agents.memory_v2.standalone_rag_memory¶

Standalone RAG-based Memory Agent using BaseRAGAgent with retrievers.

This module provides memory-capable agents built on BaseRAGAgent without dependencies on the broken memory module. All models are defined here.

Key features: 1. ConversationMemoryAgent - conversation history with time-weighting 2. FactualMemoryAgent - factual information storage and retrieval 3. PreferencesMemoryAgent - user preferences with generation 4. UnifiedMemoryRAGAgent - coordinates all memory types 5. Real BaseRAGAgent integration with different retrievers 6. Vector store persistence across different backends

NO MOCKS - All components use real BaseRAGAgent with real retrievers.

Classes¶

`ConversationMemoryAgent`	Memory agent for conversation history using BaseRAGAgent.
`FactualMemoryAgent`	Memory agent for factual information using BaseRAGAgent.
`ImportanceLevel`	Importance levels for memories.
`MemoryRAGConfig`	Configuration for RAG-based memory agents.
`MemoryType`	Types of memories that can be stored.
`MessageDocumentConverter`	Convert messages to documents for RAG storage.
`PreferencesMemoryAgent`	Memory agent for user preferences using SimpleRAGAgent for generation.
`StandaloneMemoryItem`	Standalone memory item for RAG-based storage.
`UnifiedMemoryRAGAgent`	Unified memory agent coordinating multiple specialized memory agents.

Functions¶

`create_conversation_memory_agent`([...])	Factory function to create conversation memory agent.
`create_unified_memory_agent`([user_id, llm_config, ...])	Factory function to create unified memory agent.
`demo`()	Demo.

Module Contents¶

class agents.memory_v2.standalone_rag_memory.ConversationMemoryAgent(config, name='conversation_memory')¶

Memory agent for conversation history using BaseRAGAgent.

Initialize conversation memory agent.

Parameters:

config (MemoryRAGConfig)
name (str)

async add_conversation(messages)¶

Add conversation messages to memory.

Parameters:: messages (list[langchain_core.messages.BaseMessage])
Return type:: None

async initialize()¶

Initialize the underlying RAG agent.

Return type:: None

async retrieve_conversation_context(query, k=None)¶

Retrieve relevant conversation context.

Parameters:

query (str)
k (int | None)

Return type:

list[langchain_core.documents.Document]

class agents.memory_v2.standalone_rag_memory.FactualMemoryAgent(config, name='factual_memory')¶

Memory agent for factual information using BaseRAGAgent.

Initialize factual memory agent.

Parameters:

config (MemoryRAGConfig)
name (str)

async add_memories(memories)¶

Add multiple factual memories.

Parameters:: memories (list[StandaloneMemoryItem])
Return type:: None

async add_memory(memory)¶

Add a factual memory.

Parameters:: memory (StandaloneMemoryItem)
Return type:: None

async initialize()¶

Initialize the underlying RAG agent.

Return type:: None

async retrieve_facts(query, k=None)¶

Retrieve relevant factual memories.

Parameters:

query (str)
k (int | None)

Return type:

list[dict[str, Any]]

class agents.memory_v2.standalone_rag_memory.ImportanceLevel¶

Bases: str, enum.Enum

Importance levels for memories.

Initialize self. See help(type(self)) for accurate signature.

class agents.memory_v2.standalone_rag_memory.MemoryRAGConfig(/, **data)¶

Bases: pydantic.BaseModel

Configuration for RAG-based memory agents.

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)

model_config¶: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

class agents.memory_v2.standalone_rag_memory.MemoryType¶

Bases: str, enum.Enum

Types of memories that can be stored.

Initialize self. See help(type(self)) for accurate signature.

class agents.memory_v2.standalone_rag_memory.MessageDocumentConverter(user_id=None, conversation_id=None)¶

Convert messages to documents for RAG storage.

Initialize converter.

Parameters:

user_id (str | None)
conversation_id (str | None)

convert_message(message)¶

Convert single message to document.

Parameters:: message (langchain_core.messages.BaseMessage)
Return type:: langchain_core.documents.Document

convert_messages(messages)¶

Convert multiple messages to documents.

Parameters:: messages (list[langchain_core.messages.BaseMessage])
Return type:: list[langchain_core.documents.Document]

class agents.memory_v2.standalone_rag_memory.PreferencesMemoryAgent(config, name='preferences_memory')¶

Memory agent for user preferences using SimpleRAGAgent for generation.

Initialize preferences memory agent.

Parameters:

config (MemoryRAGConfig)
name (str)

async add_preference(preference)¶

Add a user preference.

Parameters:: preference (StandaloneMemoryItem)
Return type:: None

async get_preferences_for(context)¶

Get relevant preferences and generate summary.

Parameters:: context (str)
Return type:: str

async initialize()¶

Initialize the underlying RAG agent.

Return type:: None

class agents.memory_v2.standalone_rag_memory.StandaloneMemoryItem(/, **data)¶

Bases: pydantic.BaseModel

Standalone memory item for RAG-based storage.

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)

mark_accessed()¶: Mark memory as accessed.

property age_hours: float¶

Get memory age in hours.

Return type:: float

model_config¶: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

class agents.memory_v2.standalone_rag_memory.UnifiedMemoryRAGAgent(config, user_id=None)¶

Unified memory agent coordinating multiple specialized memory agents.

Initialize unified memory agent.

Parameters:

config (MemoryRAGConfig)
user_id (str | None)

classmethod as_tool(name=None, description=None, **config_kwargs)¶

Convert this agent to a tool for use in other agents.

Parameters:

name (str | None)
description (str | None)

async get_memory_summary()¶

Get summary of all stored memories.

Return type:: dict[str, Any]

async initialize()¶

Initialize all memory agents.

Return type:: None

async process_conversation(messages)¶

Process conversation and extract memories.

Parameters:: messages (list[langchain_core.messages.BaseMessage])
Return type:: dict[str, Any]

async retrieve_context(query, memory_types=None)¶

Retrieve relevant context from all memory types.

Parameters:

query (str)
memory_types (list[str] | None)

Return type:

dict[str, Any]

agents.memory_v2.standalone_rag_memory.create_conversation_memory_agent(vector_store_provider=VectorStoreProvider.FAISS, embedding_model='sentence-transformers/all-mpnet-base-v2', enable_time_weighting=True, name='conversation_memory')¶

Factory function to create conversation memory agent.

Parameters:

vector_store_provider (haive.core.engine.vectorstore.VectorStoreProvider)
embedding_model (str)
enable_time_weighting (bool)
name (str)

Return type:

ConversationMemoryAgent

agents.memory_v2.standalone_rag_memory.create_unified_memory_agent(user_id=None, llm_config=None, vector_store_provider=VectorStoreProvider.FAISS, embedding_model='sentence-transformers/all-mpnet-base-v2')¶

Factory function to create unified memory agent.

Parameters:

user_id (str | None)
llm_config (haive.core.engine.aug_llm.AugLLMConfig)
vector_store_provider (haive.core.engine.vectorstore.VectorStoreProvider)
embedding_model (str)

Return type:

UnifiedMemoryRAGAgent

async agents.memory_v2.standalone_rag_memory.demo()¶: Demo.