1  ---
  2  title: "DocumentLengthRouter"
  3  id: documentlengthrouter
  4  slug: "/documentlengthrouter"
  5  description: "Routes documents to different output connections based on the length of their `content` field."
  6  ---
  7  
  8  # DocumentLengthRouter
  9  
 10  Routes documents to different output connections based on the length of their `content` field.
 11  
 12  <div className="key-value-table">
 13  
 14  |  |  |
 15  | --- | --- |
 16  | **Most common position in a pipeline** | Flexible |
 17  | **Mandatory run variables** | `documents`: A list of documents |
 18  | **Output variables** | `short_documents`: A list of documents where `content` is None or the length of `content` is less than or equal to the threshold.  <br /> <br />`long_documents`: A list of documents where the length of `content` is greater than the threshold. |
 19  | **API reference** | [Routers](/reference/routers-api) |
 20  | **GitHub link** | https://github.com/deepset-ai/haystack/blob/main/haystack/components/routers/document_length_router.py |
 21  
 22  </div>
 23  
 24  ## Overview
 25  
 26  `DocumentLengthRouter` routes documents to different output connections based on the length of their `content` field.
 27  
 28  It allows to set a `threshold` init parameter. Documents where `content` is None, or the length of `content` is less than or equal to the threshold are routed to "short_documents". Others are routed to "long_documents".
 29  
 30  A common use case for `DocumentLengthRouter` is handling documents obtained from PDFs that contain non-text content, such as scanned pages or images. This component can detect empty or low-content documents and route them to components that perform OCR, generate captions, or compute image embeddings.
 31  
 32  ## Usage
 33  
 34  ### On its own
 35  
 36  ```python
 37  from haystack.components.routers import DocumentLengthRouter
 38  from haystack.dataclasses import Document
 39  
 40  docs = [
 41      Document(content="Short"),
 42      Document(content="Long document " * 20),
 43  ]
 44  
 45  router = DocumentLengthRouter(threshold=10)
 46  
 47  result = router.run(documents=docs)
 48  print(result)
 49  
 50  ## {
 51  ## "short_documents": [Document(content="Short", ...)],
 52  ## "long_documents": [Document(content="Long document ...", ...)],
 53  ## }
 54  ```
 55  
 56  ### In a pipeline
 57  
 58  In the following indexing pipeline, the `PyPDFToDocument` Converter extracts text from PDF files.
 59  Documents are then split by pages using a `DocumentSplitter`.
 60  Next, the `DocumentLengthRouter` routes short documents to `LLMDocumentContentExtractor` to extract text, which is particularly useful for non-textual, image-based pages.
 61  Finally, all documents are sent to the `DocumentWriter` and written to the Document Store.
 62  
 63  ```python
 64  from haystack import Pipeline
 65  from haystack.components.converters import PyPDFToDocument
 66  from haystack.components.extractors.image import LLMDocumentContentExtractor
 67  from haystack.components.generators.chat import OpenAIChatGenerator
 68  from haystack.components.preprocessors import DocumentSplitter
 69  from haystack.components.routers import DocumentLengthRouter
 70  from haystack.components.writers import DocumentWriter
 71  from haystack.document_stores.in_memory import InMemoryDocumentStore
 72  
 73  document_store = InMemoryDocumentStore()
 74  
 75  indexing_pipe = Pipeline()
 76  indexing_pipe.add_component("pdf_converter", PyPDFToDocument(store_full_path=True))
 77  ## setting skip_empty_documents=False is important here because the
 78  ## LLMDocumentContentExtractor can extract text from non-textual documents
 79  ## that otherwise would be skipped
 80  indexing_pipe.add_component(
 81      "pdf_splitter",
 82      DocumentSplitter(split_by="page", split_length=1, skip_empty_documents=False),
 83  )
 84  indexing_pipe.add_component("doc_length_router", DocumentLengthRouter(threshold=10))
 85  indexing_pipe.add_component(
 86      "content_extractor",
 87      LLMDocumentContentExtractor(
 88          chat_generator=OpenAIChatGenerator(model="gpt-4.1-mini"),
 89      ),
 90  )
 91  indexing_pipe.add_component(
 92      "document_writer",
 93      DocumentWriter(document_store=document_store),
 94  )
 95  
 96  indexing_pipe.connect("pdf_converter.documents", "pdf_splitter.documents")
 97  indexing_pipe.connect("pdf_splitter.documents", "doc_length_router.documents")
 98  ## The short PDF pages will be enriched/captioned
 99  indexing_pipe.connect(
100      "doc_length_router.short_documents",
101      "content_extractor.documents",
102  )
103  indexing_pipe.connect("doc_length_router.long_documents", "document_writer.documents")
104  indexing_pipe.connect("content_extractor.documents", "document_writer.documents")
105  
106  ## Run the indexing pipeline with sources
107  indexing_result = indexing_pipe.run(
108      data={"sources": ["textual_pdf.pdf", "non_textual_pdf.pdf"]},
109  )
110  
111  ## Inspect the documents
112  indexed_documents = document_store.filter_documents()
113  print(f"Indexed {len(indexed_documents)} documents:\n")
114  for doc in indexed_documents:
115      print("file_path: ", doc.meta["file_path"])
116      print("page_number: ", doc.meta["page_number"])
117      print("content: ", doc.content)
118      print("-" * 100 + "\n")
119  
120  ## Indexed 3 documents:
121  ##
122  ## file_path:  textual_pdf.pdf
123  ## page_number:  1
124  ## content:  A sample PDF file...
125  ## ----------------------------------------------------------------------------------------------------
126  ##
127  ## file_path:  textual_pdf.pdf
128  ## page_number:  2
129  ## content:  Page 2 of Sample PDF...
130  ## ----------------------------------------------------------------------------------------------------
131  ##
132  ## file_path:  non_textual_pdf.pdf
133  ## page_number:  1
134  ## content:  Content extracted from non-textual PDF using a LLM...
135  ## ----------------------------------------------------------------------------------------------------
136  ```