Index

Base schema for data structures.

BaseComponent

Bases: BaseModel

Base component object to capture class names.

Source code in llama-index-core/llama_index/core/schema.py

class BaseComponent(BaseModel):
    """Base component object to capture class names."""

    @classmethod
    def __get_pydantic_json_schema__(
        cls, core_schema: CoreSchema, handler: GetJsonSchemaHandler
    ) -> JsonSchemaValue:
        json_schema = super().__get_pydantic_json_schema__(core_schema, handler)
        if "properties" in json_schema:
            json_schema["properties"]["class_name"] = {
                "title": "Class Name",
                "type": "string",
                "default": cls.class_name(),
            }
        return json_schema

    @classmethod
    def class_name(cls) -> str:
        """
        Get the class name, used as a unique ID in serialization.

        This provides a key that makes serialization robust against actual class
        name changes.
        """
        return "base_component"

    def json(self, **kwargs: Any) -> str:
        return self.to_json(**kwargs)

    @model_serializer(mode="wrap")
    def custom_model_dump(self, handler: Any) -> Dict[str, Any]:
        data = handler(self)
        data["class_name"] = self.class_name()
        return data

    def dict(self, **kwargs: Any) -> Dict[str, Any]:
        return self.model_dump(**kwargs)

    def __getstate__(self) -> Dict[str, Any]:
        state = super().__getstate__()

        # remove attributes that are not pickleable -- kind of dangerous
        keys_to_remove = []
        for key, val in state["__dict__"].items():
            try:
                pickle.dumps(val)
            except Exception:
                keys_to_remove.append(key)

        for key in keys_to_remove:
            logging.warning(f"Removing unpickleable attribute {key}")
            del state["__dict__"][key]

        # remove private attributes if they aren't pickleable -- kind of dangerous
        keys_to_remove = []
        private_attrs = state.get("__pydantic_private__", None)
        if private_attrs:
            for key, val in state["__pydantic_private__"].items():
                try:
                    pickle.dumps(val)
                except Exception:
                    keys_to_remove.append(key)

            for key in keys_to_remove:
                logging.warning(f"Removing unpickleable private attribute {key}")
                del state["__pydantic_private__"][key]

        return state

    def __setstate__(self, state: Dict[str, Any]) -> None:
        # Use the __dict__ and __init__ method to set state
        # so that all variables initialize
        try:
            self.__init__(**state["__dict__"])  # type: ignore
        except Exception:
            # Fall back to the default __setstate__ method
            # This may not work if the class had unpickleable attributes
            super().__setstate__(state)

    def to_dict(self, **kwargs: Any) -> Dict[str, Any]:
        data = self.dict(**kwargs)
        data["class_name"] = self.class_name()
        return data

    def to_json(self, **kwargs: Any) -> str:
        data = self.to_dict(**kwargs)
        return json.dumps(data)

    # TODO: return type here not supported by current mypy version
    @classmethod
    def from_dict(cls, data: Dict[str, Any], **kwargs: Any) -> Self:  # type: ignore
        # In SimpleKVStore we rely on shallow coping. Hence, the data will be modified in the store directly.
        # And it is the same when the user is passing a dictionary to create a component. We can't modify the passed down dictionary.
        data = dict(data)
        if isinstance(kwargs, dict):
            data.update(kwargs)
        data.pop("class_name", None)
        return cls(**data)

    @classmethod
    def from_json(cls, data_str: str, **kwargs: Any) -> Self:  # type: ignore
        data = json.loads(data_str)
        return cls.from_dict(data, **kwargs)

class_name classmethod

class_name() -> str

Get the class name, used as a unique ID in serialization.

This provides a key that makes serialization robust against actual class name changes.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def class_name(cls) -> str:
    """
    Get the class name, used as a unique ID in serialization.

    This provides a key that makes serialization robust against actual class
    name changes.
    """
    return "base_component"

TransformComponent

Bases: BaseComponent, DispatcherSpanMixin

Base class for transform components.

Source code in llama-index-core/llama_index/core/schema.py

class TransformComponent(BaseComponent, DispatcherSpanMixin):
    """Base class for transform components."""

    model_config = ConfigDict(arbitrary_types_allowed=True)

    @abstractmethod
    def __call__(
        self, nodes: Sequence["BaseNode"], **kwargs: Any
    ) -> Sequence["BaseNode"]:
        """Transform nodes."""

    async def acall(
        self, nodes: Sequence["BaseNode"], **kwargs: Any
    ) -> Sequence["BaseNode"]:
        """Async transform nodes."""
        return self.__call__(nodes, **kwargs)

acall async

acall(nodes: Sequence[BaseNode], **kwargs: Any) -> Sequence[BaseNode]

Async transform nodes.

Source code in llama-index-core/llama_index/core/schema.py

async def acall(
    self, nodes: Sequence["BaseNode"], **kwargs: Any
) -> Sequence["BaseNode"]:
    """Async transform nodes."""
    return self.__call__(nodes, **kwargs)

NodeRelationship

Bases: str, Enum

Node relationships used in BaseNode class.

