Chat

ChatHistory

The ChatHistory class is where all the conversation between the user and the LLM gets stored. A ChatHistory object will transparently be instanciated in the Brain every time you create one.

At each interaction with Brain.ask_streaming both your message and the LLM's response are added to this chat history. It's super handy because this history is used in the Retrieval-Augmented Generation (RAG) process to give the LLM more context, working as form of memory between the user and the system and helping it generate better responses by looking at what’s already been said.

You can also get some cool info about the brain by printing its details with the print_info() method, which shows things like how many chats are stored, the current chat history, and more. This makes it easy to keep track of what’s going on in your conversations and manage the context being sent to the LLM!

ChatHistory

ChatHistory is a class that maintains a record of chat conversations. Each message in the history is represented by an instance of the ChatMessage class, and the chat history is stored internally as a list of these ChatMessage objects. The class provides methods to retrieve, append, iterate, and manipulate the chat history, as well as utilities to convert the messages into specific formats and support deep copying.

__init__

__init__(chat_id, brain_id)

Init a new ChatHistory object.

Parameters:

Name	Type	Description	Default
chat_id	UUID	A unique identifier for the chat session.	required
brain_id	UUID | None	An optional identifier for the brain associated with the chat.	required

append

append(langchain_msg, metadata={})

Appends a new message to the chat history.

Parameters:

Name	Type	Description	Default
langchain_msg	AIMessage | HumanMessage	The message content (either an AI or Human message).	required
metadata	dict[str, Any]	Additional metadata related to the message. Defaults to an empty dictionary.	{}

get_chat_history

get_chat_history(newest_first=False)

Retrieves the chat history, optionally sorted in reverse chronological order.

Parameters:

Name	Type	Description	Default
newest_first	bool	If True, returns the messages in reverse order (newest first). Defaults to False.	False

Returns:

Type	Description
List[ChatMessage]	List[ChatMessage]: A sorted list of chat messages.

iter_pairs

iter_pairs()

Iterates over the chat history in pairs, returning a HumanMessage followed by an AIMessage.

Yields:

Type	Description
HumanMessage	Tuple[HumanMessage, AIMessage]: Pairs of human and AI messages.

Raises:

Type	Description
AssertionError	If the messages in the pair are not in the expected order (i.e., a HumanMessage followed by an AIMessage).

to_list

to_list()

Converts the chat history into a list of raw HumanMessage or AIMessage objects.

Returns:

Type	Description
List[HumanMessage | AIMessage]	list[HumanMessage | AIMessage]: A list of messages in their raw form, without metadata.

options: heading_level: 2