Method to convert ParsedDocument object to LlamaIndex Document object

[mjspeck]

Description

It would be great to have, in addition to the to_llama_index_nodes method to have a to_llama_index_document method on the openparse.schemas.ParsedDocument class that returns a valid llama_index.core.schema.Document object.

[Filimoa]

Can you point me to documentation that explains how Nodes and Documents are related in llama_index? From what I understand a Document is just a parent Node.

This is the current implementation.

    def to_llama_index_nodes(self):
        try:
            from llama_index.core.schema import Document as LlamaIndexDocument
        except ImportError as err:
            raise ImportError(
                "llama_index is not installed. Please install it with `pip install llama-index`."
            ) from err

        li_doc = LlamaIndexDocument(
            id_=self.id_,
            metadata={
                "file_name": self.filename,
                "file_size": self.file_size,
                "creation_date": self.creation_date.isoformat(),
                "last_modified_date": self.last_modified_date.isoformat(),
            },
            excluded_embed_metadata_keys=[
                "file_size",
                "creation_date",
                "last_modified_date",
            ],
            excluded_llm_metadata_keys=[
                "file_name",
                "file_size",
                "creation_date",
                "last_modified_date",
            ],
        )
        li_nodes = self._nodes_to_llama_index(li_doc)

        return li_nodes

    def _nodes_to_llama_index(self, llama_index_doc):
        try:
            from llama_index.core.schema import NodeRelationship
        except ImportError as err:
            raise ImportError(
                "llama_index is not installed. Please install it with `pip install llama-index`."
            ) from err

        li_nodes = [node.to_llama_index() for node in sorted(self.nodes)]
        for i in range(len(li_nodes) - 1):
            li_nodes[i].relationships[NodeRelationship.NEXT] = li_nodes[
                i + 1
            ].as_related_node_info()

            li_nodes[i + 1].relationships[NodeRelationship.PREVIOUS] = li_nodes[
                i
            ].as_related_node_info()

        for li_node in li_nodes:
            li_node.relationships[NodeRelationship.PARENT] = (
                llama_index_doc.as_related_node_info()
            ) # NOTE: A DOC IS JUST A NODE?

        return li_nodes