Attributes:

Name	Type	Description
SOURCE		The node is the source document.
PREVIOUS		The node is the previous node in the document.
NEXT		The node is the next node in the document.
PARENT		The node is the parent node in the document.
CHILD		The node is a child node in the document.

Source code in llama-index-core/llama_index/core/schema.py

class NodeRelationship(str, Enum):
    """Node relationships used in `BaseNode` class.

    Attributes:
        SOURCE: The node is the source document.
        PREVIOUS: The node is the previous node in the document.
        NEXT: The node is the next node in the document.
        PARENT: The node is the parent node in the document.
        CHILD: The node is a child node in the document.

    """

    SOURCE = auto()
    PREVIOUS = auto()
    NEXT = auto()
    PARENT = auto()
    CHILD = auto()

RelatedNodeInfo

Bases: BaseComponent

Parameters:

Name	Type	Description	Default
node_id	str		required
node_type	ObjectType | None		None
metadata	Dict[str, Any]		{}
hash	str | None		None

Source code in llama-index-core/llama_index/core/schema.py

class RelatedNodeInfo(BaseComponent):
    node_id: str
    node_type: Optional[ObjectType] = None
    metadata: Dict[str, Any] = Field(default_factory=dict)
    hash: Optional[str] = None

    @classmethod
    def class_name(cls) -> str:
        return "RelatedNodeInfo"

BaseNode

Bases: BaseComponent

Base node Object.

Generic abstract interface for retrievable nodes

Parameters:

Name	Type	Description	Default
id_	str	Unique ID of the node.	'c9893eb1-8c1a-4332-b84a-5e6e94802465'
embedding	List[float] | None	Embedding of the node.	None
metadata	Dict[str, Any]	A flat dictionary of metadata fields	{}
excluded_embed_metadata_keys	List[str]	Metadata keys that are excluded from text for the embed model.	[]
excluded_llm_metadata_keys	List[str]	Metadata keys that are excluded from text for the LLM.	[]
relationships	Dict[NodeRelationship, Union[RelatedNodeInfo, List[RelatedNodeInfo]]]	A mapping of relationships to other node information.	{}

Source code in llama-index-core/llama_index/core/schema.py

class BaseNode(BaseComponent):
    """Base node Object.

    Generic abstract interface for retrievable nodes

    """

    # hash is computed on local field, during the validation process
    model_config = ConfigDict(populate_by_name=True, validate_assignment=True)

    id_: str = Field(
        default_factory=lambda: str(uuid.uuid4()), description="Unique ID of the node."
    )
    embedding: Optional[List[float]] = Field(
        default=None, description="Embedding of the node."
    )

    """"
    metadata fields
    - injected as part of the text shown to LLMs as context
    - injected as part of the text for generating embeddings
    - used by vector DBs for metadata filtering

    """
    metadata: Dict[str, Any] = Field(
        default_factory=dict,
        description="A flat dictionary of metadata fields",
        alias="extra_info",
    )
    excluded_embed_metadata_keys: List[str] = Field(
        default_factory=list,
        description="Metadata keys that are excluded from text for the embed model.",
    )
    excluded_llm_metadata_keys: List[str] = Field(
        default_factory=list,
        description="Metadata keys that are excluded from text for the LLM.",
    )
    relationships: Dict[NodeRelationship, RelatedNodeType] = Field(
        default_factory=dict,
        description="A mapping of relationships to other node information.",
    )

    @classmethod
    @abstractmethod
    def get_type(cls) -> str:
        """Get Object type."""

    @abstractmethod
    def get_content(self, metadata_mode: MetadataMode = MetadataMode.ALL) -> str:
        """Get object content."""

    @abstractmethod
    def get_metadata_str(self, mode: MetadataMode = MetadataMode.ALL) -> str:
        """Metadata string."""

    @abstractmethod
    def set_content(self, value: Any) -> None:
        """Set the content of the node."""

    @property
    @abstractmethod
    def hash(self) -> str:
        """Get hash of node."""

    @property
    def node_id(self) -> str:
        return self.id_

    @node_id.setter
    def node_id(self, value: str) -> None:
        self.id_ = value

    @property
    def source_node(self) -> Optional[RelatedNodeInfo]:
        """Source object node.

        Extracted from the relationships field.

        """
        if NodeRelationship.SOURCE not in self.relationships:
            return None

        relation = self.relationships[NodeRelationship.SOURCE]
        if isinstance(relation, list):
            raise ValueError("Source object must be a single RelatedNodeInfo object")
        return relation

    @property
    def prev_node(self) -> Optional[RelatedNodeInfo]:
        """Prev node."""
        if NodeRelationship.PREVIOUS not in self.relationships:
            return None

        relation = self.relationships[NodeRelationship.PREVIOUS]
        if not isinstance(relation, RelatedNodeInfo):
            raise ValueError("Previous object must be a single RelatedNodeInfo object")
        return relation

    @property
    def next_node(self) -> Optional[RelatedNodeInfo]:
        """Next node."""
        if NodeRelationship.NEXT not in self.relationships:
            return None

        relation = self.relationships[NodeRelationship.NEXT]
        if not isinstance(relation, RelatedNodeInfo):
            raise ValueError("Next object must be a single RelatedNodeInfo object")
        return relation

