messages

Messages are objects used in prompts and chat conversations.

Class hierarchy:

BaseMessage --> SystemMessage, AIMessage, HumanMessage, ChatMessage, FunctionMessage, ToolMessage
            --> BaseMessageChunk --> SystemMessageChunk, AIMessageChunk, HumanMessageChunk, ChatMessageChunk, FunctionMessageChunk, ToolMessageChunk

Main helpers:

ChatPromptTemplate

Classes

messages.ai.AIMessage	Message from an AI.
messages.ai.AIMessageChunk	Message chunk from an AI.
messages.ai.InputTokenDetails	Breakdown of input token counts.
messages.ai.OutputTokenDetails	Breakdown of output token counts.
messages.ai.UsageMetadata	Usage metadata for a message, such as token counts.
messages.base.BaseMessage	Base abstract message class.
messages.base.BaseMessageChunk	Message chunk, which can be concatenated with other Message chunks.
messages.chat.ChatMessage	Message that can be assigned an arbitrary speaker (i.e. role).
messages.chat.ChatMessageChunk	Chat Message chunk.
messages.function.FunctionMessage	Message for passing the result of executing a tool back to a model.
messages.function.FunctionMessageChunk	Function Message chunk.
messages.human.HumanMessage	Message from a human.
messages.human.HumanMessageChunk	Human Message chunk.
messages.modifier.RemoveMessage	Message responsible for deleting other messages.
messages.system.SystemMessage	Message for priming AI behavior.
messages.system.SystemMessageChunk	System Message chunk.
messages.tool.InvalidToolCall	Allowance for errors made by LLM.
messages.tool.ToolCall	Represents a request to call a tool.
messages.tool.ToolCallChunk	A chunk of a tool call (e.g., as part of a stream).
messages.tool.ToolMessage	Message for passing the result of executing a tool back to a model.
messages.tool.ToolMessageChunk	Tool Message chunk.

Functions

messages.ai.add_ai_message_chunks(left, *others)	Add multiple AIMessageChunks together.
messages.ai.add_usage(left, right)	Recursively add two UsageMetadata objects.
messages.ai.subtract_usage(left, right)	Recursively subtract two UsageMetadata objects.
messages.base.get_msg_title_repr(title, *[, ...])	Get a title representation for a message.
messages.base.merge_content(first_content, ...)	Merge two message contents.
messages.base.message_to_dict(message)	Convert a Message to a dictionary.
messages.base.messages_to_dict(messages)	Convert a sequence of Messages to a list of dictionaries.
messages.tool.default_tool_chunk_parser(...)	Best-effort parsing of tool chunks.
messages.tool.default_tool_parser(raw_tool_calls)	Best-effort parsing of tools.
messages.tool.invalid_tool_call(*[, name, ...])
messages.tool.tool_call(*, name, args, id)
messages.tool.tool_call_chunk(*[, name, ...])
messages.utils.convert_to_messages(messages)	Convert a sequence of messages to a list of messages.
messages.utils.convert_to_openai_messages(...)	Convert LangChain messages into OpenAI message dicts.
messages.utils.filter_messages([messages])	Filter messages based on name, type or id.
messages.utils.get_buffer_string(messages[, ...])	Convert a sequence of Messages to strings and concatenate them into one string.
messages.utils.merge_message_runs([messages])	Merge consecutive Messages of the same type.
messages.utils.message_chunk_to_message(chunk)	Convert a message chunk to a message.
messages.utils.messages_from_dict(messages)	Convert a sequence of messages from dicts to Message objects.
messages.utils.trim_messages([messages])	Trim messages to be below a token count.