agentscope.message¶

The message module in agentscope.

class TextBlock[source]¶

Bases: TypedDict

The text block.

type: Required[Literal['text']]¶: The type of the block

text: str¶: The text content

class ThinkingBlock[source]¶

Bases: TypedDict

The thinking block.

type: Required[Literal['thinking']]¶: The type of the block

thinking: str¶

class Base64Source[source]¶

Bases: TypedDict

The base64 source

type: Required[Literal['base64']]¶: The type of the src, must be base64

media_type: Required[str]¶: The media type of the data, e.g. image/jpeg or audio/mpeg

data: Required[str]¶: The base64 data, in format of RFC 2397

class URLSource[source]¶

Bases: TypedDict

The URL source

type: Required[Literal['url']]¶: The type of the src, must be url

url: Required[str]¶: The URL of the image or audio

class ImageBlock[source]¶

Bases: TypedDict

The image block

type: Required[Literal['image']]¶: The type of the block

source: Required[Base64Source | URLSource]¶: The src of the image

class AudioBlock[source]¶

Bases: TypedDict

The audio block

type: Required[Literal['audio']]¶: The type of the block

source: Required[Base64Source | URLSource]¶: The src of the audio

class VideoBlock[source]¶

Bases: TypedDict

The video block

type: Required[Literal['video']]¶: The type of the block

source: Required[Base64Source | URLSource]¶: The src of the audio

class ToolUseBlock[source]¶

Bases: TypedDict

The tool use block.

type: Required[Literal['tool_use']]¶: The type of the block, must be tool_use

id: Required[str]¶: The identity of the tool call

name: Required[str]¶: The name of the tool

input: Required[dict[str, object]]¶: The input of the tool

class ToolResultBlock[source]¶

Bases: TypedDict

The tool result block.

type: Required[Literal['tool_result']]¶: The type of the block

id: Required[str]¶: The identity of the tool call result

output: Required[str | List[TextBlock | ImageBlock | AudioBlock | VideoBlock]]¶: The output of the tool function

name: Required[str]¶: The name of the tool function

class Msg[source]¶

Bases: object

The message class in agentscope.

__init__(name, content, role, metadata=None, timestamp=None, invocation_id=None)[source]¶

Initialize the Msg object.

Parameters:

name (str) – The name of the message sender.
content (str | list[ContentBlock]) – The content of the message.
role (Literal[“user”, “assistant”, “system”]) – The role of the message sender.
metadata (dict[str, JSONSerializableObject] | None, optional) – The metadata of the message, e.g. structured output.
timestamp (str | None, optional) – The created timestamp of the message. If not given, the timestamp will be set automatically.
invocation_id (str | None, optional) – The related API invocation id, if any. This is useful for tracking the message in the context of an API call.

Return type:

None

to_dict()[source]¶

Convert the message into JSON dict data.

Return type:: dict

classmethod from_dict(json_data)[source]¶

Load a message object from the given JSON data.

Parameters:: json_data (dict)
Return type:: Msg

has_content_blocks(block_type=None)[source]¶

Check if the message has content blocks of the given type.

Parameters:: block_type (Literal["text", "tool_use", "tool_result", "image", "audio", "video"] | None, defaults to None) – The type of the block to be checked. If None, it will check if there are any content blocks.
Return type:: bool

get_text_content(separator='\n')[source]¶

Get the pure text blocks from the message content.

Args:
separator (str, defaults to `

`):

The separator to use when concatenating multiple text blocks. Defaults to newline character.

Returns:

str | None:: The concatenated text content, or None if there is no text content.

Parameters:: separator (str)
Return type:: str | None

get_content_blocks(block_type: Literal['text']) → Sequence[TextBlock][source]¶

get_content_blocks(block_type: Literal['tool_use']) → Sequence[ToolUseBlock]

get_content_blocks(block_type: Literal['tool_result']) → Sequence[ToolResultBlock]

get_content_blocks(block_type: Literal['image']) → Sequence[ImageBlock]

get_content_blocks(block_type: Literal['audio']) → Sequence[AudioBlock]

get_content_blocks(block_type: Literal['video']) → Sequence[VideoBlock]

Get the content in block format. If the content is a string, it will be converted to a text block.

Parameters:: block_type (ContentBlockTypes | List[ContentBlockTypes] | None, optional) – The type of the block to be extracted. If None, all blocks will be returned.
Returns:: The content blocks.
Return type:: List[ContentBlock]