    @property
    def parent_node(self) -> Optional[RelatedNodeInfo]:
        """Parent node."""
        if NodeRelationship.PARENT not in self.relationships:
            return None

        relation = self.relationships[NodeRelationship.PARENT]
        if not isinstance(relation, RelatedNodeInfo):
            raise ValueError("Parent object must be a single RelatedNodeInfo object")
        return relation

    @property
    def child_nodes(self) -> Optional[List[RelatedNodeInfo]]:
        """Child nodes."""
        if NodeRelationship.CHILD not in self.relationships:
            return None

        relation = self.relationships[NodeRelationship.CHILD]
        if not isinstance(relation, list):
            raise ValueError("Child objects must be a list of RelatedNodeInfo objects.")
        return relation

    @property
    def ref_doc_id(self) -> Optional[str]:
        """Deprecated: Get ref doc id."""
        source_node = self.source_node
        if source_node is None:
            return None
        return source_node.node_id

    @property
    def extra_info(self) -> Dict[str, Any]:
        """TODO: DEPRECATED: Extra info."""
        return self.metadata

    def __str__(self) -> str:
        source_text_truncated = truncate_text(
            self.get_content().strip(), TRUNCATE_LENGTH
        )
        source_text_wrapped = textwrap.fill(
            f"Text: {source_text_truncated}\n", width=WRAP_WIDTH
        )
        return f"Node ID: {self.node_id}\n{source_text_wrapped}"

    def get_embedding(self) -> List[float]:
        """Get embedding.

        Errors if embedding is None.

        """
        if self.embedding is None:
            raise ValueError("embedding not set.")
        return self.embedding

    def as_related_node_info(self) -> RelatedNodeInfo:
        """Get node as RelatedNodeInfo."""
        return RelatedNodeInfo(
            node_id=self.node_id,
            node_type=self.get_type(),
            metadata=self.metadata,
            hash=self.hash,
        )

embedding class-attribute instance-attribute

embedding: Optional[List[float]] = Field(default=None, description='Embedding of the node.')

" metadata fields - injected as part of the text shown to LLMs as context - injected as part of the text for generating embeddings - used by vector DBs for metadata filtering

hash abstractmethod property

hash: str

Get hash of node.

source_node property

source_node: Optional[RelatedNodeInfo]

Source object node.

Extracted from the relationships field.

prev_node property

prev_node: Optional[RelatedNodeInfo]

Prev node.

next_node property

next_node: Optional[RelatedNodeInfo]

Next node.

parent_node property

parent_node: Optional[RelatedNodeInfo]

Parent node.

child_nodes property

child_nodes: Optional[List[RelatedNodeInfo]]

Child nodes.

ref_doc_id property

ref_doc_id: Optional[str]

Deprecated: Get ref doc id.

extra_info property

extra_info: Dict[str, Any]

TODO: DEPRECATED: Extra info.

get_type abstractmethod classmethod

get_type() -> str

Get Object type.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
@abstractmethod
def get_type(cls) -> str:
    """Get Object type."""

get_content abstractmethod

get_content(metadata_mode: MetadataMode = ALL) -> str

Get object content.

Source code in llama-index-core/llama_index/core/schema.py

@abstractmethod
def get_content(self, metadata_mode: MetadataMode = MetadataMode.ALL) -> str:
    """Get object content."""

get_metadata_str abstractmethod

get_metadata_str(mode: MetadataMode = ALL) -> str

Metadata string.

Source code in llama-index-core/llama_index/core/schema.py

@abstractmethod
def get_metadata_str(self, mode: MetadataMode = MetadataMode.ALL) -> str:
    """Metadata string."""

set_content abstractmethod

set_content(value: Any) -> None

Set the content of the node.

Source code in llama-index-core/llama_index/core/schema.py

@abstractmethod
def set_content(self, value: Any) -> None:
    """Set the content of the node."""

get_embedding

get_embedding() -> List[float]

Get embedding.

Errors if embedding is None.

Source code in llama-index-core/llama_index/core/schema.py

def get_embedding(self) -> List[float]:
    """Get embedding.

    Errors if embedding is None.

    """
    if self.embedding is None:
        raise ValueError("embedding not set.")
    return self.embedding

as_related_node_info

as_related_node_info() -> RelatedNodeInfo

Get node as RelatedNodeInfo.

