agent_with_long_term_memory

Create Agents with memory

In this folder, we provide an example of travel agent using Amazon Bedrock Agents new capabilities for memory.

When creating you agent, you can enable the memory capabilities using the memoryConfiguration parameter. You can then invoke you agent with a memoryId to track sessions and summarize them to your memory.

In this example we will create a test agent with the following architecture:

![Agent architecture](./images/architecture.png

The code below shows how to configure the memory capabilities when creating your agent using the create_agent function from boto3 SDK. You should define the enabledMemoryTypes (currently only SESSION_SUMMARY is available) and the storageDays.

    response = bedrock_agent_client.create_agent(
        agentName=agent_name,
        agentResourceRoleArn=agent_role['Role']['Arn'],
        description=agent_description,
        idleSessionTTLInSeconds=1800,
        foundationModel=agent_foundation_model,
        instruction=agent_instruction,
        memoryConfiguration={
            "enabledMemoryTypes": ["SESSION_SUMMARY"],
            "storageDays": 30
        }
    )

When invoking your agent, you need to pass the memoryId parameter as well as a sessionId.

    # invoke the agent API
    agent_response = bedrock_agent_runtime_client.invoke_agent(
        inputText="User query to the agent",
        agentId="<AGENT_ID>",
        agentAliasId="<AGENT_ALIAS_ID>",
        sessionId="<SESSION_ID>",
        enableTrace=True or False,
        endSession=True or False,
        memoryId="Your memory id",
        ...
    )

Once you invoke the agent with the endSession flag set to True, the conversation with the same sessionId is summarized and made available to the agent's memory. You can use the get_agent_memory from the bedrock-agent-runtime boto3 SDK client to get the available memory for an agent

    memory_content=bedrock_agent_runtime_client.get_agent_memory(
        agentAliasId="<AGENT_ALIAS_ID>",
        agentId="<AGENT_ID>",
        memoryId="<Same memory id used during the invoke agent calls",
        memoryType='SESSION_SUMMARY'
    )

And the delete_agent_memory to delete the full agent memory for a certain memory id.

    response = bedrock_agent_runtime_client.delete_agent_memory(
        agentAliasId="<AGENT_ALIAS_ID>", 
        agentId="<AGENT_ID>",
        memoryId="<Same memory id used during the invoke agent calls",
    )