Personal Assistant with Memory

Objectives

• Use at least two different Memory types (e.g., ConversationBufferMemory and ConversationEntityMemory) and be able to explain the appropriate use case for each
• The assistant should correctly recall the user's name and at least two personal preferences shared within the same session
• Implement file-based memory persistence so that after restarting the program, the assistant can still read and use user information saved from the previous session
• Design a personalized System Prompt for the assistant that enables it to proactively incorporate known user information into its responses
• Support multi-turn conversations via the command line or a simple UI, and print the current memory contents after each turn for verification

Hints

• Start by setting `verbose=True` to observe the full prompt that LangChain actually sends to the LLM on each call — you'll get a clear picture of how memory is being injected, which makes understanding and debugging much easier.
• ConversationEntityMemory has an internal `entity_store` dictionary. You can print `memory.entity_store` directly to see what entities the assistant currently remembers, making it easy to verify that things are working correctly.
• When implementing persistence, note that `ConversationBufferMemory` allows you to serialize conversation history via the `chat_memory.messages` list. When reloading, restore messages one by one using `memory.chat_memory.add_user_message()` and `add_ai_message()`.
• If you run into token limit errors, consider switching to `ConversationSummaryBufferMemory`, which automatically compresses older history into a summary once it exceeds a specified token count.
• To give your assistant a more distinct personality, use `ChatPromptTemplate` together with `MessagesPlaceholder` to craft a custom System Prompt that includes memory variables, rather than relying on the default template.

Resources

• python.langchain.com/docs/modules/memory/ - LangChain Memory 官方文档，涵盖所有内置记忆类型的用法
• python.langchain.com/docs/modules/memory/types/entity_summary_memory - ConversationEntityMemory 详细指南，本挑战核心组件
• python.langchain.com/docs/modules/memory/how_to/custom_memory - 如何自定义记忆类以满足特定业务需求
• python.langchain.com/docs/modules/chains/foundational/conversation - ConversationChain 与 Memory 集成的完整示例

Comments (0)