agents.memory_v2.message_document_converter

Message to Document Converter for Memory V2 System.

This module converts conversation messages into timestamped documents following LangChain patterns for long-term memory agents and graph construction.

Based on: https://python.langchain.com/docs/versions/migrating_memory/long_term_memory_agent/

Classes

ConversationDocumentBatch

Batch processor for converting entire conversations to documents.

MessageDocumentConverter

Converts conversation messages into timestamped documents for memory storage.

MessageMetadata

Enhanced metadata for message-based documents.

TimestampedDocument

Document with enhanced timestamp and metadata for memory retrieval.

Functions

create_document_index(documents)

Create searchable index of documents.

extract_documents_by_timeframe(documents[, hours_back])

Extract documents from specific timeframe.

sort_documents_by_relevance_and_time(documents[, ...])

Sort documents by relevance and recency.

Module Contents

class agents.memory_v2.message_document_converter.ConversationDocumentBatch(conversation_id=None, user_id=None)

Batch processor for converting entire conversations to documents.

Initialize batch processor.

Parameters:

  • conversation_id (str | None)

  • user_id (str | None)

process_conversation(messages, include_summary=True, chunk_size=5)

Process entire conversation into documents.

Parameters:

  • messages (list[langchain_core.messages.BaseMessage]) – Conversation messages

  • include_summary (bool) – Whether to create summary documents

  • chunk_size (int) – Size of message chunks for processing

Returns:

List of all generated documents

Return type:

list[TimestampedDocument]

class agents.memory_v2.message_document_converter.MessageDocumentConverter(conversation_id=None, user_id=None, session_id=None)

Converts conversation messages into timestamped documents for memory storage.

Initialize converter with context.

Parameters:

  • conversation_id (str | None)

  • user_id (str | None)

  • session_id (str | None)

convert_message(message)

Convert single message to timestamped document.

Parameters:

message (langchain_core.messages.BaseMessage) – LangChain message to convert

Returns:

TimestampedDocument with rich metadata

Return type:

TimestampedDocument

convert_messages(messages)

Convert multiple messages to timestamped documents.

Parameters:

messages (list[langchain_core.messages.BaseMessage]) – List of messages to convert

Returns:

List of TimestampedDocuments

Return type:

list[TimestampedDocument]

create_conversation_summary_document(messages, summary)

Create summary document from conversation.

Parameters:

  • messages (list[langchain_core.messages.BaseMessage]) – Original messages

  • summary (str) – Generated summary text

Returns:

TimestampedDocument containing the summary

Return type:

TimestampedDocument

create_memory_document(memory_item)

Convert memory item to timestamped document.

Parameters:

memory_item (agents.memory_v2.memory_state_original.EnhancedMemoryItem) – Memory item to convert

Returns:

TimestampedDocument for memory storage

Return type:

TimestampedDocument

class agents.memory_v2.message_document_converter.MessageMetadata(/, **data)

Bases: pydantic.BaseModel

Enhanced metadata for message-based documents.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

Parameters:

data (Any)

class agents.memory_v2.message_document_converter.TimestampedDocument(page_content, metadata=None)

Bases: langchain_core.documents.Document

Document with enhanced timestamp and metadata for memory retrieval.

Initialize with enhanced metadata.

Parameters:

  • page_content (str)

  • metadata (dict[str, Any] | None)

property age_days: float

Get document age in days.

Return type:

float

property age_hours: float

Get document age in hours.

Return type:

float

property timestamp: datetime.datetime

Get timestamp as datetime object.

Return type:

datetime.datetime

agents.memory_v2.message_document_converter.create_document_index(documents)

Create searchable index of documents.

Parameters:

documents (list[TimestampedDocument]) – Documents to index

Returns:

Index with metadata and statistics

Return type:

dict[str, Any]

agents.memory_v2.message_document_converter.extract_documents_by_timeframe(documents, hours_back=24)

Extract documents from specific timeframe.

Parameters:
Returns:

Filtered list of documents

Return type:

list[TimestampedDocument]

agents.memory_v2.message_document_converter.sort_documents_by_relevance_and_time(documents, time_weight=0.3, recency_decay=0.1)

Sort documents by relevance and recency.

Parameters:

  • documents (list[TimestampedDocument]) – Documents to sort

  • time_weight (float) – Weight given to recency (0.0 to 1.0)

  • recency_decay (float) – How quickly relevance decays with time

Returns:

Sorted documents (most relevant first)

Return type:

list[TimestampedDocument]