Source code in llama-index-core/llama_index/core/schema.py

def as_related_node_info(self) -> RelatedNodeInfo:
    """Get node as RelatedNodeInfo."""
    return RelatedNodeInfo(
        node_id=self.node_id,
        node_type=self.get_type(),
        metadata=self.metadata,
        hash=self.hash,
    )

TextNode

Bases: BaseNode

Parameters:

Name	Type	Description	Default
text	str	Text content of the node.	''
mimetype	str	MIME type of the node content.	'text/plain'
start_char_idx	int | None	Start char index of the node.	None
end_char_idx	int | None	End char index of the node.	None
text_template	str	Template for how text is formatted, with {content} and {metadata_str} placeholders.	'{metadata_str}\n\n{content}'
metadata_template	str	Template for how metadata is formatted, with {key} and {value} placeholders.	'{key}: {value}'
metadata_seperator	str	Separator between metadata fields when converting to string.	'\n'

Source code in llama-index-core/llama_index/core/schema.py

class TextNode(BaseNode):
    text: str = Field(default="", description="Text content of the node.")
    mimetype: str = Field(
        default="text/plain", description="MIME type of the node content."
    )
    start_char_idx: Optional[int] = Field(
        default=None, description="Start char index of the node."
    )
    end_char_idx: Optional[int] = Field(
        default=None, description="End char index of the node."
    )
    text_template: str = Field(
        default=DEFAULT_TEXT_NODE_TMPL,
        description=(
            "Template for how text is formatted, with {content} and "
            "{metadata_str} placeholders."
        ),
    )
    metadata_template: str = Field(
        default=DEFAULT_METADATA_TMPL,
        description=(
            "Template for how metadata is formatted, with {key} and "
            "{value} placeholders."
        ),
    )
    metadata_seperator: str = Field(
        default="\n",
        description="Separator between metadata fields when converting to string.",
    )

    @classmethod
    def class_name(cls) -> str:
        return "TextNode"

    @property
    def hash(self) -> str:
        doc_identity = str(self.text) + str(self.metadata)
        return str(sha256(doc_identity.encode("utf-8", "surrogatepass")).hexdigest())

    @classmethod
    def get_type(cls) -> str:
        """Get Object type."""
        return ObjectType.TEXT

    def get_content(self, metadata_mode: MetadataMode = MetadataMode.NONE) -> str:
        """Get object content."""
        metadata_str = self.get_metadata_str(mode=metadata_mode).strip()
        if not metadata_str:
            return self.text

        return self.text_template.format(
            content=self.text, metadata_str=metadata_str
        ).strip()

    def get_metadata_str(self, mode: MetadataMode = MetadataMode.ALL) -> str:
        """Metadata info string."""
        if mode == MetadataMode.NONE:
            return ""

        usable_metadata_keys = set(self.metadata.keys())
        if mode == MetadataMode.LLM:
            for key in self.excluded_llm_metadata_keys:
                if key in usable_metadata_keys:
                    usable_metadata_keys.remove(key)
        elif mode == MetadataMode.EMBED:
            for key in self.excluded_embed_metadata_keys:
                if key in usable_metadata_keys:
                    usable_metadata_keys.remove(key)

        return self.metadata_seperator.join(
            [
                self.metadata_template.format(key=key, value=str(value))
                for key, value in self.metadata.items()
                if key in usable_metadata_keys
            ]
        )

    def set_content(self, value: str) -> None:
        """Set the content of the node."""
        self.text = value

    def get_node_info(self) -> Dict[str, Any]:
        """Get node info."""
        return {"start": self.start_char_idx, "end": self.end_char_idx}

    def get_text(self) -> str:
        return self.get_content(metadata_mode=MetadataMode.NONE)

    @property
    def node_info(self) -> Dict[str, Any]:
        """Deprecated: Get node info."""
        return self.get_node_info()

node_info property

node_info: Dict[str, Any]

Deprecated: Get node info.

get_type classmethod

get_type() -> str

Get Object type.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def get_type(cls) -> str:
    """Get Object type."""
    return ObjectType.TEXT

get_content

get_content(metadata_mode: MetadataMode = NONE) -> str

Get object content.

Source code in llama-index-core/llama_index/core/schema.py

def get_content(self, metadata_mode: MetadataMode = MetadataMode.NONE) -> str:
    """Get object content."""
    metadata_str = self.get_metadata_str(mode=metadata_mode).strip()
    if not metadata_str:
        return self.text

    return self.text_template.format(
        content=self.text, metadata_str=metadata_str
    ).strip()

get_metadata_str

get_metadata_str(mode: MetadataMode = ALL) -> str

Metadata info string.

Source code in llama-index-core/llama_index/core/schema.py

def get_metadata_str(self, mode: MetadataMode = MetadataMode.ALL) -> str:
    """Metadata info string."""
    if mode == MetadataMode.NONE:
        return ""

    usable_metadata_keys = set(self.metadata.keys())
    if mode == MetadataMode.LLM:
        for key in self.excluded_llm_metadata_keys:
            if key in usable_metadata_keys:
                usable_metadata_keys.remove(key)
    elif mode == MetadataMode.EMBED:
        for key in self.excluded_embed_metadata_keys:
            if key in usable_metadata_keys:
                usable_metadata_keys.remove(key)

    return self.metadata_seperator.join(
        [
            self.metadata_template.format(key=key, value=str(value))
            for key, value in self.metadata.items()
            if key in usable_metadata_keys
        ]
    )

set_content

set_content(value: str) -> None

Set the content of the node.

Source code in llama-index-core/llama_index/core/schema.py

def set_content(self, value: str) -> None:
    """Set the content of the node."""
    self.text = value

get_node_info

get_node_info() -> Dict[str, Any]

Get node info.

