversion_options.ocr_lang = []
+        chunker = DoclingServeChunker(config)
+
+        result_data = {"chunks": [{"text": "Chunk 1", "chunk_index": 0}]}
+        submit_resp, poll_resp, result_resp = create_async_workflow_mocks(result_data)
+
+        mock_client = AsyncMock()
+        mock_client.post = AsyncMock(return_value=submit_resp)
+        mock_client.get = AsyncMock(side_effect=[poll_resp, result_resp])
+        mock_client_class.return_value.__aenter__.return_value = mock_client
+
+        converter = get_converter(Config)
+        doc = await converter.convert_text("# Test", name="test.md")
+        await chunker.chunk(doc)
+
+        call_kwargs = mock_client.post.call_args.kwargs
+        data = call_kwargs["data"]
+        # OCR options should use defaults
+        assert data["convert_do_ocr"] == "true"
+        assert data["convert_force_ocr"] == "false"
+        assert data["convert_ocr_engine"] == "auto"
+        # ocr_lang should NOT be present when empty
+        assert "convert_ocr_lang" not in data
 
     @pytest.mark.asyncio
     @patch("haiku.rag.providers.docling_serve.httpx.AsyncClient")