Merge pull request #13447 from athoik/dev

feat(ocr): add support for Docling OCR engine and language configuration

backend/open_webui/config.py

@@ -1839,6 +1839,18 @@ DOCLING_SERVER_URL = PersistentConfig(
     os.getenv("DOCLING_SERVER_URL", "http://docling:5001"),
 )
 
+DOCLING_OCR_ENGINE = PersistentConfig(
+    "DOCLING_OCR_ENGINE",
+    "rag.docling_ocr_engine",
+    os.getenv("DOCLING_OCR_ENGINE", "tesseract"),
+)
+
+DOCLING_OCR_LANG = PersistentConfig(
+    "DOCLING_OCR_LANG",
+    "rag.docling_ocr_lang",
+    os.getenv("DOCLING_OCR_LANG", "eng,fra,deu,spa"),
+)
+
 DOCUMENT_INTELLIGENCE_ENDPOINT = PersistentConfig(
     "DOCUMENT_INTELLIGENCE_ENDPOINT",
     "rag.document_intelligence_endpoint",

backend/open_webui/main.py

@@ -202,6 +202,8 @@ from open_webui.config import (
     CONTENT_EXTRACTION_ENGINE,
     TIKA_SERVER_URL,
     DOCLING_SERVER_URL,
+    DOCLING_OCR_ENGINE,
+    DOCLING_OCR_LANG,
     DOCUMENT_INTELLIGENCE_ENDPOINT,
     DOCUMENT_INTELLIGENCE_KEY,
     MISTRAL_OCR_API_KEY,
@@ -635,6 +637,8 @@ app.state.config.ENABLE_WEB_LOADER_SSL_VERIFICATION = ENABLE_WEB_LOADER_SSL_VERI
 app.state.config.CONTENT_EXTRACTION_ENGINE = CONTENT_EXTRACTION_ENGINE
 app.state.config.TIKA_SERVER_URL = TIKA_SERVER_URL
 app.state.config.DOCLING_SERVER_URL = DOCLING_SERVER_URL
+app.state.config.DOCLING_OCR_ENGINE = DOCLING_OCR_ENGINE
+app.state.config.DOCLING_OCR_LANG = DOCLING_OCR_LANG
 app.state.config.DOCUMENT_INTELLIGENCE_ENDPOINT = DOCUMENT_INTELLIGENCE_ENDPOINT
 app.state.config.DOCUMENT_INTELLIGENCE_KEY = DOCUMENT_INTELLIGENCE_KEY
 app.state.config.MISTRAL_OCR_API_KEY = MISTRAL_OCR_API_KEY

backend/open_webui/retrieval/loaders/main.py

@@ -100,7 +100,7 @@ class TikaLoader:
             headers = {}
 
         if self.kwargs.get("PDF_EXTRACT_IMAGES") == True:
-            headers['X-Tika-PDFextractInlineImages'] = 'true'
+            headers["X-Tika-PDFextractInlineImages"] = "true"
 
         endpoint = self.url
         if not endpoint.endswith("/"):
@@ -124,10 +124,14 @@ class TikaLoader:
 
 
 class DoclingLoader:
-    def __init__(self, url, file_path=None, mime_type=None):
+    def __init__(
+        self, url, file_path=None, mime_type=None, ocr_engine=None, ocr_lang=None
+    ):
         self.url = url.rstrip("/")
         self.file_path = file_path
         self.mime_type = mime_type
+        self.ocr_engine = ocr_engine
+        self.ocr_lang = ocr_lang
 
     def load(self) -> list[Document]:
         with open(self.file_path, "rb") as f:
@@ -144,6 +148,12 @@ class DoclingLoader:
                 "table_mode": "accurate",
             }
 
+            if self.ocr_engine and self.ocr_lang:
+                params["ocr_engine"] = self.ocr_engine
+                params["ocr_lang"] = [
+                    lang.strip() for lang in self.ocr_lang.split(",") if lang.strip()
+                ]
+
             endpoint = f"{self.url}/v1alpha/convert/file"
             r = requests.post(endpoint, files=files, data=params)
 
@@ -212,6 +222,8 @@ class Loader:
                     url=self.kwargs.get("DOCLING_SERVER_URL"),
                     file_path=file_path,
                     mime_type=file_content_type,
+                    ocr_engine=self.kwargs.get("DOCLING_OCR_ENGINE"),
+                    ocr_lang=self.kwargs.get("DOCLING_OCR_LANG"),
                 )
         elif (
             self.engine == "document_intelligence"