Source code in llama-index-core/llama_index/core/schema.py

def get_node_info(self) -> Dict[str, Any]:
    """Get node info."""
    return {"start": self.start_char_idx, "end": self.end_char_idx}

ImageNode

Bases: TextNode

Node with image.

Parameters:

Name	Type	Description	Default
image	str | None		None
image_path	str | None		None
image_url	str | None		None
image_mimetype	str | None		None
text_embedding	List[float] | None	Text embedding of image node, if text field is filled out	None

Source code in llama-index-core/llama_index/core/schema.py

class ImageNode(TextNode):
    """Node with image."""

    # TODO: store reference instead of actual image
    # base64 encoded image str
    image: Optional[str] = None
    image_path: Optional[str] = None
    image_url: Optional[str] = None
    image_mimetype: Optional[str] = None
    text_embedding: Optional[List[float]] = Field(
        default=None,
        description="Text embedding of image node, if text field is filled out",
    )

    @classmethod
    def get_type(cls) -> str:
        return ObjectType.IMAGE

    @classmethod
    def class_name(cls) -> str:
        return "ImageNode"

    def resolve_image(self) -> ImageType:
        """Resolve an image such that PIL can read it."""
        if self.image is not None:
            import base64

            return BytesIO(base64.b64decode(self.image))
        elif self.image_path is not None:
            return self.image_path
        elif self.image_url is not None:
            # load image from URL
            import requests

            response = requests.get(self.image_url)
            return BytesIO(response.content)
        else:
            raise ValueError("No image found in node.")

    @property
    def hash(self) -> str:
        """Get hash of node."""
        # doc identity depends on if image, image_path, or image_url is set
        image_str = self.image or "None"
        image_path_str = self.image_path or "None"
        image_url_str = self.image_url or "None"
        image_text = self.text or "None"
        doc_identity = f"{image_str}-{image_path_str}-{image_url_str}-{image_text}"
        return str(sha256(doc_identity.encode("utf-8", "surrogatepass")).hexdigest())

hash property

hash: str

Get hash of node.

resolve_image

resolve_image() -> ImageType

Resolve an image such that PIL can read it.

Source code in llama-index-core/llama_index/core/schema.py

def resolve_image(self) -> ImageType:
    """Resolve an image such that PIL can read it."""
    if self.image is not None:
        import base64

        return BytesIO(base64.b64decode(self.image))
    elif self.image_path is not None:
        return self.image_path
    elif self.image_url is not None:
        # load image from URL
        import requests

        response = requests.get(self.image_url)
        return BytesIO(response.content)
    else:
        raise ValueError("No image found in node.")

IndexNode

Bases: TextNode

Node with reference to any object.

This can include other indices, query engines, retrievers.

This can also include other nodes (though this is overlapping with relationships on the Node class).

Parameters:

Name	Type	Description	Default
index_id	str		required
obj	Any		None

Source code in llama-index-core/llama_index/core/schema.py

class IndexNode(TextNode):
    """Node with reference to any object.

    This can include other indices, query engines, retrievers.

    This can also include other nodes (though this is overlapping with `relationships`
    on the Node class).

    """

    index_id: str
    obj: Any = None

    def dict(self, **kwargs: Any) -> Dict[str, Any]:
        from llama_index.core.storage.docstore.utils import doc_to_json

        data = super().dict(**kwargs)

        try:
            if self.obj is None:
                data["obj"] = None
            elif isinstance(self.obj, BaseNode):
                data["obj"] = doc_to_json(self.obj)
            elif isinstance(self.obj, BaseModel):
                data["obj"] = self.obj.model_dump()
            else:
                data["obj"] = json.dumps(self.obj)
        except Exception:
            raise ValueError("IndexNode obj is not serializable: " + str(self.obj))

        return data

    @classmethod
    def from_text_node(
        cls,
        node: TextNode,
        index_id: str,
    ) -> "IndexNode":
        """Create index node from text node."""
        # copy all attributes from text node, add index id
        return cls(
            **node.dict(),
            index_id=index_id,
        )

    # TODO: return type here not supported by current mypy version
    @classmethod
    def from_dict(cls, data: Dict[str, Any], **kwargs: Any) -> Self:  # type: ignore
        output = super().from_dict(data, **kwargs)

        obj = data.get("obj", None)
        parsed_obj = None

        if isinstance(obj, str):
            parsed_obj = TextNode(text=obj)
        elif isinstance(obj, dict):
            from llama_index.core.storage.docstore.utils import json_to_doc

            # check if its a node, else assume stringable
            try:
                parsed_obj = json_to_doc(obj)  # type: ignore[assignment]
            except Exception:
                parsed_obj = TextNode(text=str(obj))

        output.obj = parsed_obj

        return output

    @classmethod
    def get_type(cls) -> str:
        return ObjectType.INDEX

    @classmethod
    def class_name(cls) -> str:
        return "IndexNode"

from_text_node classmethod

from_text_node(node: TextNode, index_id: str) -> IndexNode

Create index node from text node.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def from_text_node(
    cls,
    node: TextNode,
    index_id: str,
) -> "IndexNode":
    """Create index node from text node."""
    # copy all attributes from text node, add index id
    return cls(
        **node.dict(),
        index_id=index_id,
    )

