MemoryManager

MemoryManager stores useful information from conversations and searches it later using semantic retrieval.

Create a memory manager

import akasha

memory = akasha.MemoryManager(
    memory_name="assistant",
    model="gemini:gemini-2.5-flash",
    embeddings="gemini:gemini-embedding-001",
    memory_dirname="docs",
)

The memory files are stored below memory_dirname / memory_name.

Add and search memory

memory.add_memory(
    user_prompt="I prefer short technical explanations.",
    ai_response="I will keep future explanations concise.",
    language="en",
)

matches = memory.search_memory("What is my explanation preference?", top_k=3)
for item in matches:
    print(item)

Useful methods:

Method	Purpose
add_memory(user_prompt, ai_response, language="ch")	Extract and store salient information from a conversation turn.
search_memory(query, top_k=3)	Return memories relevant to a query.
show_memory(num=100)	Return stored memory entries for inspection.

Memory storage creates or updates local files and a vector store. Choose the memory directory deliberately and exclude private memory data from public repositories.