Skip to main content

Using as a Library

System Requirements​

  • Python >= 3.9

Quick Start​

  1. Install dependencies

    pip install document_parser

  2. Configuration

    parser_config = ParserConfig(image_provider=ImageStorageProvider(),
                                 ocr_model_name="gtp-4o",
                                 # Whether to enable OCR capability
                                 # If not enabled, vision_model_provider or vision_model_list doesn't need to be implemented or configured
                                 ocr_enable=True, 
                                 vision_model_provider=OpenAIVisionModelProvider())
    parser_context.register_all_config(parser_config)
    parser_context.register_user("userId") # User ID when requesting models, if not set it will affect OCR usage

  3. Execute parsing

    converter = Converter(stream=stream) # Pass in as file stream
    dom_tree = converter.dom_tree_parse( 
        remove_watermark=True,   # Whether to enable watermark removal
        parse_stream_table=False # Whether to parse streaming tables
    )

  4. Use standard domtree [Recommended]

    dom_tree_json = jsonable_encoder(dom_tree)
    standard_dom_tree = StandardDomTree.from_domtree_dict(dom_tree_json, file_info = file_info)
    json_compatible_data = jsonable_encoder(standard_dom_tree.root)
    print(json.dumps(json_compatible_data, ensure_ascii=False))

The standard domtree is a structurally and semantically more complete domtree structure. Services like bella-rag are based on this standard domtree for processing. With future iterations, the results from step 3 parsing execution will also output as standard domtree.