NodeWithScore

Bases: BaseComponent

Parameters:

Name	Type	Description	Default
node	BaseNode		required
score	float | None		None

Source code in llama-index-core/llama_index/core/schema.py

class NodeWithScore(BaseComponent):
    node: SerializeAsAny[BaseNode]
    score: Optional[float] = None

    def __str__(self) -> str:
        score_str = "None" if self.score is None else f"{self.score: 0.3f}"
        return f"{self.node}\nScore: {score_str}\n"

    def get_score(self, raise_error: bool = False) -> float:
        """Get score."""
        if self.score is None:
            if raise_error:
                raise ValueError("Score not set.")
            else:
                return 0.0
        else:
            return self.score

    @classmethod
    def class_name(cls) -> str:
        return "NodeWithScore"

    ##### pass through methods to BaseNode #####
    @property
    def node_id(self) -> str:
        return self.node.node_id

    @property
    def id_(self) -> str:
        return self.node.id_

    @property
    def text(self) -> str:
        if isinstance(self.node, TextNode):
            return self.node.text
        else:
            raise ValueError("Node must be a TextNode to get text.")

    @property
    def metadata(self) -> Dict[str, Any]:
        return self.node.metadata

    @property
    def embedding(self) -> Optional[List[float]]:
        return self.node.embedding

    def get_text(self) -> str:
        if isinstance(self.node, TextNode):
            return self.node.get_text()
        else:
            raise ValueError("Node must be a TextNode to get text.")

    def get_content(self, metadata_mode: MetadataMode = MetadataMode.NONE) -> str:
        return self.node.get_content(metadata_mode=metadata_mode)

    def get_embedding(self) -> List[float]:
        return self.node.get_embedding()

get_score

get_score(raise_error: bool = False) -> float

Get score.

Source code in llama-index-core/llama_index/core/schema.py

def get_score(self, raise_error: bool = False) -> float:
    """Get score."""
    if self.score is None:
        if raise_error:
            raise ValueError("Score not set.")
        else:
            return 0.0
    else:
        return self.score

Document

Bases: TextNode

Generic interface for a data document.

This document connects to data sources.

Parameters:

Name	Type	Description	Default
id_	str	Unique ID of the node.	'ac359e1b-365d-45f9-9bc0-58d726a4e225'

Source code in llama-index-core/llama_index/core/schema.py

class Document(TextNode):
    """Generic interface for a data document.

    This document connects to data sources.

    """

    # TODO: A lot of backwards compatibility logic here, clean up
    id_: str = Field(
        default_factory=lambda: str(uuid.uuid4()),
        description="Unique ID of the node.",
        alias="doc_id",
    )

    _compat_fields = {"doc_id": "id_", "extra_info": "metadata"}

    @classmethod
    def get_type(cls) -> str:
        """Get Document type."""
        return ObjectType.DOCUMENT

    @property
    def doc_id(self) -> str:
        """Get document ID."""
        return self.id_

    def __str__(self) -> str:
        source_text_truncated = truncate_text(
            self.get_content().strip(), TRUNCATE_LENGTH
        )
        source_text_wrapped = textwrap.fill(
            f"Text: {source_text_truncated}\n", width=WRAP_WIDTH
        )
        return f"Doc ID: {self.doc_id}\n{source_text_wrapped}"

    def get_doc_id(self) -> str:
        """TODO: Deprecated: Get document ID."""
        return self.id_

    def __setattr__(self, name: str, value: object) -> None:
        if name in self._compat_fields:
            name = self._compat_fields[name]
        super().__setattr__(name, value)

    def to_langchain_format(self) -> "LCDocument":
        """Convert struct to LangChain document format."""
        from llama_index.core.bridge.langchain import Document as LCDocument

        metadata = self.metadata or {}
        return LCDocument(page_content=self.text, metadata=metadata, id=self.id_)

    @classmethod
    def from_langchain_format(cls, doc: "LCDocument") -> "Document":
        """Convert struct from LangChain document format."""
        if doc.id:
            return cls(text=doc.page_content, metadata=doc.metadata, id_=doc.id)
        return cls(text=doc.page_content, metadata=doc.metadata)

    def to_haystack_format(self) -> "HaystackDocument":
        """Convert struct to Haystack document format."""
        from haystack.schema import Document as HaystackDocument

        return HaystackDocument(
            content=self.text, meta=self.metadata, embedding=self.embedding, id=self.id_
        )

    @classmethod
    def from_haystack_format(cls, doc: "HaystackDocument") -> "Document":
        """Convert struct from Haystack document format."""
        return cls(
            text=doc.content, metadata=doc.meta, embedding=doc.embedding, id_=doc.id
        )

    def to_embedchain_format(self) -> Dict[str, Any]:
        """Convert struct to EmbedChain document format."""
        return {
            "doc_id": self.id_,
            "data": {"content": self.text, "meta_data": self.metadata},
        }