backend/open_webui/routers/retrieval.py

@@ -378,6 +378,8 @@ async def get_rag_config(request: Request, user=Depends(get_admin_user)):
         "PDF_EXTRACT_IMAGES": request.app.state.config.PDF_EXTRACT_IMAGES,
         "TIKA_SERVER_URL": request.app.state.config.TIKA_SERVER_URL,
         "DOCLING_SERVER_URL": request.app.state.config.DOCLING_SERVER_URL,
+        "DOCLING_OCR_ENGINE": request.app.state.config.DOCLING_OCR_ENGINE,
+        "DOCLING_OCR_LANG": request.app.state.config.DOCLING_OCR_LANG,
         "DOCUMENT_INTELLIGENCE_ENDPOINT": request.app.state.config.DOCUMENT_INTELLIGENCE_ENDPOINT,
         "DOCUMENT_INTELLIGENCE_KEY": request.app.state.config.DOCUMENT_INTELLIGENCE_KEY,
         "MISTRAL_OCR_API_KEY": request.app.state.config.MISTRAL_OCR_API_KEY,
@@ -511,6 +513,8 @@ class ConfigForm(BaseModel):
     PDF_EXTRACT_IMAGES: Optional[bool] = None
     TIKA_SERVER_URL: Optional[str] = None
     DOCLING_SERVER_URL: Optional[str] = None
+    DOCLING_OCR_ENGINE: Optional[str] = None
+    DOCLING_OCR_LANG: Optional[str] = None
     DOCUMENT_INTELLIGENCE_ENDPOINT: Optional[str] = None
     DOCUMENT_INTELLIGENCE_KEY: Optional[str] = None
     MISTRAL_OCR_API_KEY: Optional[str] = None
@@ -600,6 +604,16 @@ async def update_rag_config(
         if form_data.DOCLING_SERVER_URL is not None
         else request.app.state.config.DOCLING_SERVER_URL
     )
+    request.app.state.config.DOCLING_OCR_ENGINE = (
+        form_data.DOCLING_OCR_ENGINE
+        if form_data.DOCLING_OCR_ENGINE is not None
+        else request.app.state.config.DOCLING_OCR_ENGINE
+    )
+    request.app.state.config.DOCLING_OCR_LANG = (
+        form_data.DOCLING_OCR_LANG
+        if form_data.DOCLING_OCR_LANG is not None
+        else request.app.state.config.DOCLING_OCR_LANG
+    )
     request.app.state.config.DOCUMENT_INTELLIGENCE_ENDPOINT = (
         form_data.DOCUMENT_INTELLIGENCE_ENDPOINT
         if form_data.DOCUMENT_INTELLIGENCE_ENDPOINT is not None
@@ -767,6 +781,8 @@ async def update_rag_config(
         "PDF_EXTRACT_IMAGES": request.app.state.config.PDF_EXTRACT_IMAGES,
         "TIKA_SERVER_URL": request.app.state.config.TIKA_SERVER_URL,
         "DOCLING_SERVER_URL": request.app.state.config.DOCLING_SERVER_URL,
+        "DOCLING_OCR_ENGINE": request.app.state.config.DOCLING_OCR_ENGINE,
+        "DOCLING_OCR_LANG": request.app.state.config.DOCLING_OCR_LANG,
         "DOCUMENT_INTELLIGENCE_ENDPOINT": request.app.state.config.DOCUMENT_INTELLIGENCE_ENDPOINT,
         "DOCUMENT_INTELLIGENCE_KEY": request.app.state.config.DOCUMENT_INTELLIGENCE_KEY,
         "MISTRAL_OCR_API_KEY": request.app.state.config.MISTRAL_OCR_API_KEY,
@@ -1080,6 +1096,8 @@ def process_file(
                     engine=request.app.state.config.CONTENT_EXTRACTION_ENGINE,
                     TIKA_SERVER_URL=request.app.state.config.TIKA_SERVER_URL,
                     DOCLING_SERVER_URL=request.app.state.config.DOCLING_SERVER_URL,
+                    DOCLING_OCR_ENGINE=request.app.state.config.DOCLING_OCR_ENGINE,
+                    DOCLING_OCR_LANG=request.app.state.config.DOCLING_OCR_LANG,
                     PDF_EXTRACT_IMAGES=request.app.state.config.PDF_EXTRACT_IMAGES,
                     DOCUMENT_INTELLIGENCE_ENDPOINT=request.app.state.config.DOCUMENT_INTELLIGENCE_ENDPOINT,
                     DOCUMENT_INTELLIGENCE_KEY=request.app.state.config.DOCUMENT_INTELLIGENCE_KEY,

src/lib/components/admin/Settings/Documents.svelte

@@ -161,6 +161,12 @@
 			toast.error($i18n.t('Docling Server URL required.'));
 			return;
 		}
+		if (RAGConfig.CONTENT_EXTRACTION_ENGINE === 'docling' &&
+			((RAGConfig.DOCLING_OCR_ENGINE === '' && RAGConfig.DOCLING_OCR_LANG !== '') ||
+			(RAGConfig.DOCLING_OCR_ENGINE !== '' && RAGConfig.DOCLING_OCR_LANG === ''))) {
+			toast.error($i18n.t('Both Docling OCR Engine and Language(s) must be provided or both left empty.'));
+			return;
+		}
 
 		if (
 			RAGConfig.CONTENT_EXTRACTION_ENGINE === 'document_intelligence' &&
@@ -326,6 +332,18 @@
 									bind:value={RAGConfig.DOCLING_SERVER_URL}
 								/>
 							</div>
+							<div class="flex w-full mt-2">
+								<input
+									class="flex-1 w-full rounded-lg text-sm bg-transparent outline-hidden"
+									placeholder={$i18n.t('Enter Docling OCR Engine')}
+									bind:value={RAGConfig.DOCLING_OCR_ENGINE}
+								/>
+								<input
+									class="flex-1 w-full rounded-lg text-sm bg-transparent outline-hidden"
+									placeholder={$i18n.t('Enter Docling OCR Language(s)')}
+									bind:value={RAGConfig.DOCLING_OCR_LANG}
+								/>
+							</div>
 						{:else if RAGConfig.CONTENT_EXTRACTION_ENGINE === 'document_intelligence'}
 							<div class="my-0.5 flex gap-2 pr-2">
 								<input