agents.memory_v2.standalone_memory_agent_free

Standalone memory agent using only free resources (no API keys required).

This implementation shows how to build a functional memory agent without relying on paid APIs like OpenAI or Anthropic.

Classes

FreeMemoryAgent

Memory agent using free embeddings and local storage.

Functions

test_free_memory_agent()

Test the free memory agent.

Module Contents

class agents.memory_v2.standalone_memory_agent_free.FreeMemoryAgent(user_id, storage_path=None, embedding_model='sentence-transformers/all-MiniLM-L6-v2', k_memories=5)

Memory agent using free embeddings and local storage.

This agent provides: - Memory storage with embeddings (using HuggingFace) - Similarity-based retrieval - Persistent storage to disk - No API keys required

Initialize the free memory agent.

Parameters:

  • user_id (str) – User identifier

  • storage_path (str | None) – Path to store memories (uses temp if None)

  • embedding_model (str) – HuggingFace model name for embeddings

  • k_memories (int) – Number of memories to retrieve

add_memory(content, memory_type=MemoryType.CONVERSATIONAL, importance=ImportanceLevel.MEDIUM, metadata=None)

Add a new memory.

Parameters:

  • content (str) – Memory content

  • memory_type (haive.agents.memory_v2.memory_state_original.MemoryType) – Type of memory

  • importance (haive.agents.memory_v2.memory_state_original.ImportanceLevel) – Importance level

  • metadata (dict[str, Any] | None) – Optional metadata

Returns:

Memory ID

Return type:

str

get_relevant_context(query, k=None)

Get relevant context for a query.

Parameters:

  • query (str) – Query to find context for

  • k (int | None) – Number of memories to include

Returns:

Formatted context string

Return type:

str

get_stats()

Get memory statistics.

Return type:

dict[str, Any]

async process_input(user_input)

Process user input - store if it’s information, retrieve if it’s a question.

Parameters:

user_input (str) – User’s input text

Returns:

Response string

Return type:

str

save()

Save the vector store to disk.

search_memories(query, k=None, memory_type=None, importance=None)

Search memories using similarity search.

Parameters:

  • query (str) – Search query

  • k (int | None) – Number of results (uses k_memories if None)

  • memory_type (haive.agents.memory_v2.memory_state_original.MemoryType | None) – Filter by memory type

  • importance (haive.agents.memory_v2.memory_state_original.ImportanceLevel | None) – Filter by importance

Returns:

List of memory results with scores

Return type:

list[dict[str, Any]]

async agents.memory_v2.standalone_memory_agent_free.test_free_memory_agent()

Test the free memory agent.