    @classmethod
    def from_embedchain_format(cls, doc: Dict[str, Any]) -> "Document":
        """Convert struct from EmbedChain document format."""
        return cls(
            text=doc["data"]["content"],
            metadata=doc["data"]["meta_data"],
            id_=doc["doc_id"],
        )

    def to_semantic_kernel_format(self) -> "MemoryRecord":
        """Convert struct to Semantic Kernel document format."""
        import numpy as np
        from semantic_kernel.memory.memory_record import MemoryRecord

        return MemoryRecord(
            id=self.id_,
            text=self.text,
            additional_metadata=self.get_metadata_str(),
            embedding=np.array(self.embedding) if self.embedding else None,
        )

    @classmethod
    def from_semantic_kernel_format(cls, doc: "MemoryRecord") -> "Document":
        """Convert struct from Semantic Kernel document format."""
        return cls(
            text=doc._text,
            metadata={"additional_metadata": doc._additional_metadata},
            embedding=doc._embedding.tolist() if doc._embedding is not None else None,
            id_=doc._id,
        )

    def to_vectorflow(self, client: Any) -> None:
        """Send a document to vectorflow, since they don't have a document object."""
        # write document to temp file
        import tempfile

        with tempfile.NamedTemporaryFile() as f:
            f.write(self.text.encode("utf-8"))
            f.flush()
            client.embed(f.name)

    @classmethod
    def example(cls) -> "Document":
        return Document(
            text=SAMPLE_TEXT,
            metadata={"filename": "README.md", "category": "codebase"},
        )

    @classmethod
    def class_name(cls) -> str:
        return "Document"

    def to_cloud_document(self) -> "CloudDocument":
        """Convert to LlamaCloud document type."""
        from llama_cloud.types.cloud_document import CloudDocument

        return CloudDocument(
            text=self.text,
            metadata=self.metadata,
            excluded_embed_metadata_keys=self.excluded_embed_metadata_keys,
            excluded_llm_metadata_keys=self.excluded_llm_metadata_keys,
            id=self.id_,
        )

    @classmethod
    def from_cloud_document(
        cls,
        doc: "CloudDocument",
    ) -> "Document":
        """Convert from LlamaCloud document type."""
        return Document(
            text=doc.text,
            metadata=doc.metadata,
            excluded_embed_metadata_keys=doc.excluded_embed_metadata_keys,
            excluded_llm_metadata_keys=doc.excluded_llm_metadata_keys,
            id_=doc.id,
        )

doc_id property

doc_id: str

Get document ID.

get_type classmethod

get_type() -> str

Get Document type.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def get_type(cls) -> str:
    """Get Document type."""
    return ObjectType.DOCUMENT

get_doc_id

get_doc_id() -> str

TODO: Deprecated: Get document ID.

Source code in llama-index-core/llama_index/core/schema.py

def get_doc_id(self) -> str:
    """TODO: Deprecated: Get document ID."""
    return self.id_

to_langchain_format

to_langchain_format() -> Document

Convert struct to LangChain document format.

Source code in llama-index-core/llama_index/core/schema.py

def to_langchain_format(self) -> "LCDocument":
    """Convert struct to LangChain document format."""
    from llama_index.core.bridge.langchain import Document as LCDocument

    metadata = self.metadata or {}
    return LCDocument(page_content=self.text, metadata=metadata, id=self.id_)

from_langchain_format classmethod

from_langchain_format(doc: Document) -> Document

Convert struct from LangChain document format.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def from_langchain_format(cls, doc: "LCDocument") -> "Document":
    """Convert struct from LangChain document format."""
    if doc.id:
        return cls(text=doc.page_content, metadata=doc.metadata, id_=doc.id)
    return cls(text=doc.page_content, metadata=doc.metadata)

to_haystack_format

to_haystack_format() -> Document

Convert struct to Haystack document format.

Source code in llama-index-core/llama_index/core/schema.py

def to_haystack_format(self) -> "HaystackDocument":
    """Convert struct to Haystack document format."""
    from haystack.schema import Document as HaystackDocument

    return HaystackDocument(
        content=self.text, meta=self.metadata, embedding=self.embedding, id=self.id_
    )

from_haystack_format classmethod

from_haystack_format(doc: Document) -> Document

Convert struct from Haystack document format.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def from_haystack_format(cls, doc: "HaystackDocument") -> "Document":
    """Convert struct from Haystack document format."""
    return cls(
        text=doc.content, metadata=doc.meta, embedding=doc.embedding, id_=doc.id
    )

to_embedchain_format

to_embedchain_format() -> Dict[str, Any]

Convert struct to EmbedChain document format.

Source code in llama-index-core/llama_index/core/schema.py

def to_embedchain_format(self) -> Dict[str, Any]:
    """Convert struct to EmbedChain document format."""
    return {
        "doc_id": self.id_,
        "data": {"content": self.text, "meta_data": self.metadata},
    }

from_embedchain_format classmethod

from_embedchain_format(doc: Dict[str, Any]) -> Document

Convert struct from EmbedChain document format.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def from_embedchain_format(cls, doc: Dict[str, Any]) -> "Document":
    """Convert struct from EmbedChain document format."""
    return cls(
        text=doc["data"]["content"],
        metadata=doc["data"]["meta_data"],
        id_=doc["doc_id"],
    )

to_semantic_kernel_format

to_semantic_kernel_format() -> MemoryRecord

Convert struct to Semantic Kernel document format.

