Long-term Memory

The short-term memory contains the context information about the agent current situations, while the long-term memory stores the agent past behaviors and thoughts, which can be retrieved according to the current events.

Long-term memory resembles the external vector storage that agents can rapidly query and retrieve from as needed.

In DB-GPT, the long-term memory stored in the vector storage by default.

Using Long-term Memory

To use long-term memory, you need to provide a vector store.

Prepare Embedding Model

First, you need to prepare an embedding model, which is used to convert the text into vectors. You can prepare the embedding model according Prepare Embedding Model.

Here we use the OpenAI Embedding API as an example:

import os
from dbgpt.rag.embedding import DefaultEmbeddingFactory

api_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") + "/embeddings"
api_key = os.getenv("OPENAI_API_KEY")
embeddings = DefaultEmbeddingFactory.openai(api_url=api_url, api_key=api_key)

Prepare Vector Store

Then you need to prepare a vector store, here we use the ChromaStore as an example,

Install the chroma package with the following command:

pip install chromadb

import shutil
from dbgpt.storage.vector_store.chroma_store import ChromaVectorConfig, ChromaStore

# Delete old vector store directory(/tmp/tmp_ltm_vector_stor)
shutil.rmtree("/tmp/tmp_ltm_vector_store", ignore_errors=True)
vector_store = ChromaStore(
    ChromaVectorConfig(
        embedding_fn=embeddings,
        vector_store_config=ChromaVectorConfig(
            name="ltm_vector_store",
            persist_path="/tmp/tmp_ltm_vector_store",
        ),
    )
)

Using Long-term Memory

from concurrent.futures import ThreadPoolExecutor
from dbgpt.agent import AgentMemory, LongTermMemory

# Create an agent memory, which contains a long-term memory
memory = LongTermMemory(
    executor=ThreadPoolExecutor(), vector_store=vector_store, _default_importance=0.5
)
agent_memory: AgentMemory = AgentMemory(memory=memory)

In above code, _default_importance means the default importance of one memory fragment, because we use LongTermMemory directly, so we need to set the default importance.