Docs

Volcano Engine Function as a Service

Long-term Memory Agent

Introduction to Long-term Memory Agent

This document introduces the Long-term Memory Agent provided by Cloud Search.

Application introduction

Long-term Memory Agent is a chatbot with long-term memory that supports knowledge bases. It can extract information, preferences, and plans from chat history, vectorize them into memories, and store them in OpenSearch. You can also create Documentation based on your worldview and upload it as files (knowledge bases). During chats, relevant file and memory snippets are used as RAG and input to the large model to generate personalized content.

Try it now

On the top navigation bar, select the target region.

tip

Currently, only the China North 2 (Beijing) region supports App Square.

In the left navigation pane, select Application > App Square.

Find Long-term Memory Agent on the App Square page and click Free trial to experience it.

Process architecture

Associated models and cloud products

Model

Related services	Description	Billing instruction
Doubao-pro-32k-241215	Doubao-pro-32k-241215 excels at role-playing and user interaction.	Model service billing
Doubao-1-5-thinking-pro-250415	Doubao-1-5-thinking-pro-250415 excels in text reasoning. Used to extract memories from user chat logs
Doubao-embedding-text-240715	Semantic vectorization model for vector retrieval scenarios, used to vectorize knowledge files and memories and store them in a database.
Cloud Search Opensearch instance	Opensearch supports hybrid search of files and vectors to improve the recall rate. The diskann + rabitq optimized version reduces memory usage by 75% while maintaining the same search latency and recall rate.	Cloud Search billing

Cloud Service

Related services	Description	Billing instruction
Cloud Search	Cloud Search (Cloud Search) provides storage for memory and knowledge, and hybrid search of text and vectors.	Cloud Search billing
APIG	API Gateway (APIG) is a highly scalable and available managed cloud gateway service based on cloud-native architecture.	API Gateway billing
veFaaS	Function service (Volcano Engine Function as a Service, veFaaS) is an event-driven, fully-managed serverless computing platform. It supports the rapid creation and deployment of functions and the on-demand integration of cloud middleware and database products.	Function service billing
VPC	Virtual Private Cloud (VPC) creates an isolated and self-configured virtual network environment for your cloud resources. Customize network features such as IP address ranges, security groups, and routing policies in your Virtual Private Cloud to manage cloud resources efficiently and securely.	Free
TLS (Optional)	Torch Log Service (Torch Log Service, TLS) provides a one-stop service for log data.	Torch Log Service billing
Cloud Monitor (Optional)	Cloud Monitor supports viewing basic function Monitoring.	Cloud Monitor billing
APMPlus (Optional)	If you need application monitoring services, you can enable the Application Performance Monitoring Full-Link Edition (APMPlus). Once enabled, your function's monitoring data is sent to the APMPlus monitoring service, providing integrated monitoring that includes metrics, traces, and logs.	Billing for Application Performance Monitoring Full-link Edition

Technical implementation

Web UI: handles user interaction, including chatting, uploading knowledge base files, and extracting chat memories.

Long-term Memory Agent backend core components:

File module: supports file upload, chunking, and vectorization.

Chat module: memory retrieval, memory writing, chat content extraction, and personalized chat content generation.

The Long-term Memory Agent and Web UI are deployed in veFaaS functions, providing task requests and result outputs through API Gateway triggers.

Last updated: 2026.07.09 17:20:52