DeepsetPDFDocumentToBase64Image

Convert documents sources from PDF files to base64-encoded images.

Basic Information

• Type: deepset_cloud_custom_nodes.converters.pdf_to_image.DeepsetPDFDocumentToBase64Image
• Components it can connect with:
  • Any component that outputs a list of Document objects
  • Any component that accepts a list of Base64Image objects as input
  • DeepsetFileDownloader: DeepsetPDFDocumentToBase64Image can receive documents from DeepsetFileDownloader.
  • DeepsetAzureOpenAIVisionGenerator: This Generator can receive Base64Image objects to run visual question answering on them.

Inputs

Parameter	Type	Default	Description
documents	List[Document]		A list of documents with image information in their metadata. The expected metadata is: meta = {"file_path": str, "page_number": int} If this metadata is not present, the document is skipped and the component shows a warning.

Outputs

Parameter	Type	Default	Description
documents	List[Document]		A list of text documents.
base64_images	List[Base64Image]		A list of base64 encoded images corresponding to the documents they were converted from.

Overview

DeepsetPDFDocumentToBase64Image is a converter used in visual question answering pipelines to extract images from the downloaded PDFs. These images are then sent to a visual Generator that can process them. It converts documents accompanied by metadata containing the file_path and the page_number pointing to the location of the image.

Converting documents doesn't happen if:

• The file_path doesn't exist in the metadata.
• The page_number doesn't exist in the metadata.
• The file path doesn't start with the expected root path.
• The file path doesn't end with .pdf.

Usage Example

Initializing the Component

components:
  DeepsetPDFDocumentToBase64Image:
    type: converters.pdf_to_image.DeepsetPDFDocumentToBase64Image
    init_parameters:

Using the Component in a Pipeline

This component is used in our visual question answering templates, where it receives documents from DeepsetFileDownloader and sends them to DeepsetAzureOpenAIVisionGenerator.

This is how you connect the components in Builder:

And here's the complete pipeline YAML:

components:
  bm25_retriever: # Selects the most similar documents from the document store
    type: haystack_integrations.components.retrievers.opensearch.bm25_retriever.OpenSearchBM25Retriever
    init_parameters:
      document_store:
        type: haystack_integrations.document_stores.opensearch.document_store.OpenSearchDocumentStore
        init_parameters:
          use_ssl: True
          verify_certs: False
          hosts:
            - ${OPENSEARCH_HOST}
          http_auth:
            - "${OPENSEARCH_USER}"
            - "${OPENSEARCH_PASSWORD}"
          embedding_dim: 1024
          similarity: cosine
      top_k: 20 # The number of results to return

  query_embedder:
    type: haystack.components.embedders.sentence_transformers_text_embedder.SentenceTransformersTextEmbedder
    init_parameters:
      model: "BAAI/bge-m3"
      tokenizer_kwargs:
        model_max_length: 1024

  embedding_retriever: # Selects the most similar documents from the document store
    type: haystack_integrations.components.retrievers.opensearch.embedding_retriever.OpenSearchEmbeddingRetriever
    init_parameters:
      document_store:
        type: haystack_integrations.document_stores.opensearch.document_store.OpenSearchDocumentStore
        init_parameters:
          use_ssl: True
          verify_certs: False
          hosts:
            - ${OPENSEARCH_HOST}
          http_auth:
            - "${OPENSEARCH_USER}"
            - "${OPENSEARCH_PASSWORD}"
          embedding_dim: 1024
          similarity: cosine
      top_k: 20 # The number of results to return

  document_joiner:
    type: haystack.components.joiners.document_joiner.DocumentJoiner
    init_parameters:
      join_mode: concatenate

  ranker:
    type: haystack.components.rankers.transformers_similarity.TransformersSimilarityRanker
    init_parameters:
      model: "BAAI/bge-reranker-v2-m3"
      top_k: 5
      model_kwargs:
        torch_dtype: "torch.float16"
      tokenizer_kwargs:
        model_max_length: 1024

  image_downloader:
    type: deepset_cloud_custom_nodes.augmenters.deepset_file_downloader.DeepsetFileDownloader
    init_parameters:
      file_extensions:
        - ".pdf"

  pdf_to_image:
    type: deepset_cloud_custom_nodes.converters.pdf_to_image.DeepsetPDFDocumentToBase64Image
    init_parameters:
      detail: "high"

  prompt_builder:
    type: haystack.components.builders.prompt_builder.PromptBuilder
    init_parameters:
      template: |-
        Answer the question briefly and precisely based on the pictures.
        Give reasons for your answer.
        When answering the question only provide references within the answer text.
        Only use references in the form [NUMBER OF IMAGE] if you are using information from a image. 
        For example, if the first image is used in the answer add [1] and if the second image is used then use [2], etc. 
        Never name the images, but always enter a number in square brackets as a reference. 
        Question: {{ question }}
        Answer: 

  llm:
    type: deepset_cloud_custom_nodes.generators.openai_vision.DeepsetOpenAIVisionGenerator
    init_parameters:
      api_key: {"type": "env_var", "env_vars": ["OPENAI_API_KEY"], "strict": False}
      model: "gpt-4o"
      generation_kwargs:
        max_tokens: 650
        temperature: 0.0
        seed: 0

  answer_builder:
    type: deepset_cloud_custom_nodes.augmenters.deepset_answer_builder.DeepsetAnswerBuilder
    init_parameters:
      reference_pattern: acm

connections:  # Defines how the components are connected
- sender: bm25_retriever.documents
  receiver: document_joiner.documents
- sender: query_embedder.embedding
  receiver: embedding_retriever.query_embedding
- sender: embedding_retriever.documents
  receiver: document_joiner.documents
- sender: document_joiner.documents
  receiver: ranker.documents
- sender: ranker.documents
  receiver: image_downloader.documents
- sender: image_downloader.documents
  receiver: pdf_to_image.documents
- sender: pdf_to_image.base64_images
  receiver: llm.images
- sender: prompt_builder.prompt
  receiver: llm.prompt
- sender: image_downloader.documents
  receiver: answer_builder.documents
- sender: prompt_builder.prompt
  receiver: answer_builder.prompt
- sender: llm.replies
  receiver: answer_builder.replies

inputs:  # Define the inputs for your pipeline
  query:  # These components will receive the query as input
  - "bm25_retriever.query"
  - "query_embedder.text"
  - "ranker.query"
  - "prompt_builder.question"
  - "answer_builder.query"

  filters:  # These components will receive a potential query filter as input
  - "bm25_retriever.filters"
  - "embedding_retriever.filters"

outputs:  # Defines the output of your pipeline
  documents: "pdf_to_image.documents"  # The output of the pipeline is the retrieved documents
  answers: "answer_builder.answers"  # The output of the pipeline is the generated answers

Parameters

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
detail	Literal['auto', 'low', 'high']	auto	Controls how the model processes the image and generates its textual understanding. By default, the model uses the auto setting which looks at the image input size and decides if it should use the low or high setting. See OpenAI documentation
missing_page_number	Literal['skip', 'all_pages']	skip	Controls how to handle documents that do not have a page_number in their metadata. - skip: Skip documents that do not have a page_number in their metadata. - all_pages: If the page_number is not present, extract images from all pages of the PDF.

Run Method Parameters

These are the parameters you can configure for the component's run() method. This means you can pass these parameters at query time through the API, in Playground, or when running a job. For details, see Modify Pipeline Parameters at Query Time.

Parameter	Type	Default	Description
documents	List[Document]		A list of documents with image information in their metadata. The expected metadata is: meta = {"file_path": str, "page_number": int} If this metadata is not present, the document is skipped and the component shows a warning.