Source code in llama-index-core/llama_index/core/schema.py

def to_semantic_kernel_format(self) -> "MemoryRecord":
    """Convert struct to Semantic Kernel document format."""
    import numpy as np
    from semantic_kernel.memory.memory_record import MemoryRecord

    return MemoryRecord(
        id=self.id_,
        text=self.text,
        additional_metadata=self.get_metadata_str(),
        embedding=np.array(self.embedding) if self.embedding else None,
    )

from_semantic_kernel_format classmethod

from_semantic_kernel_format(doc: MemoryRecord) -> Document

Convert struct from Semantic Kernel document format.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def from_semantic_kernel_format(cls, doc: "MemoryRecord") -> "Document":
    """Convert struct from Semantic Kernel document format."""
    return cls(
        text=doc._text,
        metadata={"additional_metadata": doc._additional_metadata},
        embedding=doc._embedding.tolist() if doc._embedding is not None else None,
        id_=doc._id,
    )

to_vectorflow

to_vectorflow(client: Any) -> None

Send a document to vectorflow, since they don't have a document object.

Source code in llama-index-core/llama_index/core/schema.py

def to_vectorflow(self, client: Any) -> None:
    """Send a document to vectorflow, since they don't have a document object."""
    # write document to temp file
    import tempfile

    with tempfile.NamedTemporaryFile() as f:
        f.write(self.text.encode("utf-8"))
        f.flush()
        client.embed(f.name)

to_cloud_document

to_cloud_document() -> CloudDocument

Convert to LlamaCloud document type.

Source code in llama-index-core/llama_index/core/schema.py

def to_cloud_document(self) -> "CloudDocument":
    """Convert to LlamaCloud document type."""
    from llama_cloud.types.cloud_document import CloudDocument

    return CloudDocument(
        text=self.text,
        metadata=self.metadata,
        excluded_embed_metadata_keys=self.excluded_embed_metadata_keys,
        excluded_llm_metadata_keys=self.excluded_llm_metadata_keys,
        id=self.id_,
    )

from_cloud_document classmethod

from_cloud_document(doc: CloudDocument) -> Document

Convert from LlamaCloud document type.

Source code in llama-index-core/llama_index/core/schema.py

@classmethod
def from_cloud_document(
    cls,
    doc: "CloudDocument",
) -> "Document":
    """Convert from LlamaCloud document type."""
    return Document(
        text=doc.text,
        metadata=doc.metadata,
        excluded_embed_metadata_keys=doc.excluded_embed_metadata_keys,
        excluded_llm_metadata_keys=doc.excluded_llm_metadata_keys,
        id_=doc.id,
    )

ImageDocument

Bases: Document, ImageNode

Data document containing an image.

Source code in llama-index-core/llama_index/core/schema.py

class ImageDocument(Document, ImageNode):
    """Data document containing an image."""

    @classmethod
    def class_name(cls) -> str:
        return "ImageDocument"

QueryBundle dataclass

Bases: DataClassJsonMixin

Query bundle.

This dataclass contains the original query string and associated transformations.

Parameters:

Name	Type	Description	Default
query_str	str	the original user-specified query string. This is currently used by all non embedding-based queries.	required
custom_embedding_strs	list[str]	list of strings used for embedding the query. This is currently used by all embedding-based queries.	required
embedding	list[float]	the stored embedding for the query.	required
image_path	str | None		None

Source code in llama-index-core/llama_index/core/schema.py

@dataclass
class QueryBundle(DataClassJsonMixin):
    """
    Query bundle.

    This dataclass contains the original query string and associated transformations.

    Args:
        query_str (str): the original user-specified query string.
            This is currently used by all non embedding-based queries.
        custom_embedding_strs (list[str]): list of strings used for embedding the query.
            This is currently used by all embedding-based queries.
        embedding (list[float]): the stored embedding for the query.
    """

    query_str: str
    # using single image path as query input
    image_path: Optional[str] = None
    custom_embedding_strs: Optional[List[str]] = None
    embedding: Optional[List[float]] = None

    @property
    def embedding_strs(self) -> List[str]:
        """Use custom embedding strs if specified, otherwise use query str."""
        if self.custom_embedding_strs is None:
            if len(self.query_str) == 0:
                return []
            return [self.query_str]
        else:
            return self.custom_embedding_strs

    @property
    def embedding_image(self) -> List[ImageType]:
        """Use image path for image retrieval."""
        if self.image_path is None:
            return []
        return [self.image_path]

    def __str__(self) -> str:
        """Convert to string representation."""
        return self.query_str

embedding_strs property

embedding_strs: List[str]

Use custom embedding strs if specified, otherwise use query str.

embedding_image property

embedding_image: List[ImageType]

Use image path for